COLONY: Trust-Based Multi-Agent Task Allocation Under Fatigue
September 15, 2026·
,·
0 min read
E. Outland
Bennet Outland
V. Arya
Abstract
This paper solves a multi-agent, multi-task allocation problem wherein the agents have uncertainty in their capabilities and their capabilities degrade upon use. We consider a large, complex multi-agent task broken down into atomic sub-tasks and arranged via a signal temporal logic (STL) directed task graph. The algorithm, Capability-based Optimization through Local Observation, Outcome tracking, and Networked sYstems (COLONY), is introduced to: learn the dynamic capabilities of the agents via trust-based consensus, allocate tasks to agents or groups of agents, and do so in a decentralized manner. This approach is demonstrated on Mars rover survey, satellite repair, and lunar habitat construction problems.
Type
Publication
Forthcoming
Authors
Graduate Researcher
Outland is a doctoral student at the University of Colorado Boulder, Smead Department of Aerospace Engineering Sciences. His research is in differential game theory, safe multi-agent systems, and computational astrodynamics, with a focus on algorithms that are tractable enough to run onboard while providing formal guarantees.