Multi-agent systems, by dividing complex problems into smaller tasks and assigning them to independent agents, provide an innovative and efficient approach to solving intricate issues. Due to advantages such as scalability, flexibility, parallel processing, and reliability, these systems have gained significant attention in various domains, including artificial intelligence, robotics, autonomous systems, transportation, intelligent traffic control, satellites, detection, surveillance, networked control systems, ecosystem management, environmental monitoring, and more. By replacing a complex system with multiple simpler systems, these frameworks enable the achievement of optimal solutions, adaptability to environmental changes, and enhanced efficiency. In some multi-agent systems, the complex behavior of agents and their interactions require nonlinear modeling. These models, with greater flexibility, are capable of accurately describing a wide range of phenomena and relationships among agents. The presence of time delays in communication, heterogeneity in agents’ internal dynamics, as well as uncertainties and disturbances with unknown bounds, poses significant challenges in the design and control of such systems. This thesis investigates the design of distributed controllers to achieve consensus or formation with a leader in high-order nonlinear and heterogeneous multi-agent systems with communication delays, under uncertainties and disturbances with unknown bounds. An adaptive sliding mode control approach is employed to address this issue. This method, with its adaptive mechanism, can handle a broad range of variable and unknown uncertainties and disturbances without requiring precise prior information about their bounds. To overcome the inherent challenges of traditional sliding mode controllers, such as chattering and constraints due to system relative degree, a super-twisting control algorithm is utilized. Additionally, this research focuses on achieving optimal consensus or formation in distributed systems by proposing a distributed adaptive sliding mode control method based on reinforcement learning. This approach not only ensures system stability and strong robustness against uncertainties and disturbances but also facilitates performance optimization. Finally, to eva‎luate the efficiency of the proposed method, a simulation example is presented.

فريد شيخ الاسلام , مجدالدين نجفي

جعفر قيصري , جواد عسگري مارناني