Examensarbeten för masterexamen


Senast publicerade

Visar 1 - 5 av 1236
  • Post
    Stimulate learning for autonomous vehicles: From building the Kiwi car to implementation of software
    (2021) Jansson, Anton; Chalmers tekniska högskola / Institutionen för mekanik och maritima vetenskaper; Chalmers University of Technology / Department of Mechanics and Maritime Sciences; Benderius, Ola; Benderius, Ola
    In this thesis, the so called Kiwi platform was studied as a tool for teaching au tonomous systems and OpenDLV to four target groups: secondary education, un dergraduate education, postgraduate education, and hobbyists. The Kiwi platform consists of a 3D printed car equipped with sensors for measuring distances as well as a camera for image processing. Moreover, the open source software environment OpenDLV, was used to develop and test algorithms for driving the Kiwi car autonomously. The project was divided into two main parts, starting with building the Kiwi car and then programming it. Afterwards a survey was conducted in order to evaluate this teaching concept. Only one test group was involved in building the car, while all groups programmed it. In order to build the car, the primary resource was step by step instructions. When programming the car, the test groups were given a selection of mandatory topics. The steps were presented both in text as well as with video tutorials. Based on the results from the survey the participants found the tutorials relevant and extensive enough for the four programming tasks. Having said that, only the two test groups with the expected highest prior knowledge completed all the tasks in time, which may indicate that the tutorials may have been to advanced. Furthermore, the negative feedback from the survey were regarding the more technically challenging sections. The test groups with highest expected prior knowledge found that the material as a simplification of their intent, this was not expressed by the other test groups. Regarding the user manual for building the Kiwi car, the data was not extensive enough to draw any general conclusions.
  • Post
    Implementation of interpretable methods for paraphrasing and text disambiguation
    (2023) Carlström, Klara; Chalmers tekniska högskola / Institutionen för mekanik och maritima vetenskaper; Chalmers University of Technology / Department of Mechanics and Maritime Sciences; Wahde , Mattias; Wahde, Mattias
    In this project, starting from an interpretable language model based on knowledge graphs, four essential methods for natural language processing (NLP) have been developed, namely (i) paraphrasing, (ii) part-of-speech tagging, (iii) semantic similarity analysis, and (iv) text simplification. The methods yield good results on a small dataset and thus offer promising prospects for continuing research on interpretable NLP. Applications of NLP are becoming increasingly embedded in our daily lives in applications such as voice assistants, automatic language translation, opinion mining and medical diagnostics. One of the reasons behind the exponentially growing interest in NLP is the development of deep neural network (DNN) models that have achieved outstanding performance on various NLP tasks. However, the domination of DNN models has been followed by deep concerns regarding the black-box nature of such systems. By contrast, the language model used here is fully interpretable, paving the way for safe and accountable NLP.
  • Post
    The use of ETA in prediction of pick-up of containerized goods at port
    (2022) Ringborg, Johannes; Svanborg, Erik; Chalmers tekniska högskola / Institutionen för mekanik och maritima vetenskaper; Chalmers University of Technology / Department of Mechanics and Maritime Sciences; Ringsberg , Henrik; Ringsberg , Henrik
    Increased global trade combined with logistical disturbances due to COVID – 19 as well as Ever Given blockade of the Suez Canal have generated increased pressure on the global supply chain during the recent years. With the global trading predicted to grow during the upcoming period, partly due to increased e-commerce and consumption, the situation within the intermodal transportation chain is more pressured than ever. At the same time, the goods – owners still desire to track their cargo as well as receive precise information regarding the predicted arrival time. This kind of transparency could be difficult to address during the current state of the world, especially within a dynamic environment, such as the maritime domain. This thesis addresses the possibilities as well as difficulties in predicting the arrival time for container vessels, based on the Estimated Time of Arrival (ETA) retrieved from the Automatic Identification System (AIS). The aim is to present a verified process model for import containerised cargo, based on the ETA from the AIS, which could simplify the pick-up process of containers at container terminals. By utilizing a mixed methods approach, through a literature review, interviews and a questionnaire survey, data, information, and opinions were collected from stakeholders within the supply chain. The result showed that there was an overall interest for the subject, and that it was a highly relevant research area due to the ongoing situation within the global supply chain. Three main interfaces were recognized, with different stakeholders interfering with each other alongside the physical movement of containers. By utilizing the AIS and retrieving a reliable ETA, through preferably machine learning, a seamless information exchange regarding the predicted arrival time for container vessel could be possible. If the ETA is reliable enough, it can provide valuable information through all three interfaces, which could simplify different processes for the included stakeholders. It could also serve as a solid foundation for a verified process model, which could simplify the pick-up process for containerised cargo at port container terminals.
