Analytical Energetic Approach for Predictive Generation of Dynamic Biped Walking - Use of Average Energies

This paper presents a control strategy for biped robot walking based on energetic approach. Our aim is to be able, in real time, to predict the length of the next step that the biped should execute to ensure its desired speed. According to known (measured) state in the previous step the proposed algorithm will establish the desired global parameters (i.e. the biped average velocity). The new approach presented in this paper benefits from a formulation of energy exchanges using the average values of all energies contributing during dynamic bipedal walking. We develop a model that takes into account kinetic and propulsion (provided by the rear leg) energies as well as the energy lost during the impact of the front leg. We introduce a recursive formulation in order to predict the energy behavior of the biped robot during its walk. Analytical and concise expression suitable for real-time application was developed. We show that the same formulation is convenient for both flat terrain and crossing over obstacles in the sagittal plane. Hence, we are able, in both cases, to determine the next step length required to ensure the desired average velocity. We validate the proposed approach through an ADAMS simulation of a planar robot in the sagittal plane. The obtained results show the efficiency of this promising approach.