Cyborg-swarm cooperation and game via affective-based brain-machine interface

The integration of biological organisms with robotic systems has enabled hybrid cyborg platforms that combine biological sensory agility with electromechanical precision. However, existing cyborg systems predominantly rely on unidirectional stimulus-driven control, treating animals as bio-actuators while neglecting their intrinsic cognitive states. To bridge this gap, we present a closed-loop cyborg-swarm architecture that utilizes the animal's internal affective state (fear) as a high-level trigger to modulate robotic swarm strategies. Specifically, we developed a lightweight, real-time wireless brain-machine interface (BMI) to record local field potentials from the mouse basolateral amygdala. To ensure robust decoding in freely moving subjects, we implemented a dual-threshold detection algorithm that identifies fear states based on elevated [Formula: see text]-band power (15-30 Hz) and suppressed high-frequency noise, effectively rejecting motion artifacts. This decoded intent drives a dual-mode control framework: under baseline conditions, the system operates in a proportional-integral-derivative (PID)-based Exploration Mode; upon detection of fear, it autonomously switches to an Interaction Mode governed by Multi-Agent Deep Deterministic Policy Gradient. In this mode, a heterogeneous robotic swarm (comprising a MouseBot and an ally micro aerial vehicle (MAV)) executes coordinated adversarial defense strategies against an enemy MAV. Experimental results in a search-interference game demonstrate that biological affective signals can successfully trigger millisecond-level control authority switching, enabling the emergence of complex bio-machine cooperative behaviors. This work marks a paradigm shift from physical-level interaction to cognitive-level bio-hybrid cooperation, validating a scalable framework for emotion-modulated cyborg swarms.