Hexapod moving in complex terrains via a new adaptive CPG gait design

Abstract

Purpose – This paper aims to use the Matsuoka’s neural oscillators as the basic units of central pattern generator (CPG), and to offer a new CPG architecture consisting of a dual neural CPG of circular three links responsible for oscillator phase adjustment, to which an external neural oscillator is added, which is responsible for oscillator amplitude adjustment, to control foot depth to balance itself when treading on an obstacle. Design/methodology/approach – It is equipped with a triaxial accelerometer and a triaxial gyroscope to obtain a real-time robot attitude, and to disintegrate the foot tilt in each direction as feedback signals to CPG to restore the robot’ horizontal attitude on an uneven terrain. The CPG controller is a distributed control method, with each foot controller consisting of a group of reciprocally coupling neural oscillators and sensors to generate different locomotion by different coupling patterns. Findings – The experiment results indicated that the gait design method succeeded in enabling a steady hexapod walking on a rugged terrain, the mode of response is such that adjustments can only be made when the tilt occurs. Practical implications – The overall control mechanism uses individual foot tilts as the feedback signal input to the neural oscillators to change the amplitude and compare against the reference oscillators of fixed amplitude to generate the foot height reference signals that can balance the body, and then convert the control signals, through a trajectory generator, to foot trajectories from which the actual rotation angle of servo motors can be obtained through inverse kinematics to achieve the effect of restoring the balance when traveling. Originality/value – The controller design based on the bionic CPG model has the ability to restore its balance when its body tilts. In addition to the model’s ability to control locomotion, from the response waveforms of this experiment, it can also be noticed that it can control the foot depth to balance itself when treading on an obstacle, and it can adapt to a changing environment. When the obstacle is removed, the robot can quickly regain its balance.