Design and Development of a Passive Dynamic Inspired Kneed Biped Robot

Abstract

In nature, legged locomotion is achieved by consciously actuating critical degrees of freedom, resulting in compliant joint manipulation. Yet few actuated walking robots exploit anthropomorphic passive dynamics, induced by mechanical characteristics intrinsic within its design, in maintaining gait. Energy efficient bipedal robot paradigms that pursue a dynamically stable gait trajectory, and reject external disturbances, are still immature. While substantial research in fully actuated robots exists, the application of bio-inspired mechanisms for gait articulation remains nascent. A vision to create a stable, energy efficient and mechanically optimised kneed biped robot, incorporating effective mass distribution and actuators, serves as a foundation for research into adaptable locomotion. In pursuit of these ambitions, novel hardware and new software were developed. A linearly actuated platform attached to the hip manipulates robot anterior-posterior center of mass. Hip and knee joint actuation are facilitated by mechanisms that also function as clutches, permitting transmission disengagement during underactuation. A simple knee lock secures the shank during leg stance phase. Tailored electronics interface and drive electromechanical peripherals, while sensors are deployed at joints and feet for leg stance estimation. Custom “real-time” software middleware, and associated drivers dedicated for the Raspberry Pi microprocessor unit, were programmed for controlling and manipulating robot hardware. Various tests, including one-step motion, were conducted to better understand established capabilities and areas requiring improvement. Finally, suggestions for progressive postgraduate research on the robot are provided. The research realises design and development of a kneed biped robot. The robot demonstrates one-step motion with 0.19m step size, corresponding to 0.4rad between thighs. Each single step takes 0.5-0.6s. Initial outcomes provide an excellent platform for further upgrades, studying control strategies like reinforcement learning and making it walk in the future. These are preliminary steps towards realising a robot capable of representing human-like gaits, with dynamically similar mechanical structures, efficient mass distributions and joint torques.