Robust PIDNet Control of a Dual-Actuator Thrust Vectoring Platform

摘要

This paper presents a robust controller for attitude tracking on a dual-actuator thrust vectoring platform. The control objective is to track a desired thrust direction on the unit sphere using two linear actuators connected through a universal joint. The proposed framework combines a bounded nonlinear PD action with a radial basis function (RBF) network that generates state-dependent integral compensation through online weight adaptation, relying solely on tracking error and its time derivative from actuator-displacement measurements. The control law is formulated as a coupled multiple-input--multiple-output structure, enabling the RBF network to compensate cross-channel effects, including direction-dependent friction and geometric coupling, with model-free implementation. Stability is established through a convex Lyapunov function whose gradient is naturally bounded along the error manifold. An H-infinity gain bound is derived for the extended error, with a local linearized interpretation in terms of tracking accuracy. The proposed controller is validated experimentally on a two-degree-of-freedom gimbal rig, achieving improved performance relative to a conventional PID and to a super-twisting controller, with online adaptation reducing the ITNE by 10% over the non-adaptive baseline at no additional control effort.