The indoor radio source search using received signal strength (RSS) is difficult due to the multipath propagation effects. Robots driven by the ordinary gradient searching methods are always stuck in local maxima. Chemotaxis search and theseus gradient search can overcome local extreme traps, however they are inefficient in terms of travel distance. In this study, we propose a force field searching algorithm which is efficient in travel distance and invulnerable to RSS local maxima. A virtual attraction force weighted by RSS gradient from each possible location of the radio source is modeled, by gradually discard the impossible radio source locations eventually the robot can reach the true location of the radio source. Force field search is robust and fast even when the RSS readings are highly influenced by multipath effects. Simulation results driven by real data show the high efficiency of this method.