Planning, Decision Making, and Sensing for Autonomous AUV Navigation
For autonomous unmanned systems, the ability to sense and understand an environment, and then plan and execute the actions needed to complete a mission in that environment is crucial. In this pilot project we will explore both acting and sensing in an underwater environment. Navigation underwater requires special sensors. The approaches can range from using acoustic beacons giving absolute position to sensing the sea floor to estimate course and distance over ground. There are commercial solutions to these problems, but using natural underwater landmarks and terrain for absolute localization in a mapped environment is a flexible solution that is not yet 'off the shelf'. Making the maps while using them (know as Simultaneous Localization and Mapping, SLAM) is an even more flexible solution. It does not require any infrastructure, but is dependent on a good match of sensors to the environment. Each solution therefore tends to be a custom fit of SLAM approaches, sensors, AUV, and environment. There is a lack of any careful analysis of the literature on underwater SLAM that gives insight into the choices made and comparisons between them. We propose to do just such an analysis. As part of this we will also investigate the SOA with regard to relevant sensors currently available for underwater SLAM.
Regarding acting, we will study the problem of autonomously planning and executing so-called coverage missions. Such missions involve using one or several AUVs to cover a given maritime area with sensors, in order to e.g., search for mines or other objects, or map the depth. While carrying out the mission, the AUVs must take the presence of moving and static obstacles into account. In the Stockholm archipelago, static obstacles include the numerous small islands and shallow water areas, and moving obstacles include other maritime vessels, both large and small. Performing the mission in a reactive way, the AUVs should avoid other vessels, while making sure they come back later and cover the related area. They should also adapt to the depth being measured in real time. If the water becomes too shallow, the AUV must turn, and follow the contours of the shallow area, making sure it covers as much of the designated area as possible, without jeopardizing its integrity by getting grounded on the seabed.
Methods: The SOA on underwater SLAM for navigation, and control will be reviewed by an extensive literature study. Sensor technologies will be explored by contacting the companies and sales representatives complemented by the experience of ourselves and colleagues working with AUV's. We will then carry out simulation studies to illustrate and test the developed autonomous planning and execution functionality.
Conducted by: John Folkesson (KTH), Petter Ögren (KTH)