  • Post
    Supply Chain Market Study on Business Intelligence Technologies:With an emphasis on Big Data Management in Ocean Freight Transportation
    (2022) Abil, Alper; Chalmers tekniska högskola / Institutionen för mekanik och maritima vetenskaper; Chalmers University of Technology / Department of Mechanics and Maritime Sciences; Ringsberg, Henrik; Ringsberg, Henrik
    In the last two decades, with the developing technology, data science and the useful tools of data science have started to be used more frequently in the business world. The benefits of BI tools such as real-time visibility, track and trace, route optimization, rate benchmarking and forecasting in processing acquired data could not be overlooked among these developments. Especially in the shipping industry, the need for these BI tools has increased considerably to determine the unpredictable freight prices, customer demand amounts, and the direction of trade flow. This thesis shows, it was concluded that BI tools are very useful for decision making, increasing efficiency, visualization, and optimization for the transportation industry. Obtaining information based on a quantitative approach was supported with the qualitative approach through the survey, in this context, the importance of Business Intelligence was emphasized for how Big Data can be used and to extract the necessary data from it. In line with the findings obtained, the use of Big Data and the place of Business Intelligence were evaluated throughout this research, while the supply chain, which has an impact on every aspect of our lives, benefits from the cutting-edge developments.
  • Post
    Modelling A-pillar overflow - Using a smoothed particle hydrodynamics based method
    (2023) Larsson, Martin; Chalmers tekniska högskola / Institutionen för mekanik och maritima vetenskaper; Chalmers University of Technology / Department of Mechanics and Maritime Sciences; Sasic, Srdjan; Naixian,, Lu
    A-pillar overflow is the event when fluid is transported from the windshield across the A-pillar, ending up on the driver side window, obscuring the driver’s vision. Simulations of A-pillar overflow can make initial predictions of how the driver’s vision will be affected during windscreen washing or rain, and reduce developmental costs by making earlier design changes. Earlier numerical simulations have been carried out using traditional Finite Volume Method (FVM) Computational Fluid Dynamics-solvers (CFD) based on hybrid methods using Lagrangian Particle Tracking (LPT) and Volume of Fluid (VOF). Since A-pillar overflow is a transient event with moving wipers, requiring a transient mesh, it increases the computational cost and can induce numerical instabilities. By applying a Smoothed Particle Hydrodynamics (SPH) solver the need for a mesh is removed, but this approach has less validated solvers and problems with particle size-dependent model constants. This thesis aims at investigating A-pillar overflow using an SPH-based solver, PreonLab, and qualitatively validate the simulation with physical tests. The purpose of the thesis is firstly to establish a feasible workflow to simulate a windscreen washing event. The wiper kinematics modelled by a multibody dynamics software, ADAMS, and the airflow computed by an FVM method in Star-CCM+ are imported to PreonLab. Model constants such as particle spacing, adhesion and roughness factor are studied using validation against simple physical test cases. Secondly, it is to simulate A-pillar overflow on the Volvo V90 and XC40, where the amount of liquid arriving on the driver side window is substantially different due to different styling around the A-pillar area. Results indicate that wiper cycle simulations could be conducted in PreonLab in the future, as the overall behaviour of the fluid is captured through tuning of model parameters. Due to a lack of validation of the surface parameters and the density used in the airflow implementation, the simulation method is not fully validated. Further studies on airflow-liquid interaction models and surface properties need to be done in order to capture the complicated physics of an A-pillar overflow simulation