Virtual Bipedal Passive Walking Control Based on Target Dynamics

This paper introduces a target-dynamics based virtual passive bipedal walking control method without reference gaits or energy. The method maps the 2-DOF bipedal robot into a 1-DOF system through input-output linearization at first. Then, the inner-leg angle of the robot is selected as the sole control output to track the target dynamics — a harmonic oscillator, which is found to be the intrinsic characteristic of the passive bipedal walking through numerical simulations and thus any other reference trajectories, e.g. the passive gaits or the energy, are no longer required. Finally, a constant ankle torque and a Port-Controlled Hamilton system based ankle torque are selected as the control inputs respectively. Simulations on different grounds are conduct to evaluate the controller performance. The results show the controller can produce stable virtual passive bipedal walking gaits on varied terrains with high trajectory tracking accuracy and energy efficiency, and each of the two types of input torques exhibits distinct advantages in terms of energy efficiency and adaptability.