Learning Vision-based Reactive Policies for Obstacle Avoidance

TL;DR

This paper introduces a unified framework for vision-based obstacle avoidance in robotic manipulators, focusing on learning reactive policies that connect visual perception with motion control to improve safety and responsiveness.

Contribution

It proposes a novel method for learning reactive obstacle avoidance policies that integrate perception and motion, advancing the state of the art in robotic obstacle avoidance.

Findings

High success rate in goal-reaching tasks with obstacles

Efficient learning of stable obstacle avoidance strategies

Maintains closed-loop responsiveness for human-robot interaction

Abstract

In this paper, we address the problem of vision-based obstacle avoidance for robotic manipulators. This topic poses challenges for both perception and motion generation. While most work in the field aims at improving one of those aspects, we provide a unified framework for approaching this problem. The main goal of this framework is to connect perception and motion by identifying the relationship between the visual input and the corresponding motion representation. To this end, we propose a method for learning reactive obstacle avoidance policies. We evaluate our method on goal-reaching tasks for single and multiple obstacles scenarios. We show the ability of the proposed method to efficiently learn stable obstacle avoidance strategies at a high success rate, while maintaining closed-loop responsiveness required for critical applications like human-robot interaction.