This work addresses the problem of deploying and controlling mobile relays in robotic networks. We consider a general data collection scenario in which a group of robots stream data towards one or more base stations. The robots form a wireless mobile ad hoc network relaying data towards the base stations in a multi-hop fashion. Additional robots are used as mobile relays with the aim of optimizing overall network performance in terms of data throughput. To dynamically control the positions of these relays, we use the combination of a mathematical programming model that includes a number of constraints and penalties to closely model the wireless environment, and heuristics to account for variations in wireless links stability. We demonstrate that the approach is computationally affordable for reasonably sized networks. The approach is evaluated in realistic simulation scenarios, studying the impact of the model parameters on the quality of the provided solutions. An experimental validation is also carried out in a real testbed, confirming the efficacy of the proposed approach for adapting relays' positions to the changes in the robot network.