Input Allocation for Partially-Identified Redundant Control and Its Application to Precision Motion Systems

摘要

System identification for all input channels in an input redundant system may be time-consuming and impractical. In this paper, a control allocation approach is proposed for redundant systems with generic input constraints, where only part of the input-output relationships are required to be identified. The unidentified input-output relationships are estimated online using an expectation maximization (EM) algorithm, and controller allocation is carried out using quadratic programming based on generic linear input constraints. Simulation studies based on robotic welding and over-actuated motion system are provided to show that the unidentified input channels can be well estimated and used to increase disturbance rejection capability and trajectory tracking performance, as compared to using only the constrained known input channels.