Alex Snow
Mentor: Douglas Blank
Computer Science

Explorations of Deep Learning in Robotics

A method of enabling machines to problem solve by exposing them to a collection of data and having them solve based on the data they learned, called deep learning, has become a popular research topic in recent years. For example, through deep learning, a computer can locate a certain object in a picture when commanded to if that computer was shown or trained with various images of the specific object. We will explore this by teaching an autonomous humanoid robot to reach out and look towards a fluorescent colored ball wherever the ball is within sight by training the robot with different images of the ball. To accomplish this task, we will write a series of code connecting the robot to an online deep learning system through a computer. From there, we will take images of the ball with varying surroundings, input them into the system, and hold the same ball in front of the robot. This will be a demonstration of deep learning in the field of robotics.