A Leapfrog Strategy for Pursuit-Evasion in a Polygonal Environment

We study pursuit-evasion in a polygonal environment with polygonal obstacles. In this turn based game, an evader $e$ is chased by pursuers $p1, p2, ..., p{\ell}$. The players have full information about the environment and the location of the other players. The pursuers are allowed to coordinate their actions. On the pursuer turn, each $pi$ can move to any point at distance at most 1 from his current location. On the evader turn, he moves similarly. The pursuers win if some pursuer becomes co-located with the evader in finite time. The evader wins if he can evade capture forever. It is known that one pursuer can capture the evader in any simply-connected polygonal environment, and that three pursuers are always sufficient in any polygonal environment (possibly with polygonal obstacles). We contribute two new results to this field. First, we fully characterize when an environment with a single obstacles is one-pursuer-win or two-pursuer-win. Second, we give sufficient (but not necessary) conditions for an environment to have a winning strategy for two pursuers. Such environments can be swept by a \emph{leapfrog strategy} in which the two cops alternately guard/increase the currently controlled area. The running time of this algorithm is $O(n \cdot h \cdot {diam}(P))$ where $n$ is the number of vertices, $h$ is the number of obstacles and ${diam}(P)$ is the diameter of $P$. More concretely, for an environment with $n$ vertices, we describe an $O(n^2)$ algorithm that (1) determines whether the obstacles are well-separated, and if so, (2) constructs the required partition for a leapfrog strategy.