Biological-inspired Rhythmic Motion & Environmental Adaptability For Quadruped Robot

Due to complex modeling, non-unique solutions and single-period planning, the traditional model-based motion-controlling approach is disadvantageous to produce quick steady motion and to adapt to unstructured environment for a multi-legged robot. In this thesis the rhythmic motion and environmental adaptability for a quadruped robot is studied by employing biological motion-controlling mechanisms and reflexes. The conducted work and contribution is: A) Improving Matsuoka's and Kimura's CPG model by building the output of an oscillator with two neural states in order to eliminate zero dead zone; and introducing the gait matrix to bridge the CPG net and robotic gait; and defining the reflex information matrix and reflex coefficient vector to provide a definite approach for the interaction between CPG and environment. B) Investigating parameter dynamics of CPG model by numerical simulation, getting the regularities between the parameters and robot motion states, and summarizing the applied method for CPG model parameters-setting in engineering. C) Advancing a new concept, rhythmic gait, to describe the motion pattern with self-oscillation and phase-lock. Building holosymmetry CPG net for quadruped robot, defining gait matrix and presenting its value-setting principles. Realizing typical quadruped gaits and completing the transition between any two gaits. D) Proposing the method of motion-mapping to simplify the control of redundant freedoms for multi-legged robot, and building the mapping function between knee and hip by imitating animal's motion. E) Advancing a framework for modeling biological reflexes, reflexing on three layers, specifying functions, features and application scope of lower reflex, medium reflex and higher reflex. Modeling flexor reflex by lower reflex to realize stepping obstacles, modeling vestibular reflex by medium reflex to realize motion on-slope. F) Proposing a motion-controlling framework inspired by biological neural system model. Designing motions by imitating quadruped mammal's motion. Building a quadruped robot under the framework.The quadruped robot can complete steady rhythmic motion with a velocity of 1/4-1/2 body length per second (0.13-0.24m/s), walk with several gaits such as trot and walk, transfer between any two gaits, have unstructured-environmental adaptability to some degree and can walk on rough terrain with 10° slope and 20mm-in-height obstacle (13.3% to leg length). The experiments proved the CPG motion-controlling method is better than the traditional method with simpler control, stronger environmental adaptability and higher velocity.