Spatial Intention Maps for Multi-Agent Mobile Manipulation

TL;DR

This paper introduces spatial intention maps, a novel representation for multi-agent vision-based deep reinforcement learning, which enhances coordination and cooperation among decentralized mobile manipulators in various tasks.

Contribution

The work proposes spatial intention maps as a new intention representation that improves multi-agent coordination by aligning intentions with visual observations and spatial action frameworks.

Findings

Improved task performance with spatial intention maps

Enhanced cooperative behaviors like object passing and collision avoidance

Effective across heterogeneous robot teams with different abilities

Abstract

The ability to communicate intention enables decentralized multi-agent robots to collaborate while performing physical tasks. In this work, we present spatial intention maps, a new intention representation for multi-agent vision-based deep reinforcement learning that improves coordination between decentralized mobile manipulators. In this representation, each agent's intention is provided to other agents, and rendered into an overhead 2D map aligned with visual observations. This synergizes with the recently proposed spatial action maps framework, in which state and action representations are spatially aligned, providing inductive biases that encourage emergent cooperative behaviors requiring spatial coordination, such as passing objects to each other or avoiding collisions. Experiments across a variety of multi-agent environments, including heterogeneous robot teams with different abilities (lifting, pushing, or throwing), show that incorporating spatial intention maps improves performance for different mobile manipulation tasks while significantly enhancing cooperative behaviors.