I''ve been scouring the net for resources concerning dynamic tactical path planning, and I figured I''d tap into the wealth of knowledge on gamedev while i was here :-D. Some ideas I''ve explored: -Modified A* as presented in Game Programming Gems 3 -Spiking NN Radar approach as presented in a paper by some German computer scientists (search for Dynamic Path Planning with Spiking Neural Networks on Citeseer) -Approach used in the Bot Navigation articles on AI-Depot, in essence choosing the path on which obstacles are farthest away. I''m looking for something that can operate in real-time for use in a simulation involving multiple bots who operate in teams to accomplish specific goals (destroy enemies, rescue hostages etc..) as part of a research project. Any other insights or tips on constructing this project would also be appreciated. Thanks guys.

Check out the D* algorithm (Dynamic A*). You should be able to achieve your stated goal using this algorithm.

Cheers,

Timkin

Spiking radar neural-net pathfinding! Marketing would love that phrase. How cool would that sound to your average game player?



My Website: ai-junkie.com | My Book: AI Techniques for Game Programming

Options two and three are reactive approaches, and I''m pretty sure they won''t provide any "tactical" capabilities.

What exactly did you have in mind?


For a path-planning solution combined with terrain decision making, see my Pathematics approach. I''ve come to understand it in terms of reinforcement learning theory; you specify rewards (points or objects in space) and costs (danger for moving, travel times), and the algorithm combines the two dynamically to suggest a route.

Alex


Artificial Intelligence Depot - Maybe it''s not all about graphics...

Well I''d like to make AI bots that work together to accomplish goals in fighting environment like UT or CS. However I''m going to graphically represent them with software I''ll write myself, so it will actually be a little less complex because the represenation will be 2D. I''ll have to check out that pathematics thing.

My favorite concept is to use many influence maps to determine the tactical importance of different areas, and to then pathfind over them. Some influence maps could reflect "learning;" say add high pathing costs to cells near locations at which you have died frequently. Many similar ideas are discussed in a paper by William van der Sterren titled "Terrain Reasoning for 3D Action Games;" google for it. I''ve also put forward a whole lot of speculative ideas in this forum; search for "influence maps" and I think you''ll find a whole lot of ideas, some of which I hope you''ll find useful.

I have the suspicion that this approach is a lot like your "pathematics," alexjc. I''m going to your site right now to find out what it''s all about.