A Probabilistic Framework for Imitating Human Race Driver Behavior

Personal use of this material is permitted. ACCEPTED JANUARY 2020 1 A Probabilistic Framework for Imitating Human Race Driver Behavior Stefan L ockel 1, Jan Peters 2, and Peter van Vliet 3 Abstract --Understanding and modeling human driver behavior is crucial for advanced vehicle development. However, unique driving styles, inconsistent behavior, and complex decision processes render it a challenging task, and existing approaches often lack variability or robustness. T o approach this problem, we propose Probabilistic Modeling of Driver behavior ( ProMoD), a modular framework which splits the task of driver behavior modeling into multiple modules. A global target trajectory distribution is learned with Probabilistic Movement Primitives, clothoids are utilized for local path generation, and the corresponding choice of actions is performed by a neural network. Experiments in a simulated car racing setting show considerable advantages in imitation accuracy and robustness compared to other imitation learning algorithms. The modular architecture of the proposed framework facilitates straightforward extensibility in driving line adaptation and sequencing of multiple movement primitives for future research. I NTRODUCTION R ELIABLE simulations are crucial for modern car development, allowing faster prototyping and a better cost efficiency for both, the design of single parts, and the testing of the overall vehicle performance. While vehicle dynamics have been studied and modeled extensively for decades and are well understood even in extreme driving situations [1], past research on modeling of human drivers did not lead to a clear result. Hence, dynamic vehicle simulations often use conventional controllers or reference maneuvers for standard driving situations with limited dynamics [2]. As these vehicle controllers incorporate human behavior only to a certain extent, which is especially important in extreme driving situations, additional simulations with a human driver in the loop (HDIL) are required in the current development process.