Method development for capturing drivers posture Method development in automotive ergonomics Master’s thesis in Product Development Petter Björsell, Johan Ramberg Department of Product and Production Development CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2016 Master’s thesis 2016 Method development for capturing drivers posture Method development in automotive ergonomics PETTER BJÖRSELL, JOHAN RAMBERG Department of Product and Production Development Division of Product Development Chalmers University of Technology Gothenburg, Sweden 2016 Method development for capturing drivers posture Method development in automotive ergonomics PETTER BJÖRSELL, JOHAN RAMBERG © PETTER BJÖRSELL, JOHAN RAMBERG, 2016. Supervisor: Pernilla Nurbo, Magnus Jerksjö, Volvo Car Cooperation Examiner: Gauti Asbjörnsson, Department of Product and Production Develop- ment Master’s Thesis 2016 Department of Product and Production Development Division of Product Development Chalmers University of Technology SE-412 96 Gothenburg Telephone +46 31 772 1000 Cover: Mediating tool used in workshop 2. Typeset in LATEX Printed by Repro Service Chalmers Gothenburg, Sweden 2016 iv Excitement ["Something that arouses a strong response from another"] -Thesaurus, excitement Method development for capturing drivers posture. Method development in automotive ergonomics Master thesis in product develop- ment. PETTER BJÖRSELL, JOHAN RAMBERG Department of Product and Production Development Chalmers University of Technology Abstract The vehicle industry is a high competitive market with shifting trends and the pres- sure on its actors is to constantly bring something new and exiting to the market. There is a need of shorten lead times, reducing cost connected to physical mock-ups and a possibility to evaluate the product with it’s users during a early stage in the development process. Volvo Cars need to collect quantitative data regarding seating postures in Volvo car seats. An identification of suitable technology for gathering data will be made, and a method will be developed based on this technology. This data will later be used in CAE tools for product development. The method will act as a add-on to earlier methods developed at the Volvo cars ergonomics department. This thesis is divided into two separate parts, one product development part which is the qualitative research. The second part is about collecting quantitative data of seating posture and developing the method. In the product development stage of the project a prestudy was conducted where the feasibility of the project was evaluated and stakeholder needs were collected. As a final activity in the development stage different tools for measuring the drivers joint angles and posture was evaluated. The technologies that have been examined and evaluated ranges from both analog to digital techniques, such as motion capturing, 3d scanning, and image processing. The final technology chosen for further devel- opment are using IMU-sensors to measure the relative motion of the test subjects limbs and predicts the joint angles. In the validation stage the accuracy of the tool was further examined and performance enhancement was also executed in form of design of experiment. The developed method was used to measure and collect data of 45 test persons in two different cars (a Volvo V40 and a Volvo XC60). To give an indication of how the posture differs between persons with different stature the sample size was divided into three groups with 14 short, 16 average and and 15 tall test persons in each group Keywords: Product development, Method development, Drivers posture, Joint an- gles, Motion Capture, Inertial sensors, Automotive, Ergonomics. vi Acknowledgements We would like to thank family and friends, for their support during the whole project. We would also like to thank our supervisor at Chalmers, for the patience and encour- agement that helped us achieve the final goal. A special thanks to our supervisors at Volvo cars for their continued support during the whole process, it has been a pleasure. Lastly we would like to thank all the people at the ergonomics depart- ment at Volvo cars and at the institution Production and product development at Chalmers. Petter Björsell, Johan Ramberg, Gothenburg, June 2016 viii x Contents List of Figures xiii List of Tables xvii 1 Introduction 1 1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Project description . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.4 Prestudy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2 Concept description 9 2.1 MVN Awinda . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.2 FARO Freestyle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.3 Microsoft Kinect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 2.4 Qualisys Miqus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.5 Goniometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 2.6 Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 3 Concept selection 21 3.1 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 3.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 4 Concept Development and Verification 27 4.1 Introduction to the problem . . . . . . . . . . . . . . . . . . . . . . . 27 4.2 Image verification of the system . . . . . . . . . . . . . . . . . . . . . 31 4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 5 Method development and validation 37 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 5.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 5.3 Developing the method capture drivers posture . . . . . . . . . . . . 38 5.4 Performing the method capture drivers posture . . . . . . . . . . . . 41 5.5 Result and validation from the questionnaire . . . . . . . . . . . . . . 41 5.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 6 Verification of the method 47 xi Contents 6.1 Measurements made in the V40 . . . . . . . . . . . . . . . . . . . . . 49 6.2 Measurements made in the XC60 . . . . . . . . . . . . . . . . . . . . 53 6.3 Comparison between the posture in the V40 and the XC60 . . . . . . 55 6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 7 Further development, recommendations, final result and final dis- cussion 59 7.1 Further development . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 7.2 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 7.3 Final Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 7.4 Final discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 Bibliography 67 A The prestudy I A.1 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I A.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III B Time plan for the project method development of capture drivers posture IX C Initial time plan for the project method development of capture drivers posture XI D Complete customers needs list XIII E Concept screening XV F Concept Scoring XVII G Result from the MVN Awinda experiment XXI H Verification XXIII H.1 Subjective verification . . . . . . . . . . . . . . . . . . . . . . . . . . XXIII H.2 ImageJ verification . . . . . . . . . . . . . . . . . . . . . . . . . . . . XXVI I The questionnaire XXXVII J Summary of the questionnaire from drivers posture clinics XXXIX xii List of Figures 1.1 Illustration of the modified product development process (Ulrich and Eppinger 2012) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2 Revised power/interest grid . . . . . . . . . . . . . . . . . . . . . . . 6 1.3 The needs cards used in workshop 1 . . . . . . . . . . . . . . . . . . . 7 1.4 Mediating tool used in workshop 2 . . . . . . . . . . . . . . . . . . . 8 2.1 Concept sketch of the MVN Awinda system . . . . . . . . . . . . . . 9 2.2 Concept sketch of the FARO Freestyle . . . . . . . . . . . . . . . . . 11 2.3 Concept sketch of the Microsoft Kinect . . . . . . . . . . . . . . . . . 12 2.4 Concept sketch of the Qualisys Miqus . . . . . . . . . . . . . . . . . . 13 2.5 Concept sketch of a goinometer . . . . . . . . . . . . . . . . . . . . . 15 2.6 A manual goniometer. (Baseline Plastic goniometers) . . . . . . . . . 15 2.7 Motion capture, with Microsoft Kinect as captured by Motion builder from Autodesk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 2.8 Motion capture, with MVN Awinda captured by MVN Studio . . . . 17 2.9 Test subject with the MVN Awinda equipped. . . . . . . . . . . . . . 17 2.10 Motion capture, with Qualisys Oqus and captured by QTM . . . . . 18 2.11 Test subject with passive markers attached for the Qualisys system . 18 2.12 Mock-up of the Miqus camera placed in a Volvo XC90, to evaluate space and FOV. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.13 Mock-up of a goinometer . . . . . . . . . . . . . . . . . . . . . . . . . 19 4.1 Illustration of a drifting system . . . . . . . . . . . . . . . . . . . . . 28 4.2 Illustartion of a design matrix for a full factorial design experiment inspired by (Bergman and Klefsjö 2010) . . . . . . . . . . . . . . . . 29 4.3 Illustartion of the design matrix as a cube inspired by (Bergman and Klefsjö 2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 4.4 The MVN mannequin placed on top of a picture of the test subject, for verification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 4.5 Measurement done in the MVN model on the test subjects ankle. . . 33 4.6 Measurement done in the picture of the test subject. . . . . . . . . . 33 5.1 The four stages in the method capture drivers posture . . . . . . . . . 39 5.2 Interaction between the seat position study and drivers posture . . . 42 5.3 The rating scale utilized for the questionnaire in the method capture drivers posture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 xiii List of Figures 6.1 N-pose with coordinate systems and arm with angle indicator . . . . 48 6.2 Short test subject positioned in the V40 . . . . . . . . . . . . . . . . 49 6.3 Short heavy test subject positioned in the V40 . . . . . . . . . . . . . 49 6.4 Mid range test subject positioned in the V40 . . . . . . . . . . . . . . 50 6.5 Tall test subject positioned in the V40 . . . . . . . . . . . . . . . . . 50 6.6 Graph over the joint angles of the right knee, measured in the V40 . . 51 6.7 Graph over the joint angle of the right hip, measured in the V40 . . . 51 6.8 Graph over the joint angle in the right shoulder . . . . . . . . . . . . 52 6.9 Short test subject positioned in the XC60 and short heavy test subject positioned in the XC60 . . . . . . . . . . . . . . . . . . . . . . . . . . 53 6.10 Mid range test subject positioned in the XC60 . . . . . . . . . . . . . 53 6.11 Tall test subject positioned in the XC60 . . . . . . . . . . . . . . . . 54 6.12 Graph over the joint angle of the right knee, measured in the XC60 . 54 6.13 Graph over the joint angle of the right hip, measured in the XC60 . . 55 6.14 Graph over the joint angle of the right shoulder . . . . . . . . . . . . 55 6.15 The knee joint angle plotted for the XC60 compared to the V40 . . . 56 6.16 The hip joint angle plotted for the XC60 compared to the V40 . . . . 56 6.17 The shoulder joint angle plotted for the XC60 compared to the V40 . 57 6.18 The joint angle of the ankle plotted for the XC60 compared to the V40 58 7.1 Summary of the concept scoring . . . . . . . . . . . . . . . . . . . . . 61 7.2 Average score from the questionnaire . . . . . . . . . . . . . . . . . . 