On the design of extended state observer-based robust finite controller: For underactuated robotic system with multiple sources of uncertainties

Abstract

Controlling an underactuated robot is always an important research and engineering issue, especially when the robot is suffering from multiple sources of uncertainties, such as unmodeled dynamics, external disturbance, and parameter uncertainties. To cope with these uncertainties in such uncertain nonlinear systems which is not fully-actuated, this paper proposes a control method that can actively estimate these uncertainties via the extended state observer (ESO), under the scheme of output-feedback control, the lumped uncertainties can be online estimated and actively compensated. Every joint of the underactuated robotic system can robustly reach the pre-given state in finite-time even though there are only fewer joints than the actual number of joints that can be controlled directly. The experimental results demonstrate the control process and validate that the proposed method is feasible for the studied underactuated robotic system.