Joints of conventional robot manipulators are controlled by PD feedback controllers, where accuracy and quick response are important. In contact task applications existing in human living environments, supervision from a higher-level controller with sophisticated sensor systems for watching contact with the environment and object is essential. Complicated control systems slow down the robot's working speed and increase the overall cost of the systems. This study proposes an alternative joint controller to simplify and reduce the cost of contact task control. The robot structure we intend to use is a musculoskeletal structure that can reduce joint friction like a human. This paper focuses on a controller that controls the speed of the proposed drive mechanism at low speed without using high-gain feedback control. Low-speed control would be useful for robots transitioning from a non-contact state to a contact state. The idea of our method is to manipulate the viscous friction coefficient by using a micro electromagnetic clutch and balance the friction at the target angular speed. This method does not rely on a position control loop, aiming for an equilibrium speed with low friction force. A model reference adaptive algorithm is used to identify the viscous coefficient online. We present theoretical results on the stability, and experimental results of validation.
Mitobe et al. (Thu,) studied this question.