This paper presents a distributed Q-Learning based approach to search mine-like objects by multiple UUVs. We consider the complex undersea environments with regional variability in sensor characteristics. By applying a receiver operator characteristic curve analysis, the detection features of the sensor are exploited to achieve informative path planning. The study utilises a network aware communications model to construct the swarm formations in the presence of low bandwidth communications. A distributed Q-Learning based planner is employed to find an informative, communication-aware and safe path for each UUV. Simulation results show that simultaneous tasks can be cooperatively performed by a team of vehicles. Compared to the boustrophedon and greedy approaches, the Q-Learning based planner is shown to be more efficient in a time-constrained search mission under a nonhomogeneous environment.