We study here the problem of solving the traditional n-queens puzzle by a group of homogeneous reactive robots. We have devised two general and decentralized behaviour-based algorithms that solve the puzzle for N mobile robots. They all make a depth-first search with backtracking “in the wild” guaranteeing “in principle” a solution. In the first one, there is a predefined precedence order in the group; each robot has local sensing (sonar), a GPS, and is able to communicate with the previous and next group elements. In the other algorithm, there is only local sensing ability and a GPS. There is neither a predefined group order nor any peer-to-peer communication between the robots. We have validated our algorithms in a simulation context.