Trajectory synthesis based on different fuzzy modeling controllers

Abstract

Trajectory synthesis based on different fuzzy modeling controllers is investigated. The learning scheme uses a fuzzy modeling neural network as the controller. Simulations focus on refinement and a thorough understanding of the fuzzy modeling structure. Given prespecified constraints such as the step length, crossing clearance, and walking speed, the learning scheme can generate a trajectory that satisfies these constraints. Different issues regarding the scheme are examined. They include the effects of using different membership functions and the effects of using different network structures. Simulation results are reported for a five-link biped robot.