Optimizing the Movement Control System of a Hexapedal Robot using Modified Coordinate Descent
MetadataShow full item record
Publisher:The Ohio State University
Series/Report no.:The Ohio State University. Department of Mechanical and Aerospace Engineering Honors Theses; 2020
Bio-inspired legged robots may potentially have capabilities that traditional wheeled robots may not be able to provide. As these robots become practical for everyday life, their body shape, control system, and movement pattern need to be optimized to fit the expected functional capabilities. The objective of this proposed research project was to test strategies to simultaneously optimize both the body shape and movement strategies for a hexapedal (six-legged) robot to walk most effectively. Specifically, we hoped to use classical iterative optimization strategies to obtain optimal shapes for 3D printed legs with different properties such as length, shape, and center of mass, and simultaneously optimize the leg movement patterns to be appropriate for the chosen leg shape. Such simultaneous hardware and control system optimization have many open problems and may inspire the design and optimization of assistive devices and other robots. Due to time constraints, we primarily considered the optimization of the movement control system (without co-optimizing the body) using a modified version of a classical optimization technique called coordinate descent. We considered optimization using three variables: leg sweep, leg down, and duty factor. We found that the robot walking speed can reach an optimal value of 0.151 m/s with the converged parameter values set. In addition to executing these coordinate descents twice, we also performed three univariate parameter sweeps, one for each of these three variables, which fixing the other two at their default values. Overall, this thesis provides evidence for the efficacy of these univariate sweeps and sequential coordinate descent in obtaining the optimal value of the parameters, but more work is needed to automate the process and also make the process itself optimized for rapid and reliable convergence.
Academic Major: Mechanical Engineering