OPTIMIZING A QUADRUPED ROBOT A COMPARISON OF TWO METHODS.pdf (1.6 MB)
Abstract-
Robots that have been optimized in simulation often underperform in the real world in comparison to their simulated counterparts. This difference in performance is often called a reality-gap. In this paper, we use two methods, genetic algorithm and topology optimization, to optimize a quadruped robot. We look at the original and optimized robots’ performance in simulation and reality and compare the results. Both methods show improvement in the robot’s efficiency, however the topology optimization behaves in a more predictable manner and shows similar results in simulation and in real laboratory testing. Modifying robot morphology with a genetic algorithm, although less predictable, has a potential for more improvement in efficiency.
KEYWORDS: reality-gap, genetic, topology, optimization, quadruped
This academic paper features our DYNAMIXEL AX-12 all-in-one smart actuators
All credit goes to: ROBERT PASTOR, ZDENKO BOBOVSKY, PETR OSCADAL, JAKUB MESICEK, MAREK PAGAC, ERIK PRADA, LUBICA MIKOVA, JAN BABJAK from the Department of Robotics, Department of Machining, Assembly and Engineering Metrology, Faculty of Mechanical Engineering, VSB-TU Ostrava, Czech Republic, Department of Mechatronics, Faculty of Mechanical Engineering, Technical University of Kosice, Slovak Republic