Regret-Based Multi-Agent Coordination with Uncertain Task Rewards

TL;DR

This paper introduces a decentralized algorithm for multi-agent task allocation under uncertain rewards, minimizing worst-case loss, and scalable to hundreds of agents and tasks.

Contribution

It extends DCOP models to handle uncertain task rewards and proposes a novel Max-Sum based decentralized solution with iterative constraint generation.

Findings

Scalable to hundreds of agents and tasks.

Effectively minimizes worst-case loss under reward uncertainty.

Outperforms standard DCOP algorithms in uncertain environments.

Abstract

Many multi-agent coordination problems can be represented as DCOPs. Motivated by task allocation in disaster response, we extend standard DCOP models to consider uncertain task rewards where the outcome of completing a task depends on its current state, which is randomly drawn from unknown distributions. The goal of solving this problem is to find a solution for all agents that minimizes the overall worst-case loss. This is a challenging problem for centralized algorithms because the search space grows exponentially with the number of agents and is nontrivial for standard DCOP algorithms we have. To address this, we propose a novel decentralized algorithm that incorporates Max-Sum with iterative constraint generation to solve the problem by passing messages among agents. By so doing, our approach scales well and can solve instances of the task allocation problem with hundreds of agents and tasks.