Motion control of multiple autonomous vehicles is necessary to assure safe interaction of agents operating in the same environment. Under the influence of fluid flow the control strategy for the agents has to adapt to the forces acting upon the vehicles. A potential-field based algorithm for multi-agent control has been extended to account for correct guidance under the influence of flow. Path following and collision avoidance are the main objectives for the control algorithm which is developed for an underwater sensor network. An approximation of the flow is used for the path planning with model predictive control, whereas path following is assured through sliding mode control. The controller has yet been tested in simulations only, but experiments are planned with a swarm of five to ten micro submarines in a water tank with natural and forced convection. The fluid field in the simulation originates from a CFD analysis of the water tank under steady flow conditions where the impact of the vehicles has been neglected. (© 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)