Sushil J. Louis

Home

I work in Evolutionary Computing and its applications to computer games and simulations for education and training (serious games), interactive evolutionary design, machine learning, engineering design, and non-linear optimization. My work combines genetic algorithms and case-based memory to learn to improve performance on similar problems. This is a broadly applicable technique and has been used in combinational logic design, combinatorial optimization (tsp, jssp), and in strike planning. The definitive work on this topic is my article on Learning with Case-Injected Genetic AlgoRithms (CIGAR). Currently, we are investigating how to affordably model human decision making by injecting cases derived from humans to bias genetic algorithms to produce solutions that are similar to human solutions. Our application areas map well to video games; specifically, 3D real-time strategy (RTS) games like Starcraft. If you are interested in Game AI for computer games, I organized the 2006 IEEE Symposium on Computational Intelligence in Games in Reno.

I direct the Evolutionary Computing Systems Lab (ECSL). Try the demos by following the "demos and movies" link on the left - you can supply your own data for most of them. The movies visualize some of our research in game AI and are part of our affordable human modeling project. In the movies, you can see boats navigating and the GA learns to avoid traps, and to generalize trap avoidance knowledge across missions.

I got my Ph.d. from the Department of Computer Science at Indiana University, Bloomington in 1993. My thesis advisor was Greg Rawlins . Others on my committee were Doug Hofstadter , Chris Haynes and Esther Thelene.