Rendezvous of mobile robots in unknown environment via graph optimization approach

In this paper is considered the problem of finding paths for the motion of two autonomous mobile robots starting from given points of an arbitrary shaped but bounded two dimensional domain in presence of unknown obstacles. The objective of any of the two cooperating robots is that of follow a path that leads it to meet the other as soon as possible. No a priori information is known in advance about the geometry of the workspace nor about the number, extension and location of obstacles. Robots have sensing devices that detect obstacles or pieces of walls lying beyond a fixed view range. The problem is embedded in a graph optimization framework and a new algorithm is proposed, characterized by little computer power requirements. The optimization approach used is highly decentralized; the two robots can perform computation and navigation in parallel fashion and communicate only by partial map updating. Theoretical and practical aspects of the algorithm are investigated.