Realtime Collision Avoidance for Mobile Robots in Dense Crowds using Implicit Multi-sensor Fusion and Deep Reinforcement Learning

TL;DR

This paper introduces CrowdSteer, a deep reinforcement learning-based collision avoidance system for mobile robots in dense crowds, utilizing multi-sensor fusion and sim-to-real transfer for real-time, smooth navigation.

Contribution

The paper presents a novel end-to-end learning approach combining multi-sensor perception and deep reinforcement learning for collision avoidance in dense environments, with successful real-world deployment.

Findings

Effective real-time collision avoidance in dense crowds

Significant improvements over prior methods in success rate and trajectory smoothness

Successful transfer from simulation to real-world robot navigation

Abstract

We present a novel learning-based collision avoidance algorithm, CrowdSteer, for mobile robots operating in dense and crowded environments. Our approach is end-to-end and uses multiple perception sensors such as a 2-D lidar along with a depth camera to sense surrounding dynamic agents and compute collision-free velocities. Our training approach is based on the sim-to-real paradigm and uses high fidelity 3-D simulations of pedestrians and the environment to train a policy using Proximal Policy Optimization (PPO). We show that our learned navigation model is directly transferable to previously unseen virtual and dense real-world environments. We have integrated our algorithm with differential drive robots and evaluated its performance in narrow scenarios such as dense crowds, narrow corridors, T-junctions, L-junctions, etc. In practice, our approach can perform real-time collision avoidance and generate smooth trajectories in such complex scenarios. We also compare the performance with prior methods based on metrics such as trajectory length, mean time to goal, success rate, and smoothness and observe considerable improvement.