62 A.1 The Input-Output model . . . . . . . . . . . . . . . . . . . . . . . . . II A.2 The mediating tool that was used during the interviews . . . . . . . . V B.1 Gantt chart for the project method development of capture drivers posture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X C.1 Initial gantt chart for the project method development of capture drivers posture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XII D.1 Complete customers needs list . . . . . . . . . . . . . . . . . . . . . . XIV E.1 Concept Screening . . . . . . . . . . . . . . . . . . . . . . . . . . . . XVI F.1 Scoring against the must needs . . . . . . . . . . . . . . . . . . . . . XVIII F.2 Scoring against the should needs . . . . . . . . . . . . . . . . . . . . . XIX F.3 Scoring against the would needs . . . . . . . . . . . . . . . . . . . . . XX G.1 Result from the MVN Awinda experiment in a V40 . . . . . . . . . . XXII G.2 Result from the MVN Awinda experiment in a V60 . . . . . . . . . . XXII G.3 Result from the MVN Awinda experiment in a XC90 . . . . . . . . . XXII H.1 Test subject positioned in car, with the MVN model layed over for comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XXIII H.2 Test subject positioned in car, with the MVN model layed over for comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XXIV xiv List of Figures H.3 Test subject positioned in car, with the MVN model layed over for comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XXV H.4 Test subject positioned in car, with the MVN model layed over for comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XXVI H.5 Verification picture from imageJ . . . . . . . . . . . . . . . . . . . . . XXIX H.6 The angle of the arm, measured in the MVN model . . . . . . . . . . XXIX H.7 The angle of the arm, measured in the verification picture . . . . . . XXX H.8 The angle of the hip, measured in the MVN model . . . . . . . . . . XXX H.9 The angle of the hip, measured in the verification picture . . . . . . . XXXI H.10 The angle of the knee, measured in the verification picture . . . . . . XXXI H.11 The angle of the knee, measured in the MVN model . . . . . . . . . . XXXII H.12 The angle of the elbow, measured in the MVN model . . . . . . . . . XXXII H.13 The angle of the knee, measured in the MVN model . . . . . . . . . . XXXIII H.14 The angle of the right knee, measured in the verification picture . . . XXXIII H.15 The angle of the right shoulder, measured in the MVN model . . . . XXXIV H.16 The angle of the right shoulder, measured in the verification picture . XXXIV H.17 The angle of the lower spine, measured in the MVN model . . . . . . XXXV H.18 The angle of the lower spine, measured in the verification picture . . XXXV I.1 The questionnaire used in the clinics . . . . . . . . . . . . . . . . . . XXXVIII J.1 Summary of questionnaire from short test subjects . . . . . . . . . . XL J.2 Summary of questionnaire from normal test subjects . . . . . . . . . XLI J.3 Summary of questionnaire from long test subjects . . . . . . . . . . . XLII xv List of Figures xvi List of Tables 4.1 High and low values for the factors in the MVN Awinda experiment . 30 4.2 length of the verification subjects . . . . . . . . . . . . . . . . . . . . 32 4.3 Table showing the minimum, maximum, and mean measured errors and their percentage in the motion range identified by Hanson, Sper- ling, and Akselsson (2006) and Barter, Emanuel, and Truett (1957) . 34 5.1 Summary of the score for questions regarding evaluation of the method 42 5.2 Summary of the score for questions regarding perception of the MVN Awinda system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 xvii List of Tables xviii Glossary CAD Computer Aided Design. CAE Computer Aided Engineering. CAN Controller Area Network. DHM Digital Human Modeling. FOV Field of view. IMU Inertial Measurement Unit. VR Virtual Reality. xix Glossary xx 1 Introduction This chapter work as an introduction to the project method development of capture drivers posture. This chapter contains a project background, method description, aim, limitations and a summary of the prestudy which lays the foundation for this thesis. 1.1 Background The background to the project method development of capture drivers posture ori- gin’s in how today’s manikin systems represent a human in a Volvo vehicle. The Digital Human Modeling (DHM) system used at Volvo cars today is a task driven human model, which means that certain actions must be defined from a user to the model and the system calculates the most probable posture. The posture is calculated with help from an anthropometric database, with respect to the environ- mental geometry and the defined task. (Rothaug 2000) The problem that Volvo cars has with the current DHM-system is that the po- sitioning of the manikin is not adapted to the seats in Volvos vehicles. This means that the positioning becomes too general, the seating position is not adapted to a Volvo vehicle. The task is to explore possible technologies and develop a method for Volvo cars to capture the drivers posture during drive. The new data collection method is going to provide the drivers posture so it can be used to update the current DHM-system. The method are thought to be a compliment to an existing method used by the ergonomics department at Volvo cars. The project has been carried out in two steps like a typical product development project. During the first step possible stakehold- ers, technologies and prerequisites are identified, to create a solid point of departure before launching the big project. In the second stage of the project a method is developed with selected technology and clinics are performed to capture data. 1 1. Introduction 1.2 Project description To be able to motivate and clarify the project among the stakeholders of method developing for capturing drivers posture, a clear aim of what the project want to achieve is crucial. The aim of a project can roughly be compared with the impact that the project should have and why it’s carried out. To realise the aim of the project it needs to be broken down in smaller pieces and formulated in a way that everyone involved in the project knows what should be delivered, this is called the objectives. When formulating the objective emphasis should be on formulate the objective as clear and simple as possible, keep it real, measurable and rootedness. To evaluate your objectives to see if they support the project a simple S.M.A.R.T-test is good to asses. Each letter in the word S.M.A.R.T stands for different criteria that the objective have to fulfill in order to fit as an objective. The letters stands for Specific, Measurable, Accepted, Realizable and Time specified. (Tonnquist 2012) 1.2.1 Aim The aim for this project is divided into two parts • Identify a suitable tool for capturing of joint angles and drivers posture and develop a method that easily can be incorporated in to the ergonomic depart- ments work flow. • Update the digital human models with new values for their joint angles and new posture in order to improve the virtual prediction model for driver position 1.2.2 Objective The objectives that are connected to the two aims presented in the previous section • Identify critical parameters which affects the driver posture in the seat. • Identify and verify a suitable technology as a data collection tool for the af- fecting parameters. • Develop a data collection method for collecting the affected parameters that works as a supplement to current method at the ergonomics department at Volvo cars. • Perform clinics using a wide range of test subjects to collect subjective and objective data for a selection of different vehicle types. 1.2.3 Limitations During the project certain limitations had to be applied due to respect to time, budget and equipment. The limitations this thesis had to take into account are listed below • Only the drivers posture will be studied, no passengers at all • Only two vehicles will be applied to the tests one "low" vehicle and one "high" vehicle 2 1. Introduction 1.3 Method This project will be carried out as a product development project. It will consist of a qualitative and quantitative data collection part. The qualitative data collection stage will take place in the prestudy, during the customer needs collection phase. The core stakeholder needs will be captured with two interviews and verified with two workshops. The quantitative data collection stage will be executed in the last stage of the project when a suitable method has been developed. 1.3.1 Process The product development process that will be used to structure the work in this the- sis origin from the generic product development process by Ulrich and Eppinger in the book product design and development (2012). The generic product development process consists of six different phases and each phase have several predefined activ- ities linked to each phase (Ulrich and Eppinger 2012). For this project the generic product development process will be modified and consist of two phases and four stages. The two phases are divide into a method development phase, a verification, and data collection phase. As for the four stages utilized for this project there are planning, concept development, testing and verification of the method and last data collection and for last and result. Picture 1.1 describes the product development process. Figure 1.1: Illustration of the modified product development process (Ulrich and Eppinger 2012) 1.3.2 Report structure The structure this report will follow is that in each chapter all used theory will be presented followed with the application of it and result and discussion will be presented at last. 3 1. Introduction 1.3.3 Time plan In the beginning of this project a time plan was established. The time plan describes in depth which activities that should be carried out and when. The revised time plan can be found in appendix B.1. 4 1. Introduction 1.4 Prestudy To initiate this project a prestudy was conducted. In the prestudy a assessment of the stakeholders needs were made, customer needs were collected and several workshops has been carried out. 1.4.1 Stakeholders One of the first activities performed in the project was a stakeholder assessment, to find possible stakeholders who are affected or affects the outcome of the project. The complete stakeholder identification process can be found in appendix A. In this section the essence of the assessment will be presented. Firstly the project team identified and classified possible stakeholders, this was done with brainstorming and post-it notes. The second step in the stakeholder identification process was to classify them by interest and power. This classification helped the project team to identify the most important stakeholders to focus on throughout the project. 1.4.1.1 Newly identified stakeholders During the course of the project a handful of new stakeholders occurred. These stakeholders were not identified through the first assessment and appeared during the course of the project. The new stakeholders that the project team have came across is Qualisys, a motion capture company located in Gothenburg. A new branch at Chalmers university of technology, that are interested in the technology that are used in the project. During the project the team has aided the school with technical expertise and also contributed with demonstration of equipment and education for staff members. As the project went further, one of the previously core stakeholders IMMA showed a lot of interest for the technology used, but their lack of power in this project still makes them a second tier stakeholder. Figure 1.2 illustrates the revised power/interest grid constructed from the prestudy A. 5 1. Introduction Figure 1.2: Revised power/interest grid 1.4.2 Needs During the prestudy the customer needs for the method to be developed were col- lected, the needs collection is part of the qualitative research of this thesis. The needs collecting process was divided into two parts, one part where the project team identified needs based upon the stakeholder statements appendix A and the second part derives from statements collected in two in depth interviews made with two of the key stakeholders. These two interviews were transcribed and statements were extracted from which became statements from interviews appendix A. The second step in the needs collection process was to categorise the collected state- ments; both the statements from the stakeholders but also the statements collected in the interviews. As a tool for this the project team utilized a affinity diagram ap- pendix A. When redundant statements had been eliminated and the rest categorised the project team created the needs following Ulrich and Eppingers guidelines ap- pendix A. In the last step the project team divided the needs into three subgroups appendix A. The final list of customer needs divided into three levels can be found in appendixD.1. 