Cooperative navigation of Underwater Autonomous Vehicles

By Luca Baglivo @ IST-Lisbon

Captain Nemo would be astonished by looking at UAVs which are currently employed in a number of applications. To mention just a few of them: pipeline inspection, harbour patrolling, mapping and sampling, bio-monitoring. The UAVs missions are more and more challenging and therefore full 3D sensing aperture, coordinated online planning and control, high robustness to failure and sensing noise and modularity are required for teams of UAVs to accomplish their tasks.

We know from complex bio-systems that “social is better”. That is the reason why cooperative is one of the most important keywords in the robotics research community since the last decade. The fundamental principle is that many cooperative robots can do better than what many grouped solo robots can do in terms, for example, of localization and mapping accuracy (e.g. here).

The DSOR team at the Instituto Superior Técnico of Lisbon has a strong expertise in the field of ocean robotics navigation and guidance. It is one of the partners involved in the just started FP7-ICT project MORPH. The goal of this project is to realize an “underwater robotic system composed of a number of spatially separated mobile robot-modules” not physically but virtually connected by means of the information flow. In this way it is possible, for example, to explore a large non-flat area while keeping a constant distance to the surface. As a visiting post-doc fellow I am working on the cooperative version of the localization problem addressed with a technique named Terrain Aided Navigation: the robot uses a prior known map to localize itself. A non-linear filter is built relying on dead-reckoning estimation and altitude measurements of the map provided by an echosounder.


The cooperative approach exploits the information coming from the measurements obtained by the other robots and from the estimated ranges between the robots. The cooperative localization problem is more general than SLAM or conventional distributed sensor networks as the entire system is dynamic; both the sensor platforms and the tracked targets are in motion.


Leave a Reply