We present an algorithmic approach to task deconfliction for teams of autonomous agents in communication-limited environments when the teams produce schedules with strong cross-schedule dependencies without the knowledge of the existence of other teams attempting to prosecute the same task set. Once agents discover others are attempting to prosecute the same task, they deconflict the task by trading assigned tasks. The method to determine the agents and tasks to trade in an efficient manner we term the Generous Agent Algorithm (GAA). Through formal analysis and simulation we show that teams utilizing the GAA can observe significant reductions in schedule end times and mission fuel requirements. We compare the GAA’s performance with a greedy deconfliction algorithm and a centralized scheduling algorithm to show that the GAA improves performance in situations when the number of agents is large within limited-communication environments.

当团队在不知道其他团队试图起诉同一任务集的情况下产生具有强烈跨时间表依赖性的时间表时，我们会在通信受限的环境中为自治代理团队的任务冲突提供一种算法方法。一旦代理发现其他人正在尝试起诉同一任务，他们就会通过交易分配的任务来使任务冲突。确定代理和任务以有效方式进行交易的方法，我们称之为“慷慨代理算法”（GAA）。通过正式的分析和模拟，我们表明，利用GAA的团队可以观察到计划结束时间和任务燃料需求的大幅减少。