6 1. Introduction 1.4.3 The workshops During the prestudy two workshops were conducted with the people who has sev- eral years of experience in the field of vehicle ergonomics. The purpose with the first workshop was to verify and rate the customer needs identified by the project team. This was done since this is the first project in the field of ergonomics that the project team undertakes and they want to be sure of the robustness of the needs. As a second part of the workshop the participants were given coloured dots so they could rate some needs that they thought were of extra importance. Full description and result from workshop one can be found in appendix A. The needs cards used in workshop 1 can be seen in figure 1.3. As for the second workshop the project team wanted answers on which parameters that were of importance and as an extra step give the participants the possibility to eliminate one of the needs form previous workshop. The tools used for workshop two was four laminated pictures of a seated manikin in a drivers seat and a front view. On these pictures the participants could mark which areas that were of importance for them and give the project team a hint of which parameters the method had to collect. Full description and result from workshop two can be found in appendix A. The mediating tool used in workshop 2 can be seen in figure 1.4. Figure 1.3: The needs cards used in workshop 1 7 1. Introduction Figure 1.4: Mediating tool used in workshop 2 8 2 Concept description In this chapter a presentation of suitable concepts identified as possible data collec- tion tool will be made. 2.1 MVN Awinda MVN Awinda is a motion capture system developed by a company called Xsens, with headquarters located in Enschede in the Netherlands. Xsens are focused on inertial sensor modules, human motion measurements and 3D character animation. (Xsens) Figure 2.1: Concept sketch of the MVN Awinda system 2.1.1 Hardware The MVN Awinda system consists of 17 inertial sensors which are attached to the body with Velcro straps. 16 of the sensors are attached to different body parts and the last one can be used as a prop, such as a golf club or sword that is captured by the program. This prop have to be attached to a body segment, so for example something that is put in the hand of the test subject will always be connected to the hand of the subject, even if it is put down. This is because the sensors only measure the relative movement of the sensors after the calibration, not the absolute position. The sensors in the MVN Awinda are all completely wireless, but there are another variation, MVN Link, that uses smaller sensors that are daisy-chained connected to a sender pack located on the back of the test subject. This system can be worn with Velcro straps like the Awinda or with a Lycra suit that the sensors are fastened at. The Lycra suit have dedicated zip-fasteners that the cables and sensors can be fastened in. (MVN user manual, user guide MVN, MVN BIOMECH, MVN Link, MVN Awinda 2015) 9 2. Concept description The inertial sensors consists of four different sensors that determine their position; a gyrometer to measure angular speed (orientation), a magnetometer that measures the magnetic field of the earth (reduces drift in the x-, y-plane), an accelerometer that measures the 3D-acceleration (position), and also to define the normal of the earths curvature(z-axis, to reduce angular drift)(Roetenberg, Luinge, and Slycke 2013), and a barometer to measure the atmospheric pressure. The output from these sensors are combined with a dynamic model, in a Kalman-filter to get the best model for the movement possible.(MVN user manual, user guide MVN, MVN BIOMECH, MVN Link, MVN Awinda 2015) The Kalman filter uses the data that are collected by the sensors, and a theoretical model over how the system ought to behave, to make the best prediction of the test subjects movement possible. (Faragher 2012) 2.1.2 Software The software used with the MVN Awinda is MVN studio. MVN studio defines a bio-mechanical model based on the bio-metrical data that are defined before the calibration phase, and the calibration pose where the position of the sensors are synced with the posture of the test subject. The sensors are then giving the relative movement of each body segment based on the movement of the sensors. (MVN user manual, user guide MVN, MVN BIOMECH, MVN Link, MVN Awinda 2015) 2.1.3 Scenario The Xsens MVN system are currently used for motion capture in animated movies and games, but also in virtual and augumented reality to capture the body move- ments of the person using the system in real-time. (Customer cases for Xsens MVN ) The scenario it will be used in this project is, that the system will be used to mea- sure the joint angles of the test person when they are positioned in the seat. This gives the test persons joint angles directly from the model and no further analysis have to be made, other than calculation of mean joint angles. 10 2. Concept description 2.2 FARO Freestyle FARO Freestyle is a 3D-scanner, developed by FARO that is a 3D measurement technology company. FARO headquarters are located in Lake Mary, Florida. (FARO company profile) Figure 2.2: Concept sketch of the FARO Freestyle 2.2.1 Hardware The FARO Freestyle is a structured light scanner. A structured light scanner sends out a IR-pattern that is captured by a camera, and through the offset between the camera and the IR-projector the distance to the object can be determined through triangulation, and the surface geometry can be determined by measuring distur- bances in the pattern. (Knicker 2014) To give an increased accuracy to the scanner the Freestyle have two IR-cameras that capture the pattern and triangulates the distance to the object. It also have one RGB-camera to capture the color of the object scanned. (Liscio 2015) 2.2.2 Software The data that are processed by the FARO Freestyle are exported to FARO Scene, where it can be processed into a point-cloud and from there be exported to a number of different file formats, that then can be used in various CAD-systems. (Scan & plan with the freestyle3D) 2.2.3 Scenario Today the FARO freestyle is used to document objects and rooms in 3D point-clouds (FARO Freestyle 3D scanner - Efficient handheld 3D laser scanning). The way it would be used in this project is that the test person will be documented with the FARO freestyle when sitting in the drivers seat, the files would then be analysed to determine the joint angles of the test person. 11 2. Concept description 2.3 Microsoft Kinect Microsoft Kinect uses a infrared camera and sensors to track the motions of the person using it. It is a motion capture camera developed primarily for the gaming industry. (Weinberg 2015) Figure 2.3: Concept sketch of the Microsoft Kinect 2.3.1 Hardware The Microsoft Kinect is a sensor developed for the gaming platform Xbox One. It uses a similar technique as the FARO Freestyle, a time of flight scanner, which sends out a IR-pattern that is tracked with a monochrome camera and gives information on the distance to objects in front of the Kinect (Choppin and Wheat 2013). The hardware is composed of a RGB-camera, a IR depth sensor, and a Multi-array microphone. (Crawford 2010) 2.3.2 Software Other companies have developed third-party software for recording using multiple Kinect sensors to increase the accuracy and the field of vision, such as iPi recorder (Motion capture for the masses). iPi also have software for transforming the record- ings into motion-capture files, and from them get the joint angles of the test subject joint angles (iPi MoCap Studio 2016). 2.3.3 Scenario Microsoft Kinect is mainly used in the gaming industry, to capture motions to control a game. But it could also be used for motion capture, for a much lower price than the conventional motion capture technologies. (Motion capture for the masses) In this project a number of Microsoft Kinects have to be used simultaneous to capture the joint angles. 12 2. Concept description 2.4 Qualisys Miqus Miqus is developed by Qualisys, a motion capture company focused on biomechanical research, sports biomechanics, and medical. Qualisys headquarters are located in Gothenburg Sweden. (We are Qualisys. The Swedish motion capture company) Figure 2.4: Concept sketch of the Qualisys Miqus 2.4.1 Hardware The Miqus is a optical motion capture camera that is using passive markers placed on the subjects body to locate the motions of the subject (Miqus 2015). A optical motion capture system using passive markers, usually consisting of small balls coated with a reflective material that the camera detects. The position of the markers is then decided by triangulation from the camera in software, this means that multiple cameras have to be used. (Kirk, O‘Brian, and Forsyth 2005) The Miqus is available in two models, the miqus M1 and the Miqus M2. The Miqus M1 has a sample rate of 250fps, a resolution of 1MP, and a FOV of 58*40°. The Miqus M2 has a sample rate of 340fps, a resolution of 2MP and a FOV of 61*39°. The dimensions for one camera is 14.0*8.7*8.4cm. (Miqus 2015) 2.4.2 Software Qualisys has developed their own software for capturing the data that the Miqus receives, called Qualisys Track Manager (QTM). QTM combines the data that are captured from the cameras and place the markers in a predefined coordinate system. To measure movement in 3D, the cameras have to be calibrated for the area that the motion capture will be performed in. This is done by moving a wand with markers attached, in the space the test will be performed in. The data can then be exported into a number of different programs such as MATLAB.(Qyalisys Track Manager - QTM ) 2.4.3 Scenario Qualisys is used in sports analysis to prevent or find injuries, and to improve tech- nique. In engineering it is used to determine position over time. In medical it is used for rehabilitation and in psychology for motion analysis. (We help people and analyse motion) To utilize this concept the cameras have to be mounted inside the 13 2. Concept description car and the test subject would be equipped with markers, to acquire the desired information. 14 2. Concept description 2.5 Goniometer A goniometer is an instrument used to measure the movement in joint angles. Used by physical therapists.(Sears 2016) Figure 2.5: Concept sketch of a goinometer 2.5.1 Hardware The goniometer consists of two arms, that are positioned at specific points of the body, while the center of the goniometer are placed over the center of the joint being measured. Compared to all other concepts this is a very cheap method. Figure 2.6: A manual goniometer. (Baseline Plastic goniometers) 2.5.2 Scenario A goniometer are used by physiotherapists to measure the range of motion for joints, from one maximum angle to the other(Measuring flexibility using a goniometer). Using the goniometer all the test subjects joints have to be measured manually, while they are seated in the driving seat. This would lead to quite inaccurate result due to that all joint angles are not measured at the same time. 15 2. Concept description 2.6 Testing To acquire knowledge about how well the different concepts would perform in the measuring environment (inside a car) this testing was initialised early in the develop- ment process. This testing were done to acquire knowledge about how the concepts worked and not to do a detailed verification of the performance of the different con- cepts. The technology behind every concept was tested first. The Microsoft Kinect was used to record a test subject performing a reference pose and then performing a set of predefined movements to evaluate how well the motion tracking followed the movement of the test subject. Figure 2.7: Motion capture, with Microsoft Kinect as captured by Motion builder from Autodesk The MVN Awinda and the Qualisys system were tested in a similar way. With the test subject performing necessary calibrations and then performing a set of predefined movements. The movements was compared to the movements that was captured in the software in real time. The test of the MVN Awinda was conducted at VCC, with a Awinda system bor- rowed from the University of Skövde. The Awinda was first tested in a lab, with a test subject walking and performing movements that was given from the test leader. The system was then tested in a stationary vehicle, to evaluate how it would work in the real environment. 