Examensarbeten för masterexamen

Browse

Senast publicerade

Visar 1 - 5 av 1730
  • Post
    A study of parameter trade-offs with Re- STIR for low-cost rendering and high-quality reflections
    (2024) Wang, Yuhan; Chalmers tekniska högskola / Institutionen för data och informationsteknik; Chalmers University of Technology / Department of Computer Science and Engineering; Assarsson, Ulf; Sintorn, Erik
    It is the aim of this thesis to conduct a thorough analysis related to the effects of parameters on render efficiency or image quality of Path Tracing algorithm known as ReSTIR. In this case, the paper is based on the prior algorithms like ReSTIR, ReSTIR GI, and ReSTIR PT, and provides an integrated viewpoint. The analysis begins by first clearly explaining the ReSTIR algorithm before conducting an extensive examination regarding its multiple parameters influencing it provided speed. As a result, considerable exchanges between/among these parameters and rendering modes are bound to occur. So, the goal is to balance the efficiency gains and quality of the images produced with a ray tracing supported GPU. In this case, therefore, we have employed plots in determining the best values for parameters that dramatically affect the efficiency and image quality of ReSTIR real time rendering. These are all beneficial in mapping which parameters need tuning in order to yield specific renderings. In conclusion, this thesis provides a clear insight about the influence of parameters in ReSTIR (PT) at render efficiency and quality. Such recommendations suggest improvement strategies for different facets of the algorithm, which offer prospects for increasing the runtime rendering speed and the quality of images in ReSTIR-based systems.
  • Post
    Transferring Resilience Metrics: Evaluating the Adaptability of Resilience Metrics from Different Domains for Assessing Vehicle Resilience
    (2024) Pasalic, Dino; Retteli, Mattias; Chalmers tekniska högskola / Institutionen för data och informationsteknik; Chalmers University of Technology / Department of Computer Science and Engineering; Almgren, Magnus; Almgren, Magnus
    Recent advancements in the automotive domain have made vehicles more dependent on the Internet to make the driving experience smoother. This increases the attack surface as this new connectivity can be targeted and it cannot be assumed that these attack surfaces are secure. To still build secure systems resilience can be used. Resilience is, the extent a system can sustain attacks and still be functional, to a certain degree. Currently there exists no metrics for cyber security resilience in vehicular design. This thesis aims to explore existing metrics in related domain to investigate the possibility of adapting those metrics into the automotive domain. To this end, a literature study is performed with 10 investigated papers with 244 metrics found. Attack-injection experiments are performed using the MODIFI tool and Simulink models on the adapted Health Index metric to evaluate the difficulty and practicality of adapting metrics. We found that it is possible to adapt resilience metrics from related domains into the automotive domain. However, depending on the complexity of the metric it can be difficult to adapt it as some modifications may be necessary to be able to use it within the automotive domain.
  • Post
    Validation and Verification Challenges in a Machine Learning Algorithm for Connected Vehicles - Design Science Research of Developing a Most Probable Path Algorithm
    (2024) Hertzberg, Axel; Bengtsson, Erik; Chalmers tekniska högskola / Institutionen för data och informationsteknik; Chalmers University of Technology / Department of Computer Science and Engineering; Berger, Christian; Feldt, Robert
    Machine Learning (ML) software in connected and automated vehicles puts new demands on safety regulators and industry standards to keep up with the explosive evolution of technology in the automotive domain. This thesis reports a practical example of developing an ML-based algorithm that predicts the most probable path for an arbitrary vehicle, without knowing the destination. This work is done in collaboration with Carmenta Automotive AB as an industry partner, a company that is aiming to increase situational awareness for vehicles on the roads. The thesis methodology follows an iterative design science research (DSR) approach, developing an artifact consisting of an ML model connected to the company’s system. The literature highlights the challenges of validating and verifying (V&V) an ML component, as there are currently no applicable standards for ML software in the automotive domain. This DSR attempts to showcase V&V activities on ML models trained with different data characteristics to assess whether the challenges surrounding V&V can be mitigated when validating the data-driven most probable path algorithm.
  • Post
    Mutation Testing in Industry CI: A Value- Centric Approach - A case study about the intersection of software quality and developer experience.
    (2024) van Heijningen, Stefan; Wiik, Theo; Chalmers tekniska högskola / Institutionen för data och informationsteknik; Chalmers University of Technology / Department of Computer Science and Engineering; Torkar, Richard; Gay, Gregory; Gomez, Francisco
    Context: Mutation testing is a robust testing technique to assess the sufficiency of test suites. The industry is still struggling to adopt mutation testing and maximize its benefits despite recent research suggesting it is becoming more mature. Objectives: This study aims to share insights and recommendations to: 1. Assist developers during the integration process of mutation testing tools. 2. Present mutation testing results to maximize their benefit and minimize the cost of using them. Methods: We perform a case study in an industry setting. We create an experience report that reflects on the integration process of a mutation testing tool at the partner company. We then focus on developer experience by using ten think-aloud sessions and semi-structured interviews to explore what information developers perceive as useful and how the information should be presented. Results: A CI pipeline was developed to run mutation testing nightly and upload results to a developed dashboard. Integrating mutation testing tools is still a challenging process. When interacting with results, developers valued interactivity, getting an overview, and wanted to associate mutation testing results with other contextual information of the codebase, such as code coverage and complexity. A set of recommendations was created to facilitate integrating and using mutation testing in industry settings. Conclusion: As-is, mutation testing is a standalone tool that should be easier to integrate and interact with other aspects of the codebase to become more adopted and successful.
  • Post
    A Software Engineering Perspective on Data Quality Processes in Environmental Research - Recommendations Based on Software Engineering Practices Applied for Improving of Open Data Practices and Communication in Environmental Research
    (2024) MOEN, MARKUS; NORÉN, MAX; Chalmers tekniska högskola / Institutionen för data och informationsteknik; Chalmers University of Technology / Department of Computer Science and Engineering; Penzenstadler, Birgit; Heyn, Hans-Martin
    The many fields within environmental research have been on the path towards open science and, most importantly, open data. With the increase in available data, there are opportunities to apply data-driven and data-intensive methods, including recent developments such as machine learning. However, the success of applying machine learning depends significantly on the quality of the available training data. The purpose of this thesis was to investigate the field of environmental research in regards to current views, practices and communication of data quality and to identify software engineering principles and practices that can form possible recommendations to progress data quality in environmental research. This process identified six challenges and proposed eight recommendations. The result shows a great deal of effort towards open data, with the FAIR principles as the main arbiter to achieve it. Most identified challenges are based on data quality handling, communication, and difficulties in achieving open science. We found suitable software engineering practices for four of the six challenges, with two key perspectives being derived from open source software and requirements engineering practices. Our results demonstrate that there is a willingness among environmental researchers to investigate and adopt software engineering practices in environmental research. Importantly, there is a broad agreement that open science is an improvement over to previous methods, and the stated challenges and recommendations need to preserve those advancements. The recommendations should be regarded as a first design iteration of these recommendations, and they should be explored further in terms of their applicability to different fields within environmental research.