Gait Optimization Method for Quadruped Locomotion

Abstract

The scope of the paper is to develop a methodology for finding optimal gaits of a quadruped robot using genetic algorithm, comparing the results to the ones resulting from natural evolution. The optimization is performed over preimposed contact forces to find the best shapes that guarantees the minimum energy consumption during a single stride cycle. The dynamic formulation of the four-dimensional model is developed without involving any specific kinematic mechanism for the legs, considering the entire gait spectrum a quadruped can exhibit. The optimization model consists of a set of constraints that ensure the feasibility and stability of the gaits. Results are presented for an optimization requiring a constant speed of 1.35 𝑚/𝑠. The optimal gait was found to be consistent to nature, suggesting that energy consumption is one of the key factors contributing to the evolution of gaiting patterns in quadrupeds. Eventually, a comparison between different existing gait patterns is carried out in terms of foot contact time and energy consumption.