Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
OPTIMIZING SWING FOOT TRAJECTORY OF HUMANOID ROBOT WALKING FOR ENERGY EFFICIENCY
Submission Author:
Caroline Silva , RJ
Co-Authors:
Caroline Silva, Marcos Maximo, Luiz Carlos Góes
Presenter: Caroline Silva
doi://10.26678/ABCM.COBEM2021.COB2021-1096
Abstract
In this article, we developed a method using an evolutionary strategy (ES) algorithm. The goal is to find the parameters that make humanoid robot walking energetically more efficient. The robotic bipedal walk is limited for several reasons, one of which is the design. The kinematic chain of a humanoid is restricted by geometry or components that have operating limits, such as servos. We intend to find parameters that help the walk, without changing the design of the robot. Our contribution is to implement the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) to find parameters that define the oscillating foot trajectory. We use energy as metrics to find three parameters, which correspond to the frequency of the steps, the toe support angle (at the beginning of the swing phase), and the heel support angle (at the end of the swing phase). We chose CMA-ES because it is considered one of the best choices against ill-conditioned, non-convex black-box optimization problems in the continuous domain. Being in the state-of-the-art of ES algorithms. As most of the robotic gait corresponds to the single support phase (or swing phase). For this reason, in this work, we consider the swing phase in the Sagittal plane, with three joints: hip, knee, and ankle. Therefore, the measured energy corresponds to the efforts to obtain the pitch angles in the swing leg joints. From the trajectory defined for the swing foot with the optimal parameters obtained by the CMA-ES, the inverse kinematics (IK) finds the swing leg joints trajectory. After that, it uses them as a reference for the position control of each servo. For the dynamics of the swing leg, the kinematic chain considered has three links. The equation of the dynamics of the movement can be derived using Lagrangian Mechanics. We validate our approach by implementing the proposed above with the parameters corresponding to our custom-made robot Chape. In the results obtained, we can change the speed in different situations, knowing which is the best step size to be defined, to the walk has the lowest energy cost.
Keywords
Humanoid Robot, metaheuristic optimization, Energetic Efficient Walking, Swing Leg Movement
