For a group of underwater vehicles, the relative positioning problem is considered. That is, the mutual relative positions of the individual vehicles and hence the group formation is determined. The observations to be used are the relative ranges between the vehicles and the dead reckoning estimates of the individual navigation systems of the single vehicles. These data are to be distributed via an acoustic under-water communication system. The solution to the relative positioning problem is a prerequisite for implementing algorithms for coordinated behaviours of multiple underwater vehicles. A crucial condition is the limited communication capability in the underwater communication network. This is due to the comparably small bandwidth of the under-water acoustic channel and the lack of reliability of those connections, caused by multipath, diffraction and other acoustic wave propagation related effects. This paper introduces a simple concept for the solution to this problem, which avoids the usage of an acoustic positioning system such as LBL or USBL. This is achieved by implementing a variant of a trilateration technique in an extended Kalman filter.
The technology of Autonomous Underwater Vehicles has achieved an impressive degree of maturity during the last decade. However, due to the limitations of state-of-the-art embedded systems, underwater re-search vehicles are still limited in both their autonomy and capabilities. It would be a leap ahead, if researchers could use a multiple vehicle approach, whereby each vehicle plays the role of a sophisticated node (with sensor, processing, and communication capabilities) in a possibly large network - this means combining the properties of different systems in a team. Therefore, the request for application of multiple autonomous marine vehicles, operating in a coordinated way, has largely been growing. A multitude of tasks can be defined which are now becoming feasible by means of a coordinated network of agents which were out of reach for one single underwater vehicle.