A Benchmark Comparison of Imitation Learning-based Control Policies for Autonomous Racing

TL;DR

This paper benchmarks various imitation learning policies for autonomous racing, demonstrating that interactive imitation learning enhances performance and sample efficiency, thereby aiding reinforcement learning in both simulation and real-world scaled environments.

Contribution

It provides a comprehensive comparison of imitation learning methods for autonomous racing and shows the benefits of interactive techniques for improving reinforcement learning.

Findings

Interactive imitation learning outperforms traditional methods.

Imitation learning improves reinforcement learning sample efficiency.

Benchmarks establish a foundation for future autonomous racing research.

Abstract

Autonomous racing with scaled race cars has gained increasing attention as an effective approach for developing perception, planning and control algorithms for safe autonomous driving at the limits of the vehicle's handling. To train agile control policies for autonomous racing, learning-based approaches largely utilize reinforcement learning, albeit with mixed results. In this study, we benchmark a variety of imitation learning policies for racing vehicles that are applied directly or for bootstrapping reinforcement learning both in simulation and on scaled real-world environments. We show that interactive imitation learning techniques outperform traditional imitation learning methods and can greatly improve the performance of reinforcement learning policies by bootstrapping thanks to its better sample efficiency. Our benchmarks provide a foundation for future research on autonomous racing using Imitation Learning and Reinforcement Learning.