An Adaptive Robotic Putter: Using Visual Input to Handle Putts Involving a Uniform Slope

摘要

Previous work in the area of machine learning has demonstrated the feasibility of using input of a visual sensory system for an adaptive learning algorithm for a robotic putter [1]. The previous environment used was a flat, almost frictionless four-foot by eight-foot surface. The sensory input was provided via an overhead vision system. That study focused on a learning algorithm that allowed the robotic putter to adapt, based on information recorded from previous putts at different areas of the putting surface in order to improve its accuracy. This research extended the previous work in the area of making the environment more complex and creating a more realistic putting situation. Overhead vision tracked the motion of single putts and this information was stored for putts of different angles. The main focus of this project was the development of an algorithm that interprets the sensory data and learns from the path of the ball the degree of slope present and how to adapt to it and complete successful putts.