Multiagent systems controlling robots can have different structures, depending on a way of generating decision in these systems. Decisions can be work out in centralized, decentralized or even hybrid way (hybrid system is a connection of both centralized and decentralized systems). In the case of controlling a team of underwater vehicles, it is significant to examine different structures of multiagent systems for choosing the best one for defined underwater task. In the paper, results of operation of three different structures (centralized, dispersed - decentralized and hybrid) of multiagent control systems of underwater vehicles team were presented. The systems were tested in predator-prey problem. In this problem, a team of three underwater vehicles had to catch another underwater robot escaping with larger velocity.