Distributed Coordination and Data Fusion for Communication-limited Underwater Search Geoffrey A. Hollinger, Srinivas Yerramalli, Sanjiv Singh, Urbashi Mitra and Gaurav S. Sukhatme Abstract— This paper presents coordination and data fusion methods for teams of vehicles performing target search tasks without guaranteed communication. A fully distributed team planning algorithm is proposed that utilizes limited shared information as it becomes available, and data fusion techniques are introduced for merging estimates of the target’s position from vehicles that regain contact after long periods of time. The proposed data fusion techniques are shown to avoid overcounting information, which ensures that combining data from different vehicles will not decrease the performance of the search. Motivated by the underwater search domain, a realistic underwater acoustic communication channel is used to determine the probability of successful data transfer between two locations. The channel model is integrated into a simulation of multiple autonomous vehicles in both open ocean and harbor search scenarios. The simulated experiments demonstrate that distributed coordination with limited communication signifi- cantly improves team performance versus prior techniques that continually maintain connectivity. I. INTRODUCTION Communication between networked robotic vehicles is rarely (if ever) perfect. One method for dealing with im- perfect communication is to constrain the movements of the vehicles so that they remain within range, line-of-sight or both. These constraints can be maintained continaully [1], [2] or periodically [3], [4]. However, any method that depends on connectivity occurring at a fixed time will be brittle if the model of the communication system is incorrect. For example, if a planning algorithm requires two vehicles to be connected at given positions, a failure would occur if com- munication is worse than planned. In reality, communication between robots cannot be predicted by simple heuristics like distance and line-of-sight [5]. The inherent unpredictability of communication in many real-world applications motivates the development of algorithms capable of operating with any level of shared information. We explore the problem of multi-vehicle coordination with limited shared information through analysis of the underwater moving target search domain. In this scenario, This research has been funded in part by the following grant, ONR N00014-09-1-0700. G. Hollinger and G. Sukhatme are with the Dept. of Computer Science, Viterbi School of Engineering, Univer- sity of Southern California, Los Angeles, CA 90089 USA {geoffrey.hollinger,gaurav}@usc.edu S. Yerramalli and U. Mitra are with the Dept. of Elec- trical Engineering, Viterbi School of Engineering, University of Southern California, Los Angeles, CA 90089 USA {srinivas.yerramalli,ubli}@usc.edu S. Singh is with the Robotics Institute, School of Computer Science, Carnegie Mellon University, Pittsburgh, PA 15213 USA ssingh@ri.cm.edu (a) Calm wind and low shipping activity results in a connected network of underwater vehicles. (b) High wind and high shipping activity disconnects the same network of vehicles. Fig. 1. Communication with underwater acoustic signals is highly imperfect, and it is often difficult to predict whether two vehicles can share information at given locations. Utilizing such communication systems during multi-robot planning requires fully distributed algorithms capable of operating at any level of communication. autonomous underwater vehicles (AUVs) need to locate a lost target (e.g., a submarine or enemy underwater vehicle). Moving target search with communication limitations re- quires both distributed path planning and methods for fusing information gathered when the vehicles were not connected. For instance, if a vehicle makes a number of observations and then comes into contact with a vehicle that was out of contact for a long time, sharing the entire history would be costly both in terms of communication and in terms of computation required to fold the observations into each vehicles’s current information map. Thus, solving this problem requires the development of both coordination and data fusion techniques. In this paper, we examine the problem of moving target search with a team of autonomous underwater vehicles (AUVs) utilizing acoustic communication. We first discuss related work in multi-robot coordination, communication- constrained planning, and search with mobile robots (Sec- tion II). We then propose a fully distributed coordination algorithm in which each AUV optimizes its path given the information it currently has available (Section III). Our approach is passive, in that it does not incorporate the communication model into planning, but rather utilizes any shared information to improve the plan. In addition, we