Evaluation of Safety Constraints in Autonomous Navigation with Deep Reinforcement Learning

TL;DR

This paper compares safe and unsafe deep reinforcement learning policies for autonomous navigation, demonstrating that safety constraints improve obstacle clearance and reduce collisions without compromising performance.

Contribution

It introduces a comparative analysis of safety-constrained versus unconstrained RL policies in autonomous navigation tasks.

Findings

Safe policy achieves greater obstacle clearance.

Safe policy results in fewer collisions.

Safety constraints do not impair overall navigation performance.

Abstract

While reinforcement learning algorithms have had great success in the field of autonomous navigation, they cannot be straightforwardly applied to the real autonomous systems without considering the safety constraints. The later are crucial to avoid unsafe behaviors of the autonomous vehicle on the road. To highlight the importance of these constraints, in this study, we compare two learnable navigation policies: safe and unsafe. The safe policy takes the constraints into account, while the other does not. We show that the safe policy is able to generate trajectories with more clearance (distance to the obstacles) and makes less collisions while training without sacrificing the overall performance.