Pole/Zero Design of Agonist/Antagonist Actuation

Abstract

Objective: This brief analyzes the open-loop dynamics of coupled (dedicated) and decoupled tendon tension control methods in an agonist/antagonist (AA) actuator pair, a fundamental control unit used in orthopedic testbeds known as joint motion simulators (JMSs). Methods: A linear mathematical model of an AA actuator pair is derived. Transfer functions from tendon tension control signal to tendon tension and joint position are derived. Sources of key dynamics are explained. Results: The system’s dynamic model shows that both the dedicated and decoupled approaches have a low-frequency pole pair, a low-frequency zero pair, and a mid-range zero pair. A frequency domain identification of the flexion/extension axis of an existing elbow JMS validates the locations of these dynamics. The interaction between tendon tension control and joint position is shown to be controllable in decoupled control, but not in dedicated control. The bandwidth reduction due to the low-frequency pole pair and low-frequency zero pair are shown to be controllable in decoupled control, but not in dedicated control. Conclusion: Decoupled control is superior to dedicated control for AA actuator pairs in JMS designs because it reduces actuator interaction and has a larger tension control bandwidth. Significance: This analysis describes the sources of the dynamics seen in the open-loop frequency response of both methods and shows the superiority of the decoupled method in tension control bandwidth and in lack of interactions with position control.