Planning Safe Maneuvers of Automated Vehicles

Motivation

A great challenge of automated vehicles is to guarantee their safety. Since automated vehicles are operating in unpredictable environments, such as road traffic, their planning methods have to deal with various uncertainties. In particular, the prediction of the future behavior of other traffic participants is a sophisticated task.

 

Objective 

In collaboration with the BMW Group, the CAR@TUM project develops methods which can guarantee the safety of automated vehicles. To consider all unforeseeable events, these methods must be formally verified. We can accomplish this by employing reachability analysis of cyber-physical systems.

The CAR@TUM research project is structured as follows:

  • Set-based prediction of other traffic participants: By computing an over-approximation of the feasible occupancies of all surrounding traffic participants over time, we can formally guarantee whether the automated vehicle can possibly collide with other traffic participants.
  • Reachable set of the ego vehicle: We propose an algorithm which efficiently computes the safe, reachable area of the automated vehicle in consideration of the ego vehicle dynamics and the occupancy of surrounding traffic participants.
  • Safe trajectory planning: Based on the reachable set of the ego vehicle, its driving maneuvers are planned. Our methods can, in contrast to others, formally verify that the planned trajectories are collision-free.
  • Falsification of the trajectory planner: A mad driver model will be developed to verify the safety of our approach.
  • Validation in the BMW simulator: All proposed methods of the above work packages will be validated in the BMW simulator.
  • Experiments with a real vehicle: Finally, the safe maneuver planner will be implemented in an experimental vehicle of BMW to drive on public roads. 

 

People


Student Projects

If you are interested in this research project, please contact Markus Koschi to discuss possible topics for a Bachelor or Master thesis or other projects.

  • Ongoing:
    • Maria Althaus: Consideration of Safe Distances in Online Verification for Motion Planning of Autonomous Vehicles, Master Thesis.
    • Christian Baumann: Risk Assessment of Traffic Scenarios by Developing a Criticality Metric and Conduction a User Study, Master Thesis.
    • Sebastian Kaster: Online Prediction of Vehicles and Pedestrians for Guaranteed Motion Satey of Autonomous Vehicles, Master Thesis.
  • Completed:
    • Maxime Allard: Traffic Rules at Intersections for Set-Based Prediction, Bachelor Thesis, 2017.
    • Mona Beikirch: Safe Evasive Maneuvers in Urban Environments, Master Thesis, 2017.
    • Hannes Bibel: Interaction of Traffic Participants in Set-Based Prediction, Master Thesis, WS 2017.
    • Vanessa Bui: Evaluating the Drivable Area considering Set-Based Prediction, Bachelor Thesis, 2017.
    • Lukas Braunstorfer: Risk Assessment of Traffic Scenarios, Bachelor Thesis, 2017.
    • Alexander Gaul: Over-Approximative Occupancy Prediction Considering Off-Tracking, Bachelor Thesis, 2018.
    • Philip Meyersieck: Falsification of Adaptive Cruise Control Systems in Automated Driving, Master Thesis, 2018.
    • Diana Papyan: Vehicle Turning Constraints for Set-Based Prediction, Master Thesis, 2017.
    • Fabian Schönert: Online Verification of Autonomous Driving in Parking Scenarios, Master Thesis, 2018.
    • Lukas Streit: Web-Based Benchmark for Trajectory Planning of Autonomous Vehicles, Bachelor Thesis, 2018.
    • Stefan Urban: Evaluation of Set-Based Prediction using Real-World Measurement Data, Master Thesis, 2018.
    • Lukas Willinger: Sensor Limitations in Set-Based Traffic Prediction, Bachelor Thesis, 2017.