Online payload estimation for the control of underactuated mechanical systems

This paper presents a payload estimation scheme for underactuated robotic manipulators with passive joints that are not driven by actuators. In the proposed scheme, only the payload, which can be quite uncertain when a robot performs various tasks, is estimated, because the manipulator's electrical and other mechanical parameters are generally known in advance. In comparison to other adaptive schemes for underactuated robotic manipulators, the proposed scheme produces satisfactory transient performance and also reduces the computational burden in real-time implementation. The proposed estimation law is also based on the theory of Variable Structure Systems. In contrast to existing adaptation laws that have an integral form, the proposed law estimates uncertain payload using lowpass filtering of a switching signal that is always bounded, which avoids the parameter-drifting problem that is often encountered when using the previous integral laws. Real-time experiments are conducted using an inverted pendulum and the experimental results demonstrate the feasibility and effectiveness of the proposed scheme.