Artificial Intelligence for Robotics

Summary of robot arm learning process

The result for our arm training is a combination of several AI techniques in one mashup process. We have reinforcement learning in the process of maximizing rewards. We are using Q-learning to compute rewards for successful completion or partial completion of a goal. We use a straight artificial neural network (ANN) to predict the outcomes of actions given a beginning state. And we introduced genetic algorithms to innovate and make movement patterns that we can teach to the robot arm:

Our general process for teaching the arm how to move is based on a type of reinforcement learning called Q-Learning. Our ...