Adaptive control with neuro-adaptive disturbance rejection

This paper presents an adaptive disturbance rejection scheme which makes use of a neural model of the disturbance. Unknown disturbances may account for the reduction in the performance of a control system where precise tracking is required. These disturbances may be nonlinear and dynamic making the rejection problem difficult for traditional methods. Also the plant being controlled may be unknown, as the model may be inaccurate or the parameters may vary over time. Classical controllers may not be able to stabilize the system and meet performance requirements under these conditions. For this purpose, the scheme presented employs an adaptive controller in conjunction with an adaptive disturbance rejector which is based on a neural model of the unknown disturbance. Numerical simulations are included to show the benefit of the scheme in terms of tracking performance.