Brain Emotional Sliding Mode Control for a Heterogeneous Multi-agent System.

The ability to adapt to a continuously changing environment and robustness to disturbances and uncertainties are always crucial features for a controller. In the last decades, an increasing interest in modeling the Limbic system in the human brain led to the development of the Brain Emotional Learning-Based Intelligent Controller (BELBIC) in order to cope with unknown dynamics. The novelty brought by this system, compared to a Neural Adaptive, is the ability to have a memory of past experiences, learn from them and adapt itself, providing a faster response when the same scenario is proposed again. In this thesis, the BELBIC controller is used to control a heterogeneous Multi-Agent System composed of a fleet of drones following a mobile robot leader, both for consensus and formation purposes. At first, the biological review of the Limbic system, which inspires the BELBIC, is described to understand its behavior and thus its computational model developed in the literature. This model is obviously simpler than the real Limbic system yet contains its main crucial parts as \textit{Thalamus, Sensory Cortex, Orbitofrontal Cortex, and Amygdala}. The role of the sensory signal is to provide information to the Amygdala and Orbitofrontal Cortex and this is done by feeding these areas with knowledge about the tracking error. This information is carried by the \textit{sliding surface} and hence, a theory review is provided in order to understand its ability to deal with uncertainties. Throughout this thesis, advanced types of sliding surfaces are considered and analyzed. Finally, the results of the heterogeneous Multi-Agent System are presented, showing that the BELBIC provides better performances compared to other types of neural adaptive control.