Abstract
This paper focuses on the decentralized control problem of multi-agent systems with multiple control objectives, including target tracking, velocity synchronization, collision avoidance, and obstacle avoidance. To enable the real-time application and optimization, a control scheme is constructed based on the idea of adaptive dynamic programming (ADP). The cost function is defined according to the control objectives and approximated by a critic neural network. An action neural network is applied to approximate the optimal controller, which is achieved by minimizing the cost function. Based on the deduced weight updating laws, the proposed control scheme can learn online with good adaptability to unknown environment. Finally, a simulation of seven agents moving in two-obstacle environment is conducted to show the validity of the intelligent controller.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have