Parallel Distributional Deep Reinforcement Learning for Mapless Navigation of Terrestrial Mobile Robots

TL;DR

This paper presents parallel distributional Deep-RL methods for mapless terrestrial robot navigation, demonstrating improved decision-making and generalization in simulation and real-world tests.

Contribution

It introduces novel parallel distributional actor-critic algorithms for robot navigation using laser data, advancing Deep-RL capabilities in real environments.

Findings

Enhanced navigation performance over non-distributional methods

Improved spatial generalization in real scenarios

Successful deployment in Gazebo and real-world environments

Abstract

This paper introduces novel deep reinforcement learning (Deep-RL) techniques using parallel distributional actor-critic networks for navigating terrestrial mobile robots. Our approaches use laser range findings, relative distance, and angle to the target to guide the robot. We trained agents in the Gazebo simulator and deployed them in real scenarios. Results show that parallel distributional Deep-RL algorithms enhance decision-making and outperform non-distributional and behavior-based approaches in navigation and spatial generalization.