Model checking knowledge in pursuit evasion games
IJCAI'11 Proceedings of the Twenty-Second international joint conference on Artificial Intelligence - Volume Volume One
| Hi-index | 0.00 |
We define a game of pursuers seeking evaders on a grid, originally motivated by, but not limited to, the problem of unmanned aerial vehicles searching for moving targets. The pursuers have a speed advantage over the evaders but are incapable of determining an evader's location unless a pursuer occupies the same location as that evader. The evaders, on the other hand, operate with unlimited luck and cunning: if it is possible for them to avoid the pursuers, they will. By treating the players as nondeterministic finite automata, we can model the game and use it as the input for a model checker. While model checkers normally are used to verify program correctness, we use the model checker to generate a pursuer-winning search strategy. We demonstrate this technique for the full model and then develop heuristics to reduce the model so we can address larger problem sizes. We further prove an upper bound on the minimum pursuer-winning speed; moreover, we show the bound is tight for a particular class of search strategies.