TY - GEN
T1 - Deep Drifting: Autonomous Drifting of Arbitrary Trajectories using Deep Reinforcement Learning
AU - Domberg, Fabian
AU - Wembers, Carlos Castelar
AU - Patel, Hiren
AU - Schildbach, Georg
N1 - Funding Information:
We sincerely thank RC vehicle enthusiast Troy from Road-sideRC and international RC drift pro-driver Timmy Woo for providing their expert knowledge and valuable feedback. This research was partly funded by Federal Ministry of Education and Research (BMBF) as part of the project ‘KI-Labor in Lübeck’. We gratefully acknowledge the additional support of this research by Elektronische Fahrwerksysteme (EFS) GmbH with the laboratory equipment, and by NVIDIA Corporation with the donation of a Jetson TX2 development kit. Furthermore, we thank Aleksandra Filippova and Ievgen Zhavzharov for their help with the hardware of the RC car.
Publisher Copyright:
© 2022 IEEE.
PY - 2022/5
Y1 - 2022/5
N2 - In this paper, a Deep Neural Network is trained using Reinforcement Learning in order to drift on arbitrary trajectories which are defined by a sequence of waypoints. In a first step, a highly accurate vehicle simulation is used for the training process. Then, the obtained policy is refined and validated on a self-built model car. The chosen reward function is inspired by the scoring process of real life drifting competitions. It is kept simple and thus applicable to very general scenarios. The experimental results demonstrate that a relatively small network, given only a few measurements and control inputs, already achieves an outstanding performance. In simulation, the learned controller is able to reliably hold a steady state drift. Moreover, it is capable of generalizing to arbitrary, previously unknown trajectories and different driving conditions. After transferring the learned controller to the model car, it also performs surprisingly well given the physical constraints.
AB - In this paper, a Deep Neural Network is trained using Reinforcement Learning in order to drift on arbitrary trajectories which are defined by a sequence of waypoints. In a first step, a highly accurate vehicle simulation is used for the training process. Then, the obtained policy is refined and validated on a self-built model car. The chosen reward function is inspired by the scoring process of real life drifting competitions. It is kept simple and thus applicable to very general scenarios. The experimental results demonstrate that a relatively small network, given only a few measurements and control inputs, already achieves an outstanding performance. In simulation, the learned controller is able to reliably hold a steady state drift. Moreover, it is capable of generalizing to arbitrary, previously unknown trajectories and different driving conditions. After transferring the learned controller to the model car, it also performs surprisingly well given the physical constraints.
M3 - Conference contribution
BT - 2022 International Conference on Robotics and Automation (ICRA)
PB - IEEE
ER -
