A three-layered architecture for real time path planning and obstacle avoidance for surveillance USVs operating in harbour fields

The use of unmanned vehicles in the field of underwater and marine applications is increasing significantly in recent years. Autonomous vehicles (like AUVs and gliders) or teleoperated ones (like ROVs) are currently employed for executing a number of different underwater tasks, like inspecting submerged pipes, executing maintenance interventions on underwater gas- or oil-platforms, collecting environmental or oceanographic data, performing surveys on sites of archeological interest. In parallel with the development of underwater vehicles, unmanned surface vehicles (USVs), are they also witnessing an increasing interest from the robotic community, especially with the goal of performing surveillance applications, like patrolling and maintaining safeguarded against intruders harbours or other “crucial” sites. The potential benefits offered by automated vessels equipped with sensors such as cameras or sonars are quite evident, since they could be used to quickly identify the level of menace of unknown radar track without exposing any human operators to possible threats. However USVs, unlike in the underwater case, have to face the problem of avoiding other vessels which in most cases are manned ones. This is a crucial point especially in that kind of application, where the automated vessel has to move quickly towards a possible menace while at the same time avoiding all the other boats normally operating in the harbour area. Unfortunately, at the current state of art, a reliable methodology to avoid the other vessels and the availability of effective and accurate obstacle detection sensors is still missing. This paper focus its attention on the case of USV used for security applications within a harbour, devising a solution that can be real-time implemented for the obstacle avoidance problem under critical situations where the vehicle as to reach its target as fast as possible while guaranteeing the safety of the other vessels. The presented solution is based on a three layered hierarchical architecture: the first layer computes a global path taking into account static obstacles known a priori, the second layer modifies this path in a locally optimal way (under certain assumptions) exploiting kinematic data of the moving obstacles, while the last layer reactively avoids obstacles for which such data is not available. The paper will be therefore organized as follows: in the first section an introduction and state of art are presented, in the successive section the work will discuss the first layer and the methods for the static obstacles avoidance, while in the third the paper will focus on the moving obstacles and the proposed avoidance algorithm, while also presenting many different detailed simulation results regarding the performances achievable by the overall architecture. Finally a concluding section will also indicate some still open problems and future work directions to be developed.