Advanced

Dynamic path planning for collision avoidance in a robotized framework for autonomous driving verification

Johansson, Daniel (2019)
Department of Automatic Control
Abstract
Self-driving vehicles is a highly anticipated technology for increasing the safety and efficiency of automotive transportation systems by removing the risk of human errors. Volvo Car Corporation is determined to produce vehicles of full autonomy and highest safety within the near future. To achieve this goal, Volvo Cars is in parallel with the development of autonomous vehicles setting up a sophisticated pipeline for verifying and testing the autonomous driving functions. This thesis revolves around the last step of this pipeline, by implementing and further developing an algorithm for collision avoidance in a robotized framework for verification of autonomous driving where the functions are tested on real vehicles on a test track. The... (More)
Self-driving vehicles is a highly anticipated technology for increasing the safety and efficiency of automotive transportation systems by removing the risk of human errors. Volvo Car Corporation is determined to produce vehicles of full autonomy and highest safety within the near future. To achieve this goal, Volvo Cars is in parallel with the development of autonomous vehicles setting up a sophisticated pipeline for verifying and testing the autonomous driving functions. This thesis revolves around the last step of this pipeline, by implementing and further developing an algorithm for collision avoidance in a robotized framework for verification of autonomous driving where the functions are tested on real vehicles on a test track. The proposed algorithm used to achieve collision avoidance in the robotized test framework is the Bicycle Optimal Reciprocal Collision Avoidance (B-ORCA) algorithm. This algorithm uses a construct called Velocity Obstacle to predict imminent collisions between vehicles in a scenario and then calculates the optimal velocities to avoid the collisions in a collaborative manner. To evaluate the performance of the algorithm, a set of experiments were performed on the driving robots that will be used in the testing framework, both in simulations and on a real vehicle. The results from these experiments show that the current implementation of the B-ORCA algorithm guarantees accurate safe trajectories up to speeds of 50km=h. To support speeds above 50km=h, the simple Kinematic bicycle model currently used to calculate the trajectories has to be replaced with a more sophisticated motion model. This new model has to better model the lateral acceleration that, with too high values, was shown to be the main parameter that made the vehicle not follow the safe trajectories as desired. (Less)
Please use this url to cite or link to this publication:
author
Johansson, Daniel
supervisor
organization
year
type
H3 - Professional qualifications (4 Years - )
subject
report number
TFRT-6087
ISSN
0280-5316
language
English
id
8986376
date added to LUP
2019-08-30 09:30:41
date last changed
2019-08-30 09:30:41
@misc{8986376,
  abstract     = {Self-driving vehicles is a highly anticipated technology for increasing the safety and efficiency of automotive transportation systems by removing the risk of human errors. Volvo Car Corporation is determined to produce vehicles of full autonomy and highest safety within the near future. To achieve this goal, Volvo Cars is in parallel with the development of autonomous vehicles setting up a sophisticated pipeline for verifying and testing the autonomous driving functions. This thesis revolves around the last step of this pipeline, by implementing and further developing an algorithm for collision avoidance in a robotized framework for verification of autonomous driving where the functions are tested on real vehicles on a test track. The proposed algorithm used to achieve collision avoidance in the robotized test framework is the Bicycle Optimal Reciprocal Collision Avoidance (B-ORCA) algorithm. This algorithm uses a construct called Velocity Obstacle to predict imminent collisions between vehicles in a scenario and then calculates the optimal velocities to avoid the collisions in a collaborative manner. To evaluate the performance of the algorithm, a set of experiments were performed on the driving robots that will be used in the testing framework, both in simulations and on a real vehicle. The results from these experiments show that the current implementation of the B-ORCA algorithm guarantees accurate safe trajectories up to speeds of 50km=h. To support speeds above 50km=h, the simple Kinematic bicycle model currently used to calculate the trajectories has to be replaced with a more sophisticated motion model. This new model has to better model the lateral acceleration that, with too high values, was shown to be the main parameter that made the vehicle not follow the safe trajectories as desired.},
  author       = {Johansson, Daniel},
  issn         = {0280-5316},
  language     = {eng},
  note         = {Student Paper},
  title        = {Dynamic path planning for collision avoidance in a robotized framework for autonomous driving verification},
  year         = {2019},
}