Conflict Resolution for Heterogeneous Teams in Communication-limited Environments: A Generous Agent Approach

摘要

We present a schedule deconfliction algorithm to enhance cooperation among heterogeneous teams of autonomous agents in situations where communication is severely limited. The proposed solution, which we call the Generous Agent Algorithm (GAA), encodes human-like negotiations during conflict resolution. Individual agents propose modified schedules, vacating their most costly task from their schedules, while simultaneously augmenting their schedules with another agent's vacated task. An act of generosity, as we define it, must not increase an agent's makespan (or schedule completion time) beyond the time set by the team. Thus the GAA ensures that schedule alterations are done without degrading a team's global makespan. We present a proof of bounds on an agent's makespan after an act of generosity, as well as simulation results demonstrating the efficacy of the GAA at reducing task-redundancy and decreasing the makespan.