Fault-Tolerant Multi-Robot Coordination with Limited Sensing within Confined Environments

TL;DR

The paper addresses fault tolerance for multi-robot coordination in confined environments where sensing and communication are limited. It introduces Active Contact Response (ACR), a decentralized method that uses a transient space-time contact map to infer the presence of a faulty, stationary robot and bias active robots to reposition it, thereby reducing obstruction during collective excavation. Experimental results show that ACR significantly improves task throughput and mitigates congestion compared to a baseline that lacks contact-based bias, though edge-case scenarios and scalability remain challenging. This work demonstrates that local, physical interactions can enable robust, scalable fault-tolerant coordination in resource-constrained, extreme environments, with potential applicability to larger teams and other multi-agent systems.

Abstract

As robots are increasingly deployed to collaborate on tasks within shared workspaces and resources, the failure of an individual robot can critically affect the group's performance. This issue is particularly challenging when robots lack global information or direct communication, relying instead on social interaction for coordination and to complete their tasks. In this study, we propose a novel fault-tolerance technique leveraging physical contact interactions in multi-robot systems, specifically under conditions of limited sensing and spatial confinement. We introduce the "Active Contact Response" (ACR) method, where each robot modulates its behavior based on the likelihood of encountering an inoperative (faulty) robot. Active robots are capable of collectively repositioning stationary and faulty peers to reduce obstructions and maintain optimal group functionality. We implement our algorithm in a team of autonomous robots, equipped with contact-sensing and collision-tolerance capabilities, tasked with collectively excavating cohesive model pellets. Experimental results indicate that the ACR method significantly improves the system's recovery time from robot failures, enabling continued collective excavation with minimal performance degradation. Thus, this work demonstrates the potential of leveraging local, social, and physical interactions to enhance fault tolerance and coordination in multi-robot systems operating in constrained and extreme environments.

Figures (4)

• Figure 1: A model constrained and confined environment for multi-robot collective excavation of cohesive granular media. Robots respond to environmental cues using onboard sensors. Robot length is 32 cm.
• Figure 2: (A) Block diagram of the individual robot controller based on Finite State Machine. $P_e$ is the probability of switching to Active Mode. $P_e'$ is the probability of switching to Conservative Mode. $P_r$ is the probability of "giving up" upon detecting contact with another robot. $R_c$ is the probability that the contact is from a faulty robot. (B) Schematic of experimental setup showing three active robots (green, blue, and red), and one faulty robot (yellow). The origin of coordinate system is indicated.
• Figure 3: 1. Comparison of excavation performance between the Baseline and ACR methods across three trials. Solid-dotted and dash-dotted lines represent the mean excavated pellets, with shaded areas indicating the standard deviation. 2. Final configuration of the faulty robot after each trial: (A) Three Baseline trial, and (B) Three ACR trials. .
• Figure 4: 1. Robot trajectories tracked overtime for (A) Baseline method, and (B) Active Contact Response (ACR) method. Tunnel spans from 0 cm (home area) to 300 cm (digging area). 2. Top-view snapshots of the ACR experiment: (A) Initial configuration of faulty robot obstructs traffic flow, resulting in clustering at the middle of the tunnel. (B) Subsequently, active robots update their collision map and maneuver around the faulty robot to reach the excavation area. (C) Finally, active robots effectively reposition faulty robot into a less obstructive configuration through active pushing and nudging.