Bridging Probabilistic Inference and Behavior Trees: An Interactive Method for Adaptive Collaborative Behavior Decision-Making of Multi-UAVs

This paper presents an interactive inference behavior tree (IIBT) framework, integrating behavior trees (BTs) with interactive inference based on the free energy principle for distributed decision-making in multi-UAV (unmanned aerial vehicle) systems. The proposed IIBT framework enhances conventional BTs by incorporating probabilistic inference, enabling online joint planning and execution among multiple UAVs. The framework maintains full compatibility with standard BT architectures, allowing seamless integration into existing UAV control systems. In this framework, cooperative behavior is modeled as a free-energy minimization process, where each UAV dynamically updates its preference matrix based on perceptual inputs and peer intentions, achieving adaptive coordination in dynamic and partially observable environments. The validation tasks, including cooperative navigation in uncertain environments and task coordination, directly mirror the decision-making and coordination challenges faced in UAV missions. Experimental results demonstrate that the IIBT framework achieves a reduction of over 70% in BT node complexity while maintaining robust, interpretable, and adaptive cooperative behavior in uncertain environments.