Movies and Demos
Demos: the genetic algorithm runs on our server but
you can track progress on this web page (requires java applet
support).
-
Traveling Salesperson: A genetic algorithm solves a TSP. We can
use your data in the
TSPLIB format for symemetric TSPs. The evolving tours are
nicely visualized on your web page using java.
- Job
Shop Scheduling: A genetic algorithm for solving JSSPs. We can
use your data if it is in the OR-Library
format for JSSPs. The evolving schedules can be visualized graphically.
-
Logic Design: A GA to design combinational logic circuits. You
have to provide a truth table.
- GA
Calibrates a CA: A GA evolves the rules of a cellullar automata
for predicting mining activity.
-
Cellular Automaton: You can see a visualization of the mining
activity with this applet.
-
A* router:
Point and click on points in the map to find least cost routes.
- Strike Defense
Movies: Visualize our current and past research.
- San Diego Transit: Watch
boats navigate in San Diego. Uses Lagoon, our open source 3D-RTS
engine.
- Game World Navigation: Cigarette
boats try to maintain a formation. Uses Lagoon, our open source 3D-RTS
engine.
- Another San Diego
Transit: You can see the Downtown San Diego Skyline get nearer
as the boats approach in formation. Uses Lagoon, our open source 3D-RTS
engine.
- Mission: This series of movies
shows the mission being flown by our platforms. Note that the
horizontal bars on the top left and right of the window show the
value and health of the platforms and targets. Blue hemispheres
denote the volumes covered by threats - any platforms that fly into
these will probably be shot down.
- Popup The second movie shows the
dynamic nature of planning. A popup threat appears (turns on its
radar) at just the right time and at just the right place to hurt
the platforms flying by on their strike mission.
- Replan: The genetic algorithm
re-plans and re-routes to go around the popup threat. Whenever the
environment changes, the game AI invokes a genetic algorithm to
replan - you can see the results.
- Trap: In this very simple research
scenario, red force has layed a trap for the lone platform that is
going to strike targets. We would like the genetic algorithm to
learn to avoid traps; in general, we would like the genetic
algorithm player to learn to avoid confined areas as they are good
places for red traps.
- Avoidance: A human player with any
experience playing the game would soon realize that good routes
avoid confined areas. By storing and learning from human game play,
the genetic algorithm also learns to avoid confined areas and
chooses a longer but safer route to the targets.
- Generalization: Not only does the
genetic algorithm player learn to avoid confined areas, our
representation generalizes this knowledge to other scenarios and
other locations.