Self-tuning in integral sliding mode control with a Levenberg–Marquardt algorithm

• SMC in motion control context.
• Self-tuning algorithm for SMC parameters.
• Wafer scanner experiments.

In high-precision motion systems, tracking often comes with the problem of overshoot or poor settling behavior. To deal with this problem, a sliding mode controller with saturated integrator is studied. For large servo signals, sufficiently removed from the sliding surface, the controller will operate in PD mode. This limits the integrator buffer, and thus the overshoot. At the sliding surface (for small servo signals) the controller operates in PID mode with the aim to avoid steady-state error. Tuning of the controller parameters: the switching gain and the saturation length, strongly affects the tracking performance. For this reason, a self-tuning method is proposed. In the method, the required gradients are partly computed using models of the system and the controller. For the remaining part, sampled data is used to deal with disturbances and model uncertainties. With a Levenberg–Marquardt algorithm numerical problems associated with the gradient computations are avoided. The sliding mode controller with optimized parameters is implemented on an industrial wafer scanner to improve throughput during the wafer stage (chuck) exchange.