Research Robots

While in graduate school at Brown University I worked with a number of robots. I recently (circa 2014) ran into pictures of two of them in a CS Department newsletter, so thought I'd capture those images here.

The first robot pictured was a do-it-yourself job, where only the base of the robot was purchased, so gave some wheels, actuators and a platform to put your computer and other machinery. I began working on this robot with a larger team for the AAAI Conference Robotics competition that was held in Washington, D.C. that year. We had the second fastest time in one of the events, yet came in near the end of the pack due the the judging being subjective (not very scientific). Our algorithm was a big hacked up heuristic, and being an AI conference, this was much less sexy that the robots that accomlished nothing, but had fancy algorithms.

The second robot was much more polished, sophisticated and stable. I used this for my dissertation work of robots navigation using Partially Observable Markov Decision Models (POMDPs). With its feable sonar sensors and error-prone actuators, I was able to get this robot to wander around the department and keep a pretty good probability distribution about where it thought it was and it could make a decision about what the best action was to reach a specific destination.