Human-Inspired Multi-Agent Navigation using Knowledge Distillation

TL;DR

This paper introduces a novel framework for multi-agent navigation that employs knowledge distillation and reinforcement learning to enable agents to mimic human-like collision avoidance and goal-directed behaviors, surpassing previous methods.

Contribution

The paper presents a new approach combining knowledge distillation with reinforcement learning to teach agents human-like navigation skills in decentralized multi-agent environments.

Findings

Agents exhibit human-like trajectories in collision avoidance tasks.

Our method outperforms expert policies and non-distilled learning agents.

Agents generalize to new tasks not seen in demonstrations.

Abstract

Despite significant advancements in the field of multi-agent navigation, agents still lack the sophistication and intelligence that humans exhibit in multi-agent settings. In this paper, we propose a framework for learning a human-like general collision avoidance policy for agent-agent interactions in fully decentralized, multi-agent environments. Our approach uses knowledge distillation with reinforcement learning to shape the reward function based on expert policies extracted from human trajectory demonstrations through behavior cloning. We show that agents trained with our approach can take human-like trajectories in collision avoidance and goal-directed steering tasks not provided by the demonstrations, outperforming the experts as well as learning-based agents trained without knowledge distillation.