16 2. Concept description Figure 2.8: Motion capture, with MVN Awinda captured by MVN Studio Figure 2.9: Test subject with the MVN Awinda equipped. The test of the Qualisys system was performed at Qualisys facilities in Gothenburg, where a time slot was available for verification. During these tests the test subject was equipped with 23 markers, some markers were also attached to a chair that was used during the session. The test subject was seated in a chair and recorded while doing a number of predefined moves that would simulate driving a car, such as moving the steering wheel, breaking, and changing gear. The Qualisys lab did not have the Miqus cameras installed in the test rig, so instead the verification was done with another of their models, the Oqus. The Oqus have the same basic function as the Miqus, but a higher sample rate and resolution(Oqus 2015). The verification was performed primarily to get a overall view of how the technology and how the software works, not the specific performance of the cameras. 17 2. Concept description Figure 2.10: Motion capture, with Qualisys Oqus and captured by QTM Figure 2.11: Test subject with passive markers attached for the Qualisys system Mock-ups where then created to evaluate how the Miqus camera could be placed in the car and how far away from the drivers seat they could be placed to maximize the FOV. A rough mock-up was made of a goinometer to get a rough understanding of how the measurements would have to be performed in the testing environment. 18 2. Concept description Figure 2.12: Mock-up of the Miqus camera placed in a Volvo XC90, to evaluate space and FOV. Figure 2.13: Mock-up of a goinometer 19 2. Concept description 20 3 Concept selection In this chapter theory, method and result from the concept selection phase will be presented. 3.1 Theory Utilized theory during concept screening and scoring. 3.1.1 Concept screening A good way to structure and keep control during an evaluation process is to asses a matrix structure. The matrix is constructed with selection criteria on the y-axis and the concepts for evaluation on the x-axis. By applying a matrix during this stage of the evaluation process will not only give a systematical work flow it will also provide a rational way of rating and discarding low performing concepts. Another benefit with applying a matrix is that it hinders other project members to favor their own ideas or concepts over the other ones. To create the criteria of evaluation a good way is brainstorming. But be aware when brainstorming the criteria of evaluation, they should have some anchoring to the main problem the concepts tries to solve. (Pugh 1991) When the evaluation criteria has been agreed upon and written down the creation of the matrix begins and evaluation of each concept against each criterion. At the beginning of the evaluation a reference concept is chosen, this is the one all other concepts will be compared against. Then the other concepts are evaluated against each criterion and the reference concept. For example if the criteria is cost and concept A is more expensive then the reference it will get a "-" but if it is cheaper it gets a "+" or cost as much as the reference then it will receive a "0". After all the concepts has been evaluated against each criterion a final score is established. The final score is calculated as the difference between the +’s and -’s and will be in a numerical value. The final score helps too determine which concepts to keep, discard or combine. (Ulrich and Eppinger 2012) 3.1.2 Concept scoring To narrow down the selection process and distinguish the remaining concepts from each other a weighting and rating matrix can be utilized. The matrix is constructed 21 3. Concept selection with concepts to be rated on the x-axis and weighted criterion’s on the y-axis. The project team asses each selection criteria with a relative importance in order to assign each concept a score usually between 1-5 or 1-10 on how well they fulfill the criteria. Then the given score is multiplied with the weighting of the criteria and provides a total score. A benefit with weighted criteria is that it helps to rate subjective criteria such as, perceived quality or environmental friendliness. When working with numerical rating of criteria it is always subjected to the judgement and experience of the team members, so another use full way can be to divide the criterion’s into three categories, high medium or low.(Pugh 1991) When selecting the criteria for the selection matrix a good idea is to utilize the customer needs list, to increase the resolution of the scoring secondary and tertiary needs could be used. As suggested by Ulrich and Eppinger a reference concept should be chosen. They also recommend to choose the scale between 1-5 were 3 is same as the reference 1-2 is worse and 4-5 is better, this is done to make the eval- uation process less time consuming. As for weighting the criterion’s suggestions to use a rating similar to the one proposed in the section above or divide 100 percent among selected criteria. The weighting can be determined by the project team or based upon collected customer needs. (Ulrich and Eppinger 2012) 3.2 Method How the theory are applied for concept screening and concept scoring. 3.2.1 Concept screening The first step in concept screening for the project were to choose which concepts to rate against each other. The project team choose the five original concepts which were evaluated during the concept development phase and added a minor change to one of them, an extra sensor was added to the Kinect concept. This concept is only theoretical and has not been evaluated during the concept development phase, but the main idea behind it is that the accuracy of the system will increase. The project team brought a total of six concepts for screening against a reference concept (Ulrich and Eppinger 2012). The reference concept, the goniometer were chosen because it is the only system that is known, out of the concepts chosen, to work in a vehicle. The drawbacks with this concept is that it is very primitive, inaccurate and time consuming to use, plus it is the only concept that is analogue. Determining the selection criteria were done during a brainstorming session. A total of nine individual criterion’s were created. The chosen criterion’s are the ones the project team identified as the best suited ones to cover as big part of the whole project as possible in as few criteria as possible. The main areas in which the criteria have anchoring in usability of the system, cost, and the impact of the system on the vehicle and driver. 22 3. Concept selection The concepts evaluated during the concept development phase are complete sys- tems, the combination option were discarded during the screening. This is due to limitations in time and limited benefits, decision was made to not combine any sys- tems at this point of the project. The result from the screening can be found in appendix E.1. The main outcome is that two concepts were discarded during this activity, the ones using the Kinect sensor. 3.2.2 Concept scoring The project team entered the scoring phase with four concepts, the reference and three other. Except for the reference the other concepts brought to the scoring were MVN Awinda, Faro freestyle and Qualisys Miqus. As for the selection criteria the project team utilized the full customer needs list appendix D.1 divided into three types of needs (Pugh 1991), must, would and should. In the first scoring session the must needs appendix F.1 were applied, in the second scoring the should needs appendix F.2, and in the third scoring the would needs appendixF.3. The most im- portant session was the first, this is because the must needs is the most important and if the concept doesn’t score good here it will have no chance to perform at the minimum level of the requirements. As for the weighting of the criteria the project team utilized the rated customer needs cards from 1.4.3, and used a 100 percent scale for each type of needs. During the first scoring of the concepts the needs in the must section were ap- plied. The project team used a grading scale between 1-5 to rank each concept against the criteria, as a reference the concept goinometer was used and first graded in consensus by the project team (Ulrich and Eppinger 2012). During the first scor- ing with the "must" criterion’s the faro freestyle concept was eliminated and the full result can be found in appendix F.1. As for the second scoring the two remaining concepts were the MVN awinda and the Qualisys Miqus system, the goinometer was kept as reference. The result from the second scoring was in i favor for the MVN awinada concept, full result can be found in appendix F.2. The final scoring was done to increase the resolution but also to give the project team the extra confidence that they made the right decision. The scoring was carried out with the goinome- ter as reference and the two remaining concepts from scoring two were kept, full result from scoring three can be found in appendix F.3. As a result from all three scoring the MVN awinda came out as a winner and this concept went to the next development stage. 3.3 Discussion In this section discussion about how the project team approached and carried out the concept selection will be held. 23 3. Concept selection 3.3.1 Reference concept As suggested by Ulrich and Eppinger (2012) a reference concept should always be chosen and used as a benchmark in the evaluation of the other concepts. The reference concept doesn’t need to be static and can be changed from screening to screening or scoring to scoring. In this project the project team chooses to keep the same concept as reference during both screening and scoring. Why this choice was made to keep the goinometer as reference and only perform one screening was because this was the only concept that the project team knew would give the joint angles and the only concept that was analogue. As for the other concepts the project team knew that they would produce some uncertainties regarding measurement accuracy and before the measurement error was known they were not suitable as references. Another issue was also that the project team didn’t knew if the other concepts would stay stable during a longer period of drive. 3.3.2 Selection criteria Here the selection criteria for both the screening and scoring will be discussed. They will be divided into the different headings, criteria for concept screening and criteria for concept scoring. 3.3.2.1 Criteria for concept screening The selection criteria for the concept screening were created as suggested by Pugh (1991). As mentioned above the project team wanted to keep the selection criteria as few as possible to not make the initial selection phase too complicated. As for the strategy to select a few main areas and anchoring the criteria to the areas was a good approach. Not only did it make the whole screening processes easier but also helped the project team make a just evaluation between the reference concepts and the one subjected for rating. Another benefit with choosing a few main areas to connect the criteria to is that the project team knew what was expected of the final concept. 3.3.2.2 Criteria for concept scoring The selection criteria applied in the concept scoring origins from the customers needs list appendix D.1. By utilizing the needs that the project team collected in the prestudy the validity of choosing the right concept for further development increases. The robustness of the selection criteria has been verified in 1.4.3 and therefore the project team felt confident with the previous statement. The weighting of the criteria was done as proposed by (Ulrich and Eppinger 2012) and by utilizing the result from 1.4.3 the importance of each criteria has been captured correctly. Furthermore a second review of the weighting has been done with the supervisors after the concept selection process in order to assure the validity of the result. 