Modeling Gait Transitions of Quadruped and Their Generalization With CMAC Neural Networks

The gait transition models of a quadruped are studied based on gait kinematics and CMAC neural networks are applied to learn and generalize these gait transition models. Three gait transition cases are studied: from wave gait to continuous follow-the-leader gait, from walk to trot, and from trot to gallop. Four solution methods are proposed for solving the gait transition models. Computer simulations are conducted to evaluate and display the gait transition models. The good transition gaits are then selected to train CMAC neural network gait transition models. The performance of the CMAC gait transition models are evaluated and found to be satisfactory.