This paper describes a stochastic static analysis of link angle driven by actuator bundles. We first demonstrate a movement of a link driven by actuator bundles in terms of high driving force. The variation of actuator parameters influences the movement of this mechanism. We confirm the link may shift and twist because of the variation of actuator parameters. We then introduce stochastic analysis to investigate the link angle considering the variation of actuator parameters. In the result, the variation of the link angle converges to zero when the number of actuator is large enough. In addition, we reinforce the result through dynamic simulation considering the variation of the parameters of actuators.

We introduce a robotic prototype of an arm with a loosely coupled joint, modeled on the human joint. A viscoelastic object functions as cartilage and soft actuators as muscles. First, we show that although viscoelastic object affords smooth movement owing to shift in the center of rotation, the repeat accuracy of the joint is poor under open-loop control. The repeat accuracy was much improved by visual feedback. Under P control, the prototype was shown to be highly robust against mechanical disturbance owing to its good mechanical compliance.

We describe here a mechanism for controlling angles of a human-like joint using length sensors in threedimensional (3D) space. This joint mechanism, which is called a loosely coupled mechanism, includes a viscoelastic object and soft actuators in place of the cartilage and muscles of a human arm. To confirm motion of the link using one length sensor, we constructed a prototype of the mechanism in two-dimensional (2D) space. Based on this prototype, we constructed a 3D loosely coupled mechanism with length sensors, and we were able to control two projecting angles of the 3D prototype. In addition, we propose an appropriate method of measurement to reduce errors in measurement due to the length sensors. Using this method, we found that, for each projecting plane, the errors were less than 1.0 deg in our 3D prototype.