24 3. Concept selection 3.3.3 Concept screening The concept screening activity was used as an elimination step in the concept se- lection process in this project. Why the team didn’t utilize the activity to combine low ranking concepts as suggested by (Ulrich and Eppinger 2012) is because the complexity of the concepts. Since the concepts consist of basically full systems as described in chapter 2 with special hardware and software, integration between them would be difficult. The project team didn’t have the time or know how to do this so the option of combination of concepts was discarded. 3.3.4 Concept scoring The activity concept scoring is one of if not the most important activity performed in this project. This is where the tool for data collection and the back bone in the method to be developed is chosen. To ensure that the scoring was done with as high resolution as possible the project team did as suggested by (Ulrich and Eppinger 2012). By increasing the detail level of the scoring the project team felt comfortable with the result. In the the must needs section 64 percent of the weighting was divided among 6 out of 16 needs. This distribution of the weight put a lot of extra emphasis on these needs, make them extra important since they would decide which concepts that would have made it to the next session. All concepts scored fairly high, of these 6 needs except for two. The most important criteria Give joint angles/segment position where the Faro Freestyle received score of two. The project team felt secure in their grading here since this concept did not have the possibility to provide either individual segments, such as and isolated leg or arm or joint angles. As for the need the concept can be applied during drive, at this need the final concept does not perform as rated in the scoring. After the final concept had been chosen the project team identified a huge problem, the final concept can not be applied during drive. This is due to technical limitations in the hardware that was discovered after the final concept had been selected. Regardless if the winning concept the MWN Awinda received a lower score, even as low as one it would not change the outcome of the first scoring session. If the score at this criterion was lowered to one the final score would become 3,34 which is the same as the second best scoring concept. In the two last scoring sessions the final concept MWN Awinda came out as clear winner every time. The benefit that the MWN Awinda had in the should session origins from that the systems is easier to set up and the data interpretation of out- put data is more manageable to handle. Lastly the project team would like to mention that the only thing that changed from the second to the third session are the criteria and weight. The two concepts were kept the same to help the team to justify that the right concept was chosen 25 3. Concept selection for further development. 26 4 Concept Development and Verification In this chapter the further work with the concept MVN Awinda is described and a verification of the systems robustness is made, through image analysis. 4.1 Introduction to the problem As described in chapter 2 the MWN Awinda concept keeps track of x- and y-axis with the help of magnetometer and the z-axis with the magnetic field of the earth. This turned out to become a problem when the system is applied in an environment with stronger magnetic fields then the earths. This problem was undiscovered by the project team during the concept evaluation but as the teams knowledge grew the problem became apparent. In order to minimize the impact of the magnetic disturbances on the MWN Awinda that is generated from the vehicles chassis and embedded electrical systems a simple factorial design experiment was constructed. 4.1.1 Design of experiment The problem identified in previous section is that the MVN Awinda system looses it’s accuracy and starts to drift, see fig.4.1 when subjected to a magnetic field. To minimize the impact from the magnetic fields in the vehicle the project team established a factorial design experiment. In a factorial design experiment the first step is to identify which factors who con- tributes to the specific problem. According to Bergman and Klefsjö (2010) can a cause and effect diagram be a good approach to identify interesting factors. After a set of factors has been established two levels of a high and a low value are chosen for each factor, when a factor is at a low value it is expressed with a "-" and when high a "+". The number of chosen factors determines how many different tests the experiment will include and is described with the formula 2n, were n is equal to the amount of factors, one testing session is called a run. At each run the different factors are either high or low until all possible combinations has been evaluated, see figure 4.2. When executing design of the experiment the order of the runs should be randomized in order to avoid interference from other factors. From these test it is then possible to calculate the effects from each factor, the effect is how much specific factor affects the final result. The main effect for a factor is calculated as the difference between the result when the factor was high and low. Factors can 27 4. Concept Development and Verification Figure 4.1: Illustration of a drifting system also interact with each other and create an effect, as for example lets take a sick patient. The patient is first treated with penicillin and nothing happens, then the patient gets a chemotherapy and still nothing happens. But with chemotherapy and penicillin in combination the patient get cured, that is the effect of interaction between factors. (Bergman and Klefsjö 2010) Picture 4.3 illustrates a factorial design experiment with the factors A, B and C the first number in the circle represent the run and the yx is the result from that run. To calculate the effect for factor A the equation (y2 −y1) is used. The arithmetic av- erage for factor A is calculated as 1/4((y2 − y1) + (y4 − y3) + (y6 − y5) + (y8 − y7)). As for calculation of interaction effects the design matrix can be utilized see figure 4.2. If another row is added for the interaction between AxB the low respective high levels of this interaction will be factor A’s level multiplied with factors B’s at a certain run. AxB level at run 1 would become a "+" since "-" x "-" = "+" and so on. Depending on if the interaction factor is high or low at the run it will affect the value of the yx. If the interaction between AxB is calculated with the formula above it will look like this 1/4(y1 − y2 − y3 + y4 + y5 − y6 − y7 + y8). (Bergman and Klefsjö 2010) 28 4. Concept Development and Verification Figure 4.2: Illustartion of a design matrix for a full factorial design experiment inspired by (Bergman and Klefsjö 2010) Figure 4.3: Illustartion of the design matrix as a cube inspired by (Bergman and Klefsjö 2010) 4.1.2 The MVN Awinda experiment The identified problem with the MVN Awinda concept is that it will start to drift when exposed to magnetic fields, since the concept have to work during drive the project team tried to identify which factors that could extend the time before the system starts to drift. The first factor sensor warm up was identified from the manual (MVN user manual, user guide MVN, MVN BIOMECH, MVN Link, MVN Awinda 2015), as mentioned in chapter 2 the system applies a Kalman filter in order to calculate the sensors position. By adding the extra warm up time the prediction is that the filters will stay stable for a longer period of time. As for the second factor calibrate outside the car which was identified through brainstorming of the project team, the reasoning was that by placing the base station outside the vehicle interference of magnetic fields would decrease. The last factor number of turns was also identified with brainstorming. The reasoning is that since the system drifts in the x and y direction the amount of turns on a track might make the system to drift faster. The high and low values for each factor is summarized in figure 4.1. 29 4. Concept Development and Verification Table 4.1: High and low values for the factors in the MVN Awinda experiment The experiment was performed in three different vehicle models in order to see if there was any difference regarding type, the tested vehicles were a Volvo V40, V60 and a XC90. All experiments were performed at the same route and when the factor number of turns was high the vehicle drove a course with many roundabouts and one extra lap in each roundabout was done, when the factor was low the vehicle took a course with a straight road. During the test the MVN Awinda system was applied to the driver and the test leader was responsible for keeping track of the system until he noticed a drift in the software. When the test leader noticed the drift the time on a stop watch that was started at the beginning of the run was written down and that is the y-value in this experiment. 4.1.3 Result from the MVN Awinda experiment The main outcome from these three experiments was that the factor that extended the time before the system started to drift in a vehcile was sensor warm up. The main effect from this factor was 67,5 in the XC90, 70,75 in the V60 and 31,5 in the V40. The factor with a negative effect on the time the system could be applied during drive was number of turns were the main effect was -26,5 in the XC90, 23,75 in the V60 and -25 in the V40. The factor calibrate outside the car, had a minor effect in the XC90 and the V60, but a bigger effect in the V40. This could be due to the smaller body of the V40 which shields the signal from the sensors and gives a more unstable system. The complete design matrices can be found in G.1, G.2 and G.3 30 4. Concept Development and Verification 4.2 Image verification of the system Image analysis was performed to evaluate how well the MVN model match the real body posture of the test subject. Two different verifications were made. One verification with the verification subjects (VS) seated in a vehicle, with the Awinda system equipped, to give an estimation of how good the Awinda system represents the posture of a test subject in a vehicle. One other verification was made, with the VS standing in a open room and sitting in a chair. This verification was done to get more exact data on how well the Awinda system represents the posture of the VS. 4.2.1 Subjective verification The first verification was done with five VS. Each VS was equipped with the Awinda, some markers to mark the joints of the verification subject, and then seated in a vehicle. A recording was started in MVN studio, and a picture was taken with a camera from outside the vehicle at the same time as a marker was set in the MVN recording, to assure that the model and the picture are synced in time. A snip was then taken in the MVN studio file from the same viewpoint as the camera had been placed. The snip was then placed over the picture using Microsoft PowerPoint. The location of the hip-joint was used as a reference point for placement, to get a comparison between the real body posture of the VS, and the MVN model. This verification method are inspired by the verification used by Kirk, O‘Brian, and Forsyth (2005). This result was entirely subjective, based on how well the manikin represented the picture of the verification subject. The manikin was seen as a good representation of the overall body posture of the test subject, but some individual joint angles was not seen as good enough. To further investigate this problem the verification in 4.2.2 was done. Figure 4.4: The MVN mannequin placed on top of a picture of the test subject, for verification. 31 4. Concept Development and Verification Five of these verifications was made with five different persons. To assure that size or body posture of the test subject would not affect the result of the verification. These verification pictures can be found in Appendix. H.1 4.2.2 Image measurements The image measurements were performed outside the car, with five verification sub- jects with different body height see 4.2 , to get a more exact result that could be used to evaluate how exact each joint angle are in MVN studio. To do this a picture was taken and a marker was set as in the previous verification. Analysis was then made using a image analysis software. Table 4.2: length of the verification subjects 4.2.2.1 software description For the image analysis used to verify the Awinda system, ImageJ was used. ImageJ is a JAVA based image analysis program (ImageJ, Image processing and analysis in Java). To verify that the measurements in ImageJ are accurate, some measurements was done in pictures were the angles are known, see fig.H.5. 