Model-based auxiliary input control: development and experimental analysis

A form of model-based control for robotic manipulators is proposed and experimentally evaluated. Algorithm development was motivated by experimental evaluation of an improved Lyapunov-based model reference adaptive control (LB-MRAC) technique. Those evaluations exposed the limitations of a decentralized LB-MRAC approach and the performance improvement potential of auxiliary inputs. Model-based auxiliary input control (MBAIC) is formed by augmenting the fixed-gain PD feedback loop of a normal model-based algorithm with the auxiliary input of the LB-MRAC. Feedback gain adaptation is not necessary. The tracking performance of several forms of LB-MRAC and MBAIC was evaluated over the standard PUMA test suite. Without feedforward compensation LB-MRAC tracking accuracy is inferior to that of a fixed-gain model-based controller. Addition of an auxiliary input significantly enhances model-based controller tracking accuracy and eliminates the large endpoint error previously associated with operation in uncertain payload configurations