Robust Decentralized Abstractions for Multiple Mobile Manipulators

TL;DR

This paper presents a decentralized control framework for multiple mobile manipulators that ensures collision avoidance, connectivity, and high-level planning compatibility using local sensing and finite transition systems.

Contribution

It introduces a novel decentralized abstraction method for multi-agent mobile manipulators with 2nd order dynamics, enabling scalable high-level planning.

Findings

Simulation results verify the effectiveness of the decentralized controllers.

The approach guarantees collision avoidance and connectivity maintenance.

Finite transition systems successfully abstract multi-agent motion for planning.

Abstract

This paper addresses the problem of decentralized abstractions for multiple mobile manipulators with 2nd order dynamics. In particular, we propose decentralized controllers for the navigation of each agent among predefined regions of interest in the workspace, while guaranteeing at the same time inter-agent collision avoidance and connectivity maintenance for a subset of initially connected agents. In that way, the motion of the coupled multi-agent system is abstracted into multiple finite transition systems for each agent, which are then suitable for the application of temporal logic-based high level plans. The proposed methodology is decentralized, since each agent uses local information based on limited sensing capabilities. Finally, simulation studies verify the validity of the approach.