4.2.2.2 Performing the measurements The VS was asked to perform a T-pose and then to sit in a wooden chair, with picture taken and a marker placed in the MVN Studio recording, at each step, to sync the recording with the picture. Measurements was then performed using ImageJ, as can be seen in fig.4.5 and fig.4.6. The measurement error between these figure are then calculated. The full table of measurements and figures can be found in appendix. H.2 To assure that the picture was taken at the same angle as the viewpoint in MVN studio the origin was set at the point of the camera. This was achieved by performing the calibration at a marked point, which place the origin of the coordinate-system, and then placing the camera at that point. 4.2.2.3 Analysis of the verification The result from this second verification was compared with the range of joint angles, measured in a car identified by Hanson, Sperling, and Akselsson (2006) for the sitting position and Barter, Emanuel, and Truett (1957) for the standing T-pose, to get an 32 4. Concept Development and Verification Figure 4.5: Measurement done in the MVN model on the test subjects ankle. Figure 4.6: Measurement done in the picture of the test subject. indication of how much the error would affect the result of the measurements. As can be seen in table 4.3 the error is much larger in the seated position, than in the standing position. 33 4. Concept Development and Verification Table 4.3: Table showing the minimum, maximum, and mean measured errors and their percentage in the motion range identified by Hanson, Sperling, and Akselsson (2006) and Barter, Emanuel, and Truett (1957) 4.3 Discussion In this section discussion about the MVN Awinda experiment will be held but also for the image verification of the accuracy of the MVN Awinda. 4.3.1 Selecting the factors As for the identification of the factors for the MVN Awinda experiment the project team consulted the manual provided with the system. The manual provided infor- mation about if the system feels unstable a longer calibration time might solve the problem, from this information the first factor was identified. It is known that the systems accuracy starts to drift when exposed to magnetic fields the project team started to think of what is creating these fields in and around a vehicle. The first thought was all cables and electrical systems but also the body of the vehicle. After some discussion the factor calibrate outside the car was discovered. As mentioned by Bergman and Klefsjö (2010) a good way to identify factors is to apply a cause and effect diagram, this was not done by the project team. Instead brainstorming and discussion about the matter was done. 4.3.2 Factors with high and low effect The factor that had the highest positive effect in all three experiments were Sensor warm up. The result was expected since this factor origins from the (MVN user manual, user guide MVN, MVN BIOMECH, MVN Link, MVN Awinda 2015) and it advices the user to take some extra time, standing still after the calibration, in order to give the Kalman filters a chance to collect good data in order to stay stable for a longer time. Why the effect in a V40 is almost 50% lower then in the XC90 and V60 might depend on the size of the cars. Since the driver has less space in the 34 4. Concept Development and Verification V40 and the test subject might be subjected to the magnetic field from the electrical wiring inside the vehicle easier then in the other models. As a result from less space in the drivers seat and closer contact with the magnetic fields a lower effect for the factor sensor warm up is obtained from the V40. The last factor Calibrate outside the car was only important in the V40 where it had a positive effect, in the other two vehicles the factor did not have an noticeable effect. This could be due to the smaller body of the V40 shielding the signals from the sensors to the receiver. The other interesting effect is number of turns, in the XC90 and the V40 it is negative but in the V60 it is positive. The project team is not completely sure why the effects turn out like this for the factor. But the project team would like to imply that a high value on this factor will decrease the stability of the system. To be complete confident about this statement the team would like top perform one or two more experiments to see if this was just a coincidence that the effect became positive in the V60. 4.3.3 Verification The subjective verification shown in 4.4 was performed to get an initial feeling of how well the MVN mannequin represents the body posture of the test subject. The verification using ImageJ was done to get hard data on how well the system performed. The error was bigger in the seated position as can be seen in appendix. H.2, this is most likely due to movement of the sensors when the test subject sits down in the chair. Especially the pelvis sensor had a tendency to move when the test subject was seated in such a steep angle between the hip and the spine(the test subject was seated in a 90° angle to make it easy to measure in ImageJ). The first verification subject had experience using the Awinda system, and may therefore been more aware of the movement of the pelvis sensor and deliberately have countered this. The first verification subject also had the Velcro straps with the sensors fastened directly on the skin (wearing shorts) to minimise the movement of the straps during seating. The other test participants were wearing pants with the Velcro straps fastened over, this may have resulted in that the pants moved the Velcro straps when the verification subject sat down in the chair. As a result the measurement errors of the first test subject are relatively even for the T-pose and the sitting verification, with the larges errors being 6.3° and 6.8° respectively. All the results from the verification can be seen in H.2 Some uncertainty came from the accuracy that joint positions was identified by the project team. Markers were placed where the joints was found on the VS to help the image analysis, but these markers could like the Velcro straps, have moved between the different postures and added to the error in the seated position. The verification could have been done using the data from MVN studio directly, but the project group choose not to, because it was very hard to see the exact ro- 35 4. Concept Development and Verification tation of the joint in ImageJ and compare that to the measurements that was done in MVN. So a 2D representation was done with the manikin instead. The biggest errors in the seated verification test are to big to be said to give a good representation of the joint angles of the VS, but all the visual representations that was done inside the car are a fairly well representation of the body posture of the VS. This could mean that the position of the joints were identified incorrectly, or that that the sensors moved more when the VS was seated at a steeper angle, or a combination of the two. 36 5 Method development and validation In this chapter theory, development approach and usage of the developed method capture drivers posture will be presented and discussed. 5.1 Introduction To be able to improve the DH-models with a better representation of the drivers posture data needs to be collected from real people when they drive a vehicle. To collect this data and to validate the method a study with n=45 participants will be performed utilizing the developed method capture drivers posture. As a validation tool to ensure the quality and evaluate the method a questionnaire is constructed and handed to the participants at the end of each study. This chapter contains identified limitations and prerequisites, applied theory and the development process of the method capture drivers posture. 5.1.1 Limitations and prerequisites for developing the method capture drivers posture Chapter 4 describes the work when the project team wanted to see how the impact of magnetic disturbance generated from the vehicle could be minimized. No real solution to the problem was identified, but if the warm up time after calibration was increased the system stayed stable a bit longer when subjected to magnetic fields. Unfortunately the findings in chapter 4 didn’t improve the system enough for it to be suitable for data collection during drive. A limitation before developing the method of capture drivers posture is that no data collection with the MVN Awinda system can be done during drive. Another limitation to consider when developing the method capture drivers posture is that only the joint angles will be captured with the MVN Awinda system. The last limitation derives from chapter 4 as well, due to the lack of robustness in the system uncertainties regarding the output data is fairly high and therefore the final result from the method will be seen as uncertain. A prerequisite on the outcome of the method is that in the collected data differ- ences between drivers posture in a low and a high vehicle can be identified. To be able to collect this type of data from the studys two different vehicles were used; a Volvo V40 and a XC60. 37 5. Method development and validation 5.2 Theory Description of theory behind the questionnaire design. 5.2.1 Questionnaire design To collect quantitative data from the studys a questionnaire was constructed. As suggested by McQuarrie (2012) a questionnaire should follow some basic rules. The first set of guidelines gives advice on the layout of the questionnaire and the second set gives guidance on how to phrase the questions and thirdly how to structure the answering categories. When constructing a questionnaire it is important to explain the value of com- pleting the questionnaire for the participants, usually this is done with a short but informative introduction in the beginning. Next step is to try to keep the survey as short as possible, this is due to economic incentives and if the task sees overwhelming to the participants it is more likely that they will not complete the survey. Other useful guidelines is to keep a red thread throughout the form, try to put the classifi- cation questions at the end and keep a professional layout if time and budget allows. (McQuarrie 2012). Creating the questions is as important as having a good layout. First of is to phrase the questions as simple as possible, a good rule of thumb is to adapt the language to a eight-graders reading level. Secondly be as specific as possible in the phrasing, even if the question gets a bit longer it will provide less room for misinterpretation from the participant. Open-ended questions is of great value when conducting a qualitative research but as for a quantitative survey they will demand too much from the participant and instead of answering the question the test subject might include biases which makes comparisons between results harder, but in the end of the survey it can be good to include one or two open-ended questions. As for answering categories in the questionnaire suggested by McQuarrie (2012), a ranking scale is preferred over a rating scale. Rating scales is good to utilize if the number of items to compare is less than five otherwise the task quickly can become too challenging. Another aspect of choosing a rating scale over a ranking is when analysing the ranking data, it has a tendency to become too time consuming. (McQuarrie 2012) 5.3 Developing the method capture drivers pos- ture Here development of the data collection method capture drivers posture will be described. 38 5. Method development and validation 5.3.1 Learnings and limitations In the previous chapter 4 a description of the instability problem with the MVN Awinda system and the possible solution to this is explored. Unfortunately these new findings changes the condition that data collection shall be done in a moving vehicle and adds a new limitation to the method capture drivers posture. The new approach that the project team has to work with is how to replicate the driving condition as good as possible. After a brainstorming session in the project team and discussion with the supervisors for the project consensus was reached, that the best solution is to let the test subjects drive the vehicles first and then apply the system and gather data of joint angles. 