William Kim

Date of Award

2024

First Advisor

Michael Bergman

Second Advisor

Kenneth Knox

Abstract

In this project, a bipedal robot is designed, built, and trained to walk. The core components of the robot were eight serial bus servomotors with a 240-degree range, a rechargeable battery, and a single-board computer (Raspberry Pi) as a controller. CAD software (SolidWorks) was used to construct and 3D-print the limbs and hull of the robot. The design process was iterative, with multiple revisions made to fix issues that only became apparent when operating the robot. Components that flexed were remade thicker, and parts that needed more clearance were adjusted accordingly. Once the robot was assembled and its components were confirmed to work, its gait was optimized for speed and stability first manually and then by using a hill climber optimization algorithm, which iterated improvements by making small adjustments to parameters and testing them in a simulation. Manual adjustment resulted in a speed of 0.35 cm/s. Computer optimization yielded an improved 0.91 cm/s, a 160% increase in speed. The hill climber algorithm worked well in smoothing out motion for stable movement, but fell prey to the local minima issue. Its performance depended heavily on the quality of the initial parameters it received. If a large adjustment in walk style was desired (e.g. longer steps, less lateral motion), the algorithm had to be encouraged to change by being given parameters that exaggerated the target movement. Despite these factors, the hill climber algorithm greatly improved walking performance.

Download
Simon's Rock Off-campus Download

Simon's Rock students and employees can log in from off-campus by clicking on the Off-campus Download button and entering their Simon's Rock username and password.

Share

COinS