Building Better Bots

Dr. Mark Goadrich, Kathryn Hardey and Dr. Matthew Jadud

Kathryn Hardey, a brand new member of the Moonbot Interactive team, recently attended the Genetic and Evolutionary Computation Conference (GECCO) in Philadelphia. While there, Kathryn and her co-author Eren Corapcioglu, presented their research on exploring and evolving process-oriented control for real and virtual fire fighting robots. Their paper, Exploring and Evolving Process-oriented Control for Real and Virtual Fire Fighting Robots was published in GECCO's fourteenth annual conference proceedings.

Mentored by Dr. Mark Goadrich of Centenary College and Dr. Matthew Jadud of Allegheny College, their research began by building a robot for the Trinity College Fire Fighting Home Robot Contest. The goal of the contest was to create a robot that could put out a candle located within a maze similar to a house.
From that experience, Kathryn and her colleagues realized that they could build a better robot using artificial intelligence to evolve the robot brain controller  — in the same way that "survival of the fittest" works in the real world. By creating a simulation of their robot, they could test the effectiveness of using process networks as the basis for evolutionary robotics.

Kathryn says her research helps with her work here at the moonbase: "It helped me to think about problems from several different perspectives. When working on my tasks at Moonbot, I can think of many different ways to solve the problems I come across, just as my AI research was about discovering different, more effective controllers."

Learn more about their research on the Rescue Robotics blog, view a video of their first robot or read the paper detailing their research.