Dynamic control of an artificial muscle arm

We propose an original structure composed of two pneumatic joints. Each joint is moved by two pneumatic cylinders in opposition, representing the agonist and the antagonist muscles. This structure has the aim of reproducing natural human joints in an arm. A dynamic model of the structure is proposed decoupling the dynamic effects of the skeletal (as interactions between segments) from the dynamic effects of the muscles involved. A nonlinear controller based on a computed torque method taking into account the actuator and the mechanical models is then presented to ensure dynamic tracking of position and force for this arm which may interact with the environment or cooperate with it. An analogy with the physiological muscle is proposed. The originality of the controller lays in the consideration of impedance behaviour at each joint during free and constrained tasks. It leads to continuous control laws between contact and non-contact phases which enable real time computations. The asymptotic stability is ensured using Popov criteria. Several simulations have been led to point out the performances of the agonist and antagonist actuators on the rigidity and on the position tracking accuracy.