Autonomous ships, intentions and situational awareness

Building on world-leading competence in sensor fusion and maritime collision avoidance from the Autosea project (2015-2019), the Autosit project is conducting research on situational awareness for autonomous surface vehicles. In most of the research literature, situational awareness refers to the awareness that a human operator has of an operation that he or she is in charge of. However, autonomous ships must themselves be in possession of situational awareness to interact smoothly with manned and unmanned vessels without accidents. In particular, autonomous ships should be aware of the intentions of other vessels (manned or unmanned), at the very minimum to the same extent as a human seafarer would be. This awareness can be generated from a variety of data sources. First, historical patterns of how ships behave can be used to predict how a ship is likely to behave at future times. Second, systems for information exchange such as the automatic identification system (AIS) can be used. Third, data from high-resolution sensors such as lidars and cameras carry a lot of information about a ship's behavior and capabilities. The Autosit project will develop reliable and efficient algorithms for fusing such diverse information types together into a consistent world image for the autonomous ship. During 2021 the work in the Autosit project has focuseed on three topics: Long term prediction of ship motion from AIS data, fusion of radar and AIS data, and detection of ship parts in camera images. Within long term prediction we have especially studied systematic techniques such as particle filters which can be provide a meaningful representation of anticipated future motion by means of a probability distribution. Within radar-AIS fusion we have developed algorithms which combine the strengths of traditional target tracking algorithms with insights regarding the fundamental differences between AIS and radar.

Autonomous surface vehicles (ASV’s) need to be in possession of situational awareness in order to interact safely with other vessels, whether manned or unmanned. Building on cutting edge research in sensor fusion and collision avoidance, the Autosit project will deliver algorithms for situational awareness that enable ASV’s to guess and predict the intentions