Joshua Bongard
Sunday May 20, 3pm ­ 4 pm
University of Zurich, Switzerland
www.ifi.unizh.ch/ailab/people/bongard

Biological Growth and the Automatic Design of Robots

My research is concerned with using ideas from developmental biology to build computer tools that allow engineers to automatically design, or "grow" simulated robots. Much of my work is based on a commercial software package, MathEngine, which allows the user to construct, or, in this case, grow three-dimensional objects, and then evaluate their behaviour in a physically-realistic, virtual environment. By incorporating an abstraction of natural selection into my software, I enable the computer to automatically grow virtual robots, or agents, that can accomplish various tasks. I will demonstrate some of the agents that the computer has "grown". I will also show some preliminary work on an educational software package which allows both the user and artificial evolution to work together to produce interesting agents.

Biography

A native of Toronto, Josh Bongard received an honours undergraduate degree in Computer Science from McMaster University, after which he spent a year as a software engineer for Computer Devices Canada in Calgary. After six months of travelling through India and Asia, he spent a year completing a Masters degree at the University of Sussex in the United Kindom, in Evolutionary and Adaptive Systems. He is currently working on a PhD in the Artificial Intelligence Lab at the University of Zurich, Switzerland. In addition to organizing and running a robotics summer school for high school students last year in Zurich, Josh has demonstrated his artificial ontogeny software in Europe, the United States and Japan.