In this paper we compare three approaches employed in multirobot control: leader follower scheme, virtual structure and behavioral approach. We describe these approaches, show their advantages and weaknesses and point out suitable applications for each of them. We also present the architectures used in multirobot systems, their variations, and we discuss necessity of communication between subsystems and layers of the system. The problem of robustness and fault-tolerance in distributed systems is also mentioned. One selected variation of behavioral approach employed in security against invaders in surveillance area is presented.