Frugal model predictive control and active disturbance rejection for laser beam steering systems

关键词:  Frugal model predictive control (FMPC) · Active disturbance rejection control (ADRC) · Laser beam steering system (LBS) · Real-time application · Constrained systems

Key words:  Frugal model predictive control (FMPC) · Active disturbance rejection control (ADRC) · Laser beam steering system (LBS) · Real-time application · Constrained systems

摘要:

This paper aims to fuse two well-established and, at the same time, opposed control techniques, namely, model predictive control (MPC) and active disturbance rejection control (ADRC), to develop a dynamic motion controller for a laser beam steering system. The proposed technique uses the ADRC philosophy to lump disturbances and model uncertainties into a total disturbance. Then, the total disturbance is estimated via a discrete extended state disturbance observer (ESO), and it is used to (1) handle the system constraints in a quadratic optimization problem and (2) injected as a feedforward term to the plant to reject the total disturbance, together with the feedback term obtained by the MPC. The main advantage of the proposed approach is that the MPC is designed based on a straightforward integrator-chain model such that a simple convex optimization problem is performed. Several experiments show the real-time closed-loop performance regarding trajectory tracking and disturbance rejection. Owing to simplicity, the self-contained approach MPC+ESO becomes a Frugal MPC, which is computationally economical, adaptable, efficient, resilient, and suitable for applications where on-board computational resources are limited.

Abstract:

This paper aims to fuse two well-established and, at the same time, opposed control techniques, namely, model predictive control (MPC) and active disturbance rejection control (ADRC), to develop a dynamic motion controller for a laser beam steering system. The proposed technique uses the ADRC philosophy to lump disturbances and model uncertainties into a total disturbance. Then, the total disturbance is estimated via a discrete extended state disturbance observer (ESO), and it is used to (1) handle the system constraints in a quadratic optimization problem and (2) injected as a feedforward term to the plant to reject the total disturbance, together with the feedback term obtained by the MPC. The main advantage of the proposed approach is that the MPC is designed based on a straightforward integrator-chain model such that a simple convex optimization problem is performed. Several experiments show the real-time closed-loop performance regarding trajectory tracking and disturbance rejection. Owing to simplicity, the self-contained approach MPC+ESO becomes a Frugal MPC, which is computationally economical, adaptable, efficient, resilient, and suitable for applications where on-board computational resources are limited.