Master Thesis - A threat-assessment system based on supervised Bayesian deep learning
Master Thesis & Research Opportunities - Gothenburg, Sweden
Recently, autonomous active safety systems have become an important corner stone in the pursue of fewer traffic-related fatalities. New sensor technology and cheaper computational power have accelerated the development of highly effective advanced collision avoidance systems. The major challenge for any active safety system is how to accurately predict future threats so to correctly assist the driver. However, the system performance is directly dependent on hardware and software limitations such as, e.g., sensor noise, driver’s actions, or different environmental conditions.
In this thesis, we aim to develop an autonomous collision avoidance system (CAS) based on an artificial Bayesian Neural Network (BNN). The BNN should be trained with real log-data and the results compared with a baseline platform. The main idea is to use the neural network for the decision making process, e.g., to determine if and when to trigger an autonomous steering intervention.
Based on real data, the purpose of this thesis is to benchmark the real-life performance of an BNN when applied to a CAS. The goals are:
- Develop different BNN structures and elaborate on different parameter and signal configurations
- Derive benchmark figures and compare them to related works, e.g., number of correctly triggered interventions, etc.
We are looking for (one or two) students with the following criteria:
- A highly motivated student from any master program with background and interest in deep learning algorithms, data analysis and probability theory.
- The ideal candidate should have interest in both the theoretical and experimentation aspects of the problem.
- Solid programming skills are required and a particular experience with python is an asset.
- Effective communication skills in English both in oral and especially written are also appreciated.
The thesis students will gain
- Competences on
o Autonomous active safety functions
o Developing deep learning algorithms for market leading technology
o Automotive engineering
- Industrial experience within a new technological company for intelligent and self-driving vehicles
Further information and contacts
Final application date: 31 October 2019
Please send in individual applications with CV, motivational letter and grade transcripts. If you wish to partner with someone, simply note that in your application.
Planned start: 15 January 2020, with some flexibility.
Duration: 30 ECTS
This is a thesis project driven by Zenuity in collaboration with Chalmers University of Technology at the department of Electrical engineering. For questions regarding the project, please contact:
Zenuity John Dahl (firstname.lastname@example.org, +4672-391 54 73)