Event-triggered optimal consensus for multi-agent systems: A prescribed time optimal scheme based on DRL

For the challenge of cooperative optimization of transient performance, energy consumption, and communication resources in multi-agent systems, this paper proposes an event-triggered prescribed time optimal consensus control scheme within the framework of deep reinforcement learning-based optimal backstepping. Firstly, a distributed event-triggered communication mechanism is proposed by designing the output sampling event-driven function with bandwidth sensing characteristics to realize the elastic adjustment of the communication load from the topological dimension. Subsequently, to find the optimal control solution for the co-optimization of stability and energy consumption, an optimal consensus control protocol is constructed based on the actor-critic neural network iterative learning algorithm and Bellman optimality theory. Furthermore, by introducing a time-varying gain scaling function, the Hamilton-Jacobi-Bellman equation analytical framework with explicit time-constrained characteristics is reformulated and an optimal consensus controller with a strict prescribed time convergence guarantee is derived that balances transient performance, energy consumption, and communication resources while attaining multi-agent system optimal consensus. Finally, the effectiveness of the proposed scheme is validated through comparative numerical simulations and a model simulation of a multi-agent electromechanical system.