CrowdMove: Autonomous Mapless Navigation in Crowded Scenarios

TL;DR

CrowdMove introduces a generalized training framework for autonomous mapless navigation, enabling various mobile robots to safely operate in crowded, dynamic environments through a robust policy gradient approach.

Contribution

The paper presents a novel 3M training framework and a robust policy gradient algorithm for effective mapless navigation across multiple robots and scenarios.

Findings

Successfully navigates four types of mobile platforms in crowded environments

Demonstrates superior safety and efficiency in complex scenarios

Validates approach with real-world experiments and videos

Abstract

Navigation is an essential capability for mobile robots. In this paper, we propose a generalized yet effective 3M (i.e., multi-robot, multi-scenario, and multi-stage) training framework. We optimize a mapless navigation policy with a robust policy gradient algorithm. Our method enables different types of mobile platforms to navigate safely in complex and highly dynamic environments, such as pedestrian crowds. To demonstrate the superiority of our method, we test our methods with four kinds of mobile platforms in four scenarios. Videos are available at https://sites.google.com/view/crowdmove.