Volvo Cars - Safe Vehicle Automation
Master's Thesis Proposals 2024
Welcome to explore the world of Volvo Cars by writing your thesis with us! As a thesis worker in Volvo’s organization, you are supported by a supervisor who follows you during your project. All thesis projects are arranged in business-critical areas and therefore you will be able to contribute to their company purpose – providing freedom to move in a safe, sustainable and personal way – from day one!
- Read through all of the theses, and then fill out the form at the end of the page.
- If you already have a master’s thesis partner, it’s only necessary for one of you to submit the info.
- Discuss the subjects with your thesis partner, and choose all of the theses subjects you are interested in. If you only choose one, we cannot guarantee you will get it, depending on how popular the thesis is.
- Deadline to decide which theses you are interested in is Friday 27th of October, but the earlier you decide the better since we will be presenting the thesis pairs to the team managers and supervisors continuously.
System Safety for Artificial Intelligence based systems
Background: The explosive development of Artificial Intelligence (AI) technologies has revolutionized numerous industries, and one area profoundly impacted is human mobility. The application of various AI techniques has significantly enhanced autonomous capabilities of modern vehicles. However, this progress has also brought forth new challenges in terms of safety. AI solutions possess inherent weaknesses, including prediction uncertainties, lack of interpretability and traceability, and vulnerabilities to adversarial attacks. These weaknesses amplify the safety challenges associated with autonomous decision making in vehicles.
To address these challenges, the industry has developed and adapted specific safety standards for the AI era. Examples of such standards include ISO26262 and ISO 21448. These standards aim to provide formalized approaches for mitigating safety risks associated with AI in automotive applications.
Objective: The primary objective of this thesis is to apply suitable safety strategies to automotive functions realized through AI techniques to ensure compliance with safety standards. By doing so, we aim to build a simulation platform with a high degree of completion for testing and demonstrating AI safety.
- Review and analyze existing safety standards and literatures related to AI safety, to gain a comprehensive understanding of the state of the art.
- Design the safety strategies for AI functions, build simulation platform for test and demonstration.
- Provide recommendations and guidelines based on the research findings to improve the safety practices and regulatory frameworks concerning AI-based functions.
- Open minded
- Team player
- The ability to actively learn and ask questions
- Background knowledge of vehicles is a plus
- Hard skills within machine learning, preferebly computer vision
- Experience of training neural networks
- Knowledge about simulation modeling and control
- Experience/knowledge in mechatronics is a plus
- Knowledge of system safety or vehicle simulation tools is a plus
Deep Learning-Based Traffic Accident Detection
Background: The development of autonomous driving and advanced driver assistance systems relies on the accurate detection of abnormal events, including traffic violations and accidents, in real-world driving scenes. Current video anomaly detection methods often assume scenarios that do not align with vehicle-mounted cameras and depend heavily on manually labeled training datasets. This thesis proposal aims to overcome these limitations by introducing a deep learning approach for traffic accident detection.
Problem Statement: To develop a deep neural network approach for traffic accident detection in first-person.
Insights: The effectiveness of predicting the future locations of traffic participants as a means of anomaly detection.
Data Source: Gather a dataset of first-person dashboard-mounted camera videos, including diverse traffic accidents and normal driving scenes.
- Eager to learn
- Studies courses within machine learning
- Knowledge on developing deep neural networks
Develop a scalable model that predicts energy consumption for the Autonomous Drive sensors and core compute platform and verify it in real world use cases
Background: Sustainability is a core value of Volvo Car Corporation and Safe Vehicle Automation is looking for a data driven model that can help us predict energy usage as we move into new computer platforms or improve the efficiency of the AD perception SW.
Problem statement: The world is not given by our parents, but borrowed from our children.
Objectives: Identify Key Performance Indicators (KPI’s) and what scenarios that will allow us to collect relevant data points. Create a scalable model where contributions from: (Environment; Sensors; Compute Platform (SoC); SW process efficiency; Actuators). Finally, evaluate the model and its predictive capability with real world data collected in a test vehicle.
- Studying control systems
- Driver’s license
- Self-motivated; able to search for info on your own before asking
- The ability to co-operate with other teams and departments
- An interest in engineering and cars
- C++ or Python
- Control systems
- Experience/knowledge of CI/CD is a plus
Identification of human gestures by Ultrasonics to be used as sensory input for functional actuation
Background: Ultrasonic sensors have been used as parking assistance sensors in Volvo cars for more than 10 years and have subsequently been utilized for low speed collision avoidance features. It is resided in the AD & ADAS domain and at present being considered for other features like hood/trunk opening.
Objectives: The objective of this master thesis shall be to use the detections of human made gestures (intentional) to be used as a sensory input to actuate any kind of feature (for example window roll down, open door, turn on/off light and potentially be customizable through the Volvo App). The master thesis is encourage to use any approach to be able to detect patterns through the cluster of echo data available from the ultrasonic sensors: Classic/Machine learning/Deep Learning)
Scope: At present the gesture recognition is only considered for ultrasonics only and not a fused system but this can be extended based on internal (Volvo Cars) interest or student interest.
- Studying software, systems, deep learning, or mechatronics
- Driver’s license
- Be able to see opportunities rather than problems
- Experience/knowledge of data analysis and visualization of data
- Python or C++
Verification Methods for Complex System Design within AD and ADAS
Background: Autonomous driving and Advanced Driver Aid Systems (ADAS) are focus areas for Volvo cars and we are striving to be leading within this area. Within our department we are developing the solution to these customer functions. The solution is a complex product described in several different abstraction levels, that requires verification on all levels and in multiple environments such as software in the loop (SIL), hardware in loop (HIL) or full vehicle.
Today there is full coverage at function and software level, but in between there is a system layer with a combination of hardware and software that works to realize the functions. In this level the test scope becomes more complex than on SW level and more detailed than on function level, so the same method does not automatically apply.
The space spanned by the possible test vectors at this level is so large that there is a need to find the most efficient way of verifying requirements. Both to reduce the number of needed test cases, but also in what environment and tools that are best suited for each test.
We see a need to explore several aspects of this, and depending on the interests and academic focus of the individual this thesis can explore either the technical and tool aspects, or the methodical aspects if the applicant has a more project management focus.
Objective: The primary objective is to provide input to our testing strategy that will improve our efficiency in testing our system level, either by improving tools or methods.
Additionally, a crucial aspect of this project is to enhance our team’s expertise in this domain. Consequently, this thesis offers a valuable opportunity for long-term skill development.
- Ability to analyze complex problems
- Pedagogical and communicative, should be able to explain things in simple ways
- Project management skills is a plus
- Experience from testing is a plus
- Teaching experience is a plus