MATE-BT: A Behavior-Tree-Based Framework for Hierarchically Decentralized Mission Autonomy in Heterogeneous Multi-Robot Teams
Multi-robot teams executing predefined missions must adapt to runtime changes and sustain mission management despite robot failures. This paper proposes MATE-BT (Mission, Allocation, and Task Execution with Behavior Trees), a hierarchically decentralized framework that separates mission execution, allocation, and task execution. An active onboard leader manages mission progression and team-level task assignment, with authority transferable to another leader-capable robot after failure. Interchangeable decentralized methods resolve target assignments, while each robot adapts its local behavior through planning and acting with behavior trees. Webots simulations and real-robot experiments demonstrate mission continuity despite new targets and robot failures, including active-leader failure. In the physical experiments, all MATE-BT computation runs onboard. Hardware-in-the-loop resource measurements further support onboard execution on Jetson Orin Nano and Raspberry Pi 5. By updating only the affected layers without full-plan regeneration, MATE-BT maintained a response time of approximately 130 ms with 30 robots and 120 targets. Comparisons of four target-assignment plugins reveal scenario-dependent trade-offs between mission completion time and total travel distance.