5.3.2 Questionnaire creation and development of the four stages in the method capture drivers posture A good way to structure and describe the method capture drivers posture is to divide it into four different stages. Figure 5.1 illustrates the different stages. Figure 5.1: The four stages in the method capture drivers posture 5.3.2.1 Questionnaire creation To be able to evaluate the method and assure that the test subjects assumed a good driving position a questionnaire was created with respect to the suggested guidelines by McQuarrie (2012). First the project team identified four major areas that the questionnaire should address, theses areas are evaluation of the method, validation of the test subjects driving position, limitations of the vehicle and perception of the MVN Awinda system. The full questionnaire with all questions can be found in appendix I.1. To each area a set of questions were created, each area contained two or three questions each, however emphasis was placed on the evaluation of the method and perception of the MVN Awinda system who contains three questions each. The final question of the questionnaire was an open-ended where the test subject was allowed to leave a comment of any kind as suggested by (McQuarrie 2012). For ranking scale the project team selected a 1-10 scale commonly used at similar studys. This was done since the test subjects would recognize the scale and give a more accurate answer. 39 5. Method development and validation 5.3.2.2 The drive stage The first stage of the capture drivers posture is to let the test subjects drive the two different vehicles and adjust the driver seat in a comfortable position. To give the test subject time to assume a good driving position and get them in to a driving mode a course of five minutes per vehicle were set. To give the test subject the opportunity to assume a comfortable driving position the driver seats in both cars were set to a starting position of furthest back, no tilt and at the lowest height. Before the test subject starts to drive the initial position of the motors in the seat is recorded from the CAN bus slot in the vehicle with Volvo cars software. 5.3.2.3 Applying the system After the drive stage it was time to apply the MVN Awinda system to the test subject. To perform this stage the project team allocated a testing facility with parking slots right outside. After the system was applied a calibration was performed inside the testing facility and the factor warm up time was applied. 5.3.2.4 Data collection After the MVN Awinda system was applied and calibrated the test subject was shown outside to the two vehicles and seated inside them. The subject was asked to assume their previous driving position as good as they can and then a sample of 15 seconds was recorded in the capturing software for the MVN Awinda system. Also a reference photo was taken from the side with a stationary GoPro camera. This was done for both vehicles. After the recording session the test subject was lead inside and the driver seats new position was measured and the seats were restored to their starting position. 5.3.2.5 Answering the questionnaire As final stage the system was removed from the test subject and a short introduction to the questionnaire was given. In the introduction the purpose and the ranking scale was explained for the test subject. 40 5. Method development and validation 5.4 Performing the method capture drivers pos- ture In this section description of performing the study capture drivers posture will be explained. 5.4.1 Acquiring test subjects The test subjects for the study were summoned from Volvo cars internal list of volunteers for studys. A total of n=45 subjects participated in the studys where the distribution consisted of 14 short people, 16 normal and 15 tall. A doodle poll with available time slots was created and an invitation was emailed. Due to low response rate the project team had to add an extra session of three days to gather the right amount participant. 5.4.2 Time allocation and execution The time allocated for each run in the study was set to 40 minutes, in the later stage the project team managed to lower the time to 30 minutes for each session. Since the method will become a compliment to an existing seat position method at Volvo cars the final time allocation will be 15 minutes where 10 minutes is for applying the system and 5 is for the data collection step. The study drivers posture was performed as described in previous section 5.3.2. Regarding needed equipment the method can utilize existing equipment that is used in the seat positioning method, the only extra equipment needed is the MVN Awinda system and a computer with the capturing software. Since the seating position of the driver is changed when the test subject was driving a low versus a high vehicle, the driver was positioned closer to the floor in a low vehicle then in a high. The two vehicles utilized in the study drivers posture were one low and one high, a V40 and a XC60. But in further studys the seat position study will dictate which vehicle to use. Picture 5.2 describes the interaction between the seat position study and the study drivers posture. 5.5 Result and validation from the questionnaire The result from the study capture drivers posture will be presented and discussed in chapter 6. In this section validation of the result from the questionnaire will be presented. 5.5.1 Result from the questionnaire The result from the questionnaire can be found in appendix J.1 for the short pop- ulation, J.2 for the normal and J.3 for the tall. The interesting questions for this 41 5. Method development and validation Figure 5.2: Interaction between the seat position study and drivers posture section is the ones regarding the areas of evaluation of the method and perception of the MVN Awinda system. 5.5.1.1 Results regarding evaluation of the method The questions concerning the method evaluation are How would you rate your overall experience of this study, Did you find the test course easy to drive and Was the test time allocated to the test drive sufficient for you to assume your perfect driving position, the questions will be referred to as 1,2 and 3 in the order as presented in the text. Figure 5.1 present the average score for each question. Table 5.1: Summary of the score for questions regarding evaluation of the method 5.5.1.2 Results regarding perception of the MVN Awinda system The questions of interest in the perception of the MVN Awinda system are Did the measurement equipment affect your driving position, Do you feel uncomfortable wearing the measurement equipment and Did you feel uncomfortable when the mea- surement equipment was applied. Same as in previous section the questions will be referred to as 1, 2 and 3 in the order as presented in the text. Figure 5.2 presents the average score for each question. 42 5. Method development and validation Table 5.2: Summary of the score for questions regarding perception of the MVN Awinda system 5.6 Discussion In this section discussion about the design of the questionnaire and the validation of the developed method capture drivers posture will be presented. 5.6.1 The questionnaire The first step in the creation of the questionnaire was to identify which areas should be examined. By identifying these main areas and structure the survey after them, it gives the questionnaire the red thread as suggested by McQuarrie (2012). The red thread is followed except for the first question which asks about the overall ex- perience of the study and it belongs to the group evaluation of the method area. This was done at purpose by the project team. Even that the question belongs to a certain area the project team thought it was so general and good start question. As for the phrasing of the questions the project team tried to keep the language as simple as possible and avoiding technical terms. This choice was made due to the test subjects different background and experience level in order to avoid misin- terpretation of the question. By applying a commonly used rating scale see figure 5.3, the project team be- lieve that it became easier for the test subjects to give a more accurate answer than they would done if not that scale was utilized. As suggested by McQuarrie (2012) when utilizing a ranked scale instead of rating between questions analysing the output data gets much easier. The scale is simple and ranging from 1-10 but for each number it has an explanation of how the impact of each score is. The project team still would like to say that the answers from the test subjects might not be as accurate as it would, since the observations were made when the test subjects who work in departments where this scale is commonly used in their daily job made a more restrictive rating then the other subjects did. A possible explanation to the observation might be that the technology used for data collection is new and exiting which may influence the test subject to become more generous in their rating. Over all the project team got the impression that the test subjects found the study to be innovative and fun which might explain the high overall experience score. 43 5. Method development and validation Figure 5.3: The rating scale utilized for the questionnaire in the method capture drivers posture 5.6.2 Evaluation of the method Judging from the score for the three questions regarding the evaluation of the method is that the project team can make some conclusions. Regarding the overall experi- ence of the study which got a 9,3 in average score out of 10 that all test subjects were pleased with the study. The same conclusion can be drawn for the choice of the test course, it is good to know that the test subjects found it easy to drive and that it didn’t generated any extra stress and was suitable regarding the test subjects driving skills. If the test course would have been too demanding to drive for the participants maybe the focus would have been more on the driving section in the study than it would have been on finding a suitable driving position. The lowest scoring question was the last question which asks the test subject if the right amount of time was allocated for them to assume their perfect driving position. Except for the first question that is about the general satisfaction of the study the project team would like to state that this is the most important question in the whole survey. Initially the data collection in the method would occur during drive, due to limitations in the technology this part had to be modified and changed to replicate a driving scenario. The score on this question helps to determine the quality of the captured data. If this question would have received a low score, the quality of the gathered data may not be accurate enough to represent the drivers posture in a driving scenario. Even though this question was the lowest scoring one with a 8,6 out of 10 the project team still feels confident that most of the test subjects actually had the time to assume their most comfortable driving position in both the vehicles. 