We discuss search strategies for finding sources of particles transported in a random environment and detected by the searcher(s). The mixing of the particles in the environment is supposed to be strong, so that strategies based on concentration-gradient ascent are not viable. These dilute conditions are common in natural environments typical of searches performed by insects and birds. The sparseness of the detections constitutes the major stumbling block in developing efficient olfactory robots to detect mines, chemical leaks, etc. We first discuss a search strategy, ‘infotaxis’, recently introduced for the search of a single source by a single robot. Decisions are made by locally maximizing the rate of acquisition of information on the location of the source and they balance exploration and exploitation. We present numerical simulations demonstrating the efficiency of the method and, most importantly, its robustness to lack of detailed modeling of the transport of particles in the random environment. We then introduce a novel formulation of infotaxis for collective searches where a swarm of robots is available and must be coordinated. Gains in the search time are impressive and the method can be further generalized to deal with conflicts arising in the identification of multiple sources.