Bipedal Walking Robot Driven by Elastic Cables

Abstract

This paper presents a design concept and analysis of a bipedal walking robot with a novel type of actuation using elastic cables. Each leg has 6-dof, the trunk has 3-dof, and each arm has 1-shoulder-dof. Conventional walking robots consist of joint-attached drives at revolute joints. This yields relatively heavy legs and arms with high moments of inertia, which makes balancing robot dynamic walking difficult due to the high inertial forces of distal segments. The cable-based actuation system is designed for the most kinetically-active biped segments, such as lower legs. These consist of DC motors located on the trunk, elastic cables (with serially-connected springs) and cable routing with specially-designed pulleys. Since the trunk segment accelerations are significantly lower than the leg segments accelerations, it is expected that the overall energy required by the cable-actuated robot is significantly lower than the energy input to a directly-actuated biped. Another novelty in the biped actuation system design is the use of elastic rather than non-elastic cables, for two reasons: smoothing out the sharp impulses due to the foot-ground collision and reduction of the number of motors to actuate each joint. Non-elastic cable-based drives require each cable to be pulled by a separate motor, which would double the number of motors and increase the weight. This problem can be solved using elastic cables and specially-shaped pulleys to reduce the number of motors with a slight increase in controller complexity. The bipedal walking robot architecture with cable drives mimics the human body architecture, where the hip joint is a 3-dof spherical joint, 1-dof knee joint, and 2-dof ankle joint. The architecture is more compact compared to the conventional joint attached drive architecture, wherein all revolute joints are separated. Based on the kinematic and dynamic analysis of the robot, a controller is designed and the perturbation robustness tested. A feedback linearization controller design is used, requiring system dynamics knowledge. Steps toward hardware implementation have been made, since we have implemented an elastic cable actuation system on a robotic cat prior to the concept design for the bipedal robot. The difficulties are discussed, including future plans for improvements and hardware testing.