5.6.3 Perception of the MVN Awinda system Since the technology used for data gathering in this study are a new technology, therefore it is important to know how the test subjects perceived it and if it caused any limitations in the study. One main concern was that the sensors would become a issue when the subject was sitting in the driver seat and could affect the test sub- jects driving position. The test subjects doesn’t believe the Awinda system affects their driving position, since the system received a 9,4 out of 10 the project team can make the conclusion that this type of data gathering tool can be suitable for this type of studys. The system consists of 16 individual IMU-sensors strapped to the body with Velcro straps. When the verification tests were done with the system, it was avoided to 44 5. Method development and validation have fabric between the straps and the body, instead of long legged pants shorts and t-shirt were used in the test. The project team have also tried out the system with fabric between the body and the strap and it feels a bit awkward but after some time that feeling is gone. Why the test subjects ranked this question as the lowest can be because it is the first time they try the system and the data collection time wasn’t long enough for them to "forget" about the system. Over all the score is high and gives a good indication to the project team that this type of equipment can be used for this type of data gathering with subjects that are unfamiliar with the system. The last question about how the systems is applied was the one that the project team had greatest concerns with. Since the system consists of Velcro straps that are applied to the legs and arms, body contact will occur between the test leader and subject. The initial thought was that some of the subjects didn’t expect this type of contact and would have declined when the project team informed them about that some body contact will occur. Luckily this was not the case, as seen by the ranking of the question 9,4 of 10 but also from comments made by the subjects when they were informed in the beginning of the study. The project team thinks of two reasons why the test subjects didn’t have any problem with this. Firstly the test subjects who signs up for this type of studies did it on their own initiative and they know what they have signed up for, another reason can be how the subjects perceive the approach from the test leader. If the test leader would act as the test subject felt uncomfortable and insecure with the task it would affected the mind set of the test subject but instead if the test leader have some experience with the system and acts in a professional the application of the system gets de-dramatized. 45 5. Method development and validation 46 6 Verification of the method Here the measurements from the clinics will be presented. A visual exemplification will be done to represent how the body posture representation differ between test subjects with different body shapes. Graphs will be shown to illustrate how the measured joint angles differs between test subjects and the low (V40) and high (XC60) vehicle. Data from the settings of the seat will not be presented. The graphs presented in this chapter are chosen to illustrate the difference in body posture, between different seating height and different test subjects. Therefore the joints chosen for representation are the joints that show the biggest difference between test subjects with different body shapes. A presentation of the measurements in each vehicle will be done, and then a comparison will be made between the different vehicles. This is done to show that the method are capable to show differences in posture between high and low vehicle and between different test subjects. The expected result are that the taller test subjects will have a less bent knee and hip due to longer legs, and that the smaller V40 will have less bent knee and hip joints, than the higher XC60. In fig.6.1 the MVN-manikin can be seen standing in the ’N-pose’ which is the refer- ence for the measurements made. Each joint have a coordinate system as described in the picture. In the ’N-pose’ every joint angle are zero around every coordinate axis, if a body segment are moved the system will give the corresponding joint angle difference from the starting angle. If for example the test subject lift their right arm straight ahead, there will be a angle rotation around the Z-axis(blue in the picture), and the joint angle will increase. A bigger joint angle means that the joint have rotated to the front of the body in reference to the starting position. 47 6. Verification of the method Figure 6.1: N-pose with coordinate systems and arm with angle indicator 48 6. Verification of the method 6.1 Measurements made in the V40 A clear difference can be seen in the seating position of the short test subject seen in fig.6.2 and the short heavy test subject seen in fig.6.3. The test subject in fig.6.2 are seated close to the steering wheel and have a steep angle in the hip joint and in the knees. The test subject presented in fig.6.3 have a almost straight hip- and knee-angle, and are more laid back in the seat than the test subject in fig.6.2. The seating position of these two test subject are very different although they are almost the same height. Figure 6.2: Short test subject positioned in the V40 Figure 6.3: Short heavy test subject positioned in the V40 49 6. Verification of the method The mid ranged test subject, seen in fig.6.4 have a steeper hip- and knee joint angle than the test subject in fig.6.3, but not as steep as the test subject in fig.6.2. The mid range test subject have a lower seating position, further away from the steering wheel than the short test subjects. Figure 6.4: Mid range test subject positioned in the V40 The tall test subject, seen in fig.6.5, has a very low seating position compared to the others, which gives a quite steep hip- and knee angle. Figure 6.5: Tall test subject positioned in the V40 The graph 6.6 show a very low R2-value, this means that the data is not showing any significant correlation between the joint angle of the knee and the body height of the test subjects. 50 6. Verification of the method Figure 6.6: Graph over the joint angles of the right knee, measured in the V40 In the graph over the hip joint-angle, fig.6.7 Figure 6.7: Graph over the joint angle of the right hip, measured in the V40 51 6. Verification of the method 6.8 shows the joint angle of the shoulder in the V40. The graph shows a slight increase of the joint angle for the test subjects that are taller. Due to the R2-value no correlation can be said to occur. Figure 6.8: Graph over the joint angle in the right shoulder 52 6. Verification of the method 6.2 Measurements made in the XC60 The short test subject, see fig.6.9, Figure 6.9: Short test subject positioned in the XC60 and short heavy test subject positioned in the XC60 The mid range test subject seen in fig.6.10 have a smaller angle at the knee in the XC60 which means a straighter leg. Figure 6.10: Mid range test subject positioned in the XC60 In fig.6.11 the tall test subject can be seen positioned in the XC60. 53 6. Verification of the method Figure 6.11: Tall test subject positioned in the XC60 The graph over the knee-angle of the test subjects in the XC60, show just like in the V40 that the R2-value are too low to show any correlation between body height and knee-angle. Figure 6.12: Graph over the joint angle of the right knee, measured in the XC60 The graph in 6.14 show the joint angle of the test subjects in the XC60. The correlation are bigger than for the knee-joint but the R2-value are still too low to show any correlation between body height and hip-angle. 54 6. Verification of the method Figure 6.13: Graph over the joint angle of the right hip, measured in the XC60 Figure 6.14: Graph over the joint angle of the right shoulder 6.3 Comparison between the posture in the V40 and the XC60 The graph in 6.15 show the correlation between the knee angle in the XC60 compared to the knee angle in the V40. The graph shows that the test subjects have a slightly 55 6. Verification of the method more bent knee in the V40 compared to the XC60. Figure 6.15: The knee joint angle plotted for the XC60 compared to the V40 The graph in 6.16 show the correlation between the hip angle in the XC60 compared to the hip angle in the V40. Like the knee angle it shows a slightly more bent joint angle in the V40 compared to the XC60. Figure 6.16: The hip joint angle plotted for the XC60 compared to the V40 The graph in 6.17 show the correlation of the joint angle in the shoulder in the 56 6. Verification of the method XC60 compared to the joint angle in the V40. The R2-value are too low to show a correlation between the joint angle in the two vehicles. Figure 6.17: The shoulder joint angle plotted for the XC60 compared to the V40 Fig. 6.18 show that the measured ankle joint angle of the test subjects are bigger in the V40 than in the XC60. 57 6. Verification of the method Figure 6.18: The joint angle of the ankle plotted for the XC60 compared to the V40 6.4 Discussion The measurements presented in 6.1 and 6.2 show that the chosen method is capable to detect the joint angles of the test subjects, and show a difference in posture for test subjects with different body height. The data does not show a clear correlation between the body height of the test subjects, this is probably due to the adjustment possibilities of the vehicles. The data presented in 6.3 shows that the method is capable of showing a difference in the posture of test subjects in different vehicles, with different seating height. In the beginning of the chapter suggested theory indicates that the test subject would have a straighter leg and therefore a lower knee joint angle in the V40 than in the XC60. The results from the data collection displays that the test subjects have the lowest and highest knee angle in the XC60. This is believed to occur due to the greater adjustment area in the XC60 compared to the V40. 58 7 Further development, recommendations, final result and final discussion Here further development to improve the method and use of the MVN Awinda, recommendations will be given on how Volvo car can continue with this system and other application areas, and a final discussion will be presented. 7.1 Further development Here further development and alternative uses with the method and the MVN Awinda system are presented. There are many possible applications for the MVN Awinda at Volvo car and other industries. The motion capture function could be used for ergonomics analysis in many different scenarios for example ingress and egress(analysis of vehicle ingress and egress). The system could also be used to sim- ulate many different scenarios in real-time (Reliability, you value it when you don’t have it), examples of this is walk through of production facilities, simulation of break down scenarios with users and as a visualization tool for evaluation of concepts. 7.1.1 Creation of a new world After the concept selection of the MVN Awinda system discussion of different areas of application started to interest the project team. The project team knew that the MVN Awinda could be applied with Unreal Engine 4, but how to do it?. Unreal Engine 4 is a game engine with tools for creation of games, rendering engine for the graphics of the game, physics engine which defines for example gravity and how objects behave when they interact with each other in the created world or game and much more (What is Unreal Engine 4 ). The MVN Awinda have support for linking the system with this game engine in order to replicate the movement to the character in game done by the person using the system. Another important tool when creating this new world is a VR-headset, examples of these headsets could be the HTC Vive or the Oculus rift. The VR headset creates a link between the in game characters head and eyes and allows the user to gain first person view in game and experience the world or game as close as the real world. 59 7. Further development, recommendations, final result and final discussion By combining the unreal engine 4, the MVN Awinda and a VR headset creation of a world, room or game where the movement and vision of the user is replicated is now possible. With this new world basically scenario can be created and expe- rienced as it was the real world. Evaluation of a new factory where the engineers could evaluate placement of machinery, width of isles and much more is now possible as long as the plant is designed with CAE tools. 7.1.2 Further development of the MVN Awinda There are also some improvements from Xsens that can make the system better. Xsens are currently developing sensors that use GPS-data instead of the magne- tometers to determine true north (Xsens Research). This eliminates the problem with the magnetic disturbances in the car, and gives the possibility to conduct mea- surements while the test subject are actually driving the car. This feature have hardware support in the MVN Link but