Geographic routing scales well in sensor networks, mainly due to its stateless nature. Still, most of the algorithms are concerned with finding some path, while the optimality of the path is difficult to achieve. In this paper we are presenting a novel geographic routing algorithm with obstacle avoidance properties. It aims at finding the optimal path from a source to a destination when some areas of the network are unavailable for routing due to low local density or obstacle presence. It locally and gradually with time (but, as we show, quite fast) evaluates and updates the suitability of the previously used paths and ignores non optimal paths for further routing. By means of extensive simulations, we are comparing its performance to existing state of the art protocols, showing that it performs much better in terms of path length thus minimizing latency, space, overall traffic and energy consumption.