Innovation starts with being curious.
At Volvo Cars, we believe that being curious and truly committed to understanding people is the key to future success. We are people who care about other people, working together to create new technologies and innovations for safe, sustainable and convenient mobility. Want to join us?
In order to assess autonomous drive (AD) vehicle safety with confidence, statistical analyses have shown that fully autonomous vehicles would have to be driven for more than hundreds of millions of kilometers. This is not feasible, particularly in cases when we need to assess different system design proposals or in case of system changes, since the same amount of distance need to be driven again by the AD vehicle for the verification sign-off. Therefore, one cannot simply drive to safety and more advanced methods of demonstrating safety and reliability assurance need to be adopted. An alternative verification approach to address some of the above mentioned issues is known as scenario-based verification approach, where a scenario bank is created by extracting driving scenarios/events that the ego AD car is exposed to, in naturalistic driving traffic situations. Once a scenario database is developed, it can be used for test case generation and verification of the AD functionality in virtual and real driving test environments. Each scenario model class (e.g. cut-in or cut-out scenarios) is specified by the relative movement of surrounding interacting vehicles, with respect to the AD ego vehicle, and can be extracted from the post-processed logged data in different ways, i.e. knowledge-based (explicit-rule) approach or Machine Learning based approach, that can complement each other.
As AD function & sensor platform Agile Release Train (ART), we have responsibility for development, maintenance, verification and safety assurance of Autonomous drive function. Considering the above-mentioned issues, the scenario analysis team, in AD function & sensor platform Agile Release Train (ART), is responsible for investigating and expanding the pallet of tools for establishing and maintaining the “scenario database” up-to-date, for further verification purposes.
A machine learning based scenario tagging tool has already been developed, in Python environment, helping us group different driving trajectories into different scenario model classes. The performance of the method, applied to a proof of concept study for tagging scenarios, seems promising. Also, an automatic procedure for estimating a reasonable value for the number of scenario groups/clusters has been implemented. The focus of this thesis will be on benchmarking different machine-learning scenario trajectory tagging tools and compare the performance of the developed tool with other alternatives.
Purpose/Aim of this thesis project
As the continuation of previous work, the aim of this thesis will be on:
• Benchmarking different machine/deep learning methods for our scenario trajectory/sequence tagging purposes by implementing the new approaches.
• Compare the implemented methods with our current running trajectory tagging tool, assess the tool performance and give suggestion for further improvements.
• Summarize the cons/pros of the studied methods and try address important question regarding their applicability for our scenario-tagging purposes. Some questions to answer 1) what is the limitations regarding the length of sequences and for each method how many time step in each scenario sequence is reasonable? 2) Practical solutions in handling longer sequence? 3) investigate the required data size for different methods, for our scenario tagging problem (since the quantity of data that is needed, depends on the complexity of the network that varies from problem to problem, depending on different function complexity)
• Investigate the performance of the implemented tools for characterization and refined analysis of more complex driving scenarios (e.g. lane merge).
• The currently developed model selection procedure, for finding the number of scenario groups, uses an approximate solution, which depending on student’s interests can be further improved by implementing a full Bayesian approach for estimating the number of scenario classes. (The Matlab implementation of the full Bayesian model selection is ready, but needs to be implemented in Python for integration with the ML tagging tool)
Starting date: November or to be agreed between Volvo Cars and the student/students
Number of students: 1 or 2 (preferably 2)
Do you fit the profile?
We are looking for you who is currently studying Computer Science, Physics, Mathematics or Electrical engineering. You also have taken courses in machine learning and AI, and Statistics/Mathematics. Moreover you have good programming skills (Python is preferred, Matlab is a merit). Good knowledge in automotive and safety is a merit.
How you can learn more and apply?
If you have any questions please contact academic supervisor Morteza Haghir Chehreghani at email@example.com and industrial supervisors Sadegh Rahrovani at firstname.lastname@example.org or Martin Magnusson at email@example.com. Please apply with your CV & cover letter while using the electronic link further down in this ad as soon as possible but no later than 2019-11-01.
Who are we?
Everything we do starts with people. Our purpose is to provide freedom to move, in a personal, sustainable and safe way. We are committed to simplifying our customers’ lives by offering better technology solutions that improve their impact on the world and bringing the most advanced mobility innovations to protect them, their loved ones and the people around them.
Volvo Cars’ continued success is the result of a collaborative, diverse, and inclusive working environment. The people of Volvo Cars are committed to making a difference in our world. Today, we are one of the most well-known and respected car brands, with over 40,000 employees across the globe. We believe in bringing out the best in each other and harnessing the true power of people. At Volvo Cars your career is designed around your talents and aspirations so you can reach your full potential. Join us on a journey of a lifetime as we create safety, autonomous driving and electrification technologies of tomorrow.