This paper considers the problem of multi-robot safe mission planning in uncertain dynamic environments. This problem arises in several applications including safety-critical exploration, surveillance, and emergency rescue missions. Computation of a multi-robot optimal control policy is challenging not only because of the complexity of incorporating dynamic uncertainties while planning, but also because of the exponential growth in problem size as a function of the number of robots. Leveraging recent works obtaining a tractable safety maximizing plan for a single robot, we propose a scalable two-stage framework to solve the problem at hand. Specifically, the problem is split into a low-level single-agent planning problem and a high-level task allocation problem. The low-level problem uses an efficient approximation of stochastic reachability for a Markov decision process to handle the dynamic uncertainty. The task allocation, on the other hand, is solved using polynomial-time forward and reverse greedy heuristics. The safety objective of our multi-robot safe planning problem allows an implementation of the greedy heuristics through a distributed auction-based approach. Moreover, by leveraging the properties of the safety objective function, we ensure provable performance bounds on the safety of the approximate solutions proposed by these two heuristics. Our result is illustrated through case studies.
翻译:本文探讨了在不确定的动态环境中多机器人安全飞行任务规划的问题。 这个问题出现在几个应用中, 包括安全临界勘探、 监视和紧急救援任务。 计算多机器人最佳控制政策具有挑战性, 不仅因为在规划过程中纳入动态不确定性的复杂性,而且由于机器人数目的函数作用,问题规模的指数性增长。 利用最近的工作为单一机器人获得可移动的安全最大化计划,我们提出了一个可扩缩的两阶段框架来解决手头的问题。 具体地说, 问题被分为一个低层次的单一剂规划问题和一个高层次的任务分配问题。 低层次的问题使用高效的随机可达性近似, 用于马尔科夫决策程序处理动态不确定性。 另一方面, 任务分配是使用多元时间前行和逆向贪婪的超自然论。 我们多机器人安全规划问题的安全目标使得通过分散的拍卖方法可以实施贪婪的超自然论。 此外, 利用安全客观功能的特性的高效近似近似近似近似近似可达标, 我们通过这些安全性结果的研究确保了我们提出的近似性结果。