Image-based Regularization for Action Smoothness in Autonomous Miniature Racing Car with Deep Reinforcement Learning

TL;DR

This paper introduces image-based regularization (I-RAS) and impact ratio control to enhance action smoothness in autonomous miniature racing cars using deep reinforcement learning, significantly improving success rates and real-world performance.

Contribution

It proposes a novel image-based regularization method (I-RAS) and an adaptive impact ratio control to address jerky actions in high-dimensional control tasks.

Findings

Success rate increased from 59% to 95%.

Outperformed other methods in real-world track tests.

Won the 2022 AWS DeepRacer Final Championship Cup.

Abstract

Deep reinforcement learning has achieved significant results in low-level controlling tasks. However, for some applications like autonomous driving and drone flying, it is difficult to control behavior stably since the agent may suddenly change its actions which often lowers the controlling system's efficiency, induces excessive mechanical wear, and causes uncontrollable, dangerous behavior to the vehicle. Recently, a method called conditioning for action policy smoothness (CAPS) was proposed to solve the problem of jerkiness in low-dimensional features for applications such as quadrotor drones. To cope with high-dimensional features, this paper proposes image-based regularization for action smoothness (I-RAS) for solving jerky control in autonomous miniature car racing. We also introduce a control based on impact ratio, an adaptive regularization weight to control the smoothness constraint, called IR control. In the experiment, an agent with I-RAS and IR control significantly improves the success rate from 59% to 95%. In the real-world-track experiment, the agent also outperforms other methods, namely reducing the average finish lap time, while also improving the completion rate even without real world training. This is also justified by an agent based on I-RAS winning the 2022 AWS DeepRacer Final Championship Cup.