Abstract
Emotional models have demonstrated fast response times, quick adaptation, and promising performance, particularly in uncertain and complex environments. However, these results have been for single agents, and the social aspect of emotion-based architectures in multi-agent control systems (MACS) has yet to be adequately explored. MACS comprises a class of problems with considerably higher operational and environmental uncertainties due to agent variations, larger state-space dimensions, and limited communication with neighbors, thereby making its design and theoretical analysis considerably more challenging. Here, we propose an emotion-based cooperative adaptive neuro-controller (ECANC) for such higher-order heterogeneous, nonlinear leader-follower systems with entirely unknown dynamics, local communication, and limited access to the leader's state. This innovative approach is consistent with the non-decreasing adaptation law of the amygdala in the brain, opening the possibility of further inspiration from nature and posing a considerably higher challenge for controller design and stability analysis. For this purpose, we employ the continuous radial basis emotional neural network (CRBENN) paradigm in the MAS domain to approximate the unknown dynamics. We further employ artificial potential functions (APFs) to model agents' interactions, enabling a more straightforward mathematical analysis than graph-based methods. A new control law addresses local interactions among agents and the stability of the leader-follower system and is rigorously derived from Lyapunov stability theory. Simulation results demonstrate that ECANC achieves lower tracking error with less control effort, particularly when handling sudden disturbances, compared with two competing approaches.