Dynamic Deep ESN-Enhanced PID Optimization and Incremental Saturation Control Allocation for UAVs Under Actuator Faults

In this paper, a novel design method combining proportional-integral-derivative (PID) optimization with an incremental saturation control allocation (ISCA) is proposed for unmanned aerial vehicles (UAVs) to accommodate actuator faults. Firstly, an echo state network (ESN) with deep time-varying reservoir parameters (DTVRP-ESN) is leveraged to approximate system dynamics and optimize PID parameters. The output weights and deep reservoir parameters of the DTVRP-ESN are synchronously updated based on the approximation and tracking errors to accelerate the computational process. Secondly, actuator faults are addressed by the proposed ISCA method, which considers both the direction of change and the cost of reaching saturation boundaries for unsaturated actuators. By sequentially saturating the most easily saturated actuators along the direction specified by the PID controller, matrix inversion calculations are avoided, and simple algebraic computations are used to quickly obtain optimal solutions. Finally, the effectiveness of the proposed control scheme is demonstrated through a numerical example.