Spring Symposium on UR and Community Engagement has ended
Back To Schedule
Tuesday, April 24 • 2:40pm - 3:00pm
Artificial Intelligence in a Robot Playing Arm

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

While artificial intelligence is primarily a branch of computer science, its application in robotics remains heavily focused only in the engineering discipline. With intelligent systems being so prevalent in our current society and the future at large, we as computer scientists are potential robotics developers capable of contributing to this new paradigm. This project seeks to blend the two studies and explore some challenges faced in programming for robotics. We will do so by attempting to solve how a robotic arm would play an optimal game of Jenga. In this case, we define the optimal state to be a tower with only one block at each level, and if another piece is to be drawn, the tower would fall. The decision model for the game-play environment uses the basics of Markov Decision Process (MDP) with the addition of reinforcement learning to estimate the missing transition probability function. We implemented the temporal difference (TD) algorithm of reinforcement learning in order for the arm to teach itself the optimal path through a series of simulated test runs aiming for reward maximization. Analysis of results will demonstrate successful learning through the progression of total rewards achieved after each run. Once the learning is optimized in the simulation, integration with the physical robot arm would follow. The scripts are written in Python (3.0) and the simulation uses the open source Robot Operating System (ROS) framework and Gazebo simulator. This platform AI program will be used as a baseline for Dr. Kenneth Bogert’s undergraduate research and as part of a demo running behind glass for the UNC-Asheville Computer Science department. The presentation will discuss the MDP model and effects of TD-learning through a simulated demo.

Tuesday April 24, 2018 2:40pm - 3:00pm PDT
125 Rhoades Robinson Hall