Dynamic multi-robot task allocation under uncertainty and temporal constraints

We consider the problem of dynamically allocating tasks to multiple agents under time window constraints and task completion uncertainty. Our objective is minimize number unsuccessful at end operation horizon. present a multi-robot allocation algorithm that decouples key computational challenges sequential decision-making uncertainty multi-agent coordination, addresses them in hierarchical manner. The lower layer computes policies for individual using dynamic programming with tree search, upper resolves conflicts plans obtain valid allocation. algorithm, Stochastic Conflict-Based Allocation (SCoBA), optimal expectation complete some reasonable assumptions. In practice, SCoBA computationally efficient enough interleave planning execution online. On metric successful completion, consistently outperforms baseline methods shows strong competitive performance against an oracle lookahead. It also scales well agents. validate our results over wide range simulations on two distinct domains: multi-arm conveyor belt pick-and-place multi-drone delivery dispatch city.