Vibration‐based motion generation without any sensor or microprocessor

摘要

Abstract This article describes a decentralized controller that generates periodic motion patterns of general mechanical systems with multiple actuators, and its extremely simple implementation. The controller synchronizes the coupled dynamics of an uncertain body, the environment, and the actuators to its natural vibration modes autonomously. The major contribution of this method is that it makes possible decentralized and adaptive motion generation with multiple actuators without any sensor, microprocessor, or high‐torque and high‐precision motor. The proposed implementation of the decentralized controller is composed of a crank arm and low‐torque DC motor, and we focus on the intrinsic dynamics of the DC motor. When a reaction torque from the environment is applied to the motor shaft, the rotational speed of the motor will be changed passively. In our approach, we exploit these dynamics as a feedback controller that adjusts the phase of the DC motor. Experiments and simulations validate the fundamental features of the proposed method with a spring‐mass system, and the effects of the initial phase difference and improvement of the convergence property are reported. Moreover, an application example with a nonlinear legged robot simulation is demonstrated.