DeepMNavigate: Deep Reinforced Multi-Robot Navigation Unifying Local & Global Collision Avoidance

TL;DR

DeepMNavigate introduces a deep reinforcement learning-based algorithm for multi-robot navigation that effectively combines local and global information, demonstrating superior performance in dense, complex environments with many agents.

Contribution

It presents a novel DRL-based multi-robot navigation algorithm that unifies local and global collision avoidance using a CNN and multi-scenario training.

Findings

Outperforms prior learning and geometric methods in dense scenarios

Handles raw sensor data directly for local observations

Effective in environments with narrow passages and many agents

Abstract

We present a novel algorithm (DeepMNavigate) for global multi-agent navigation in dense scenarios using deep reinforcement learning (DRL). Our approach uses local and global information for each robot from motion information maps. We use a three-layer CNN that takes these maps as input to generate a suitable action to drive each robot to its goal position. Our approach is general, learns an optimal policy using a multi-scenario, multi-state training algorithm, and can directly handle raw sensor measurements for local observations. We demonstrate the performance on dense, complex benchmarks with narrow passages and environments with tens of agents. We highlight the algorithm's benefits over prior learning methods and geometric decentralized algorithms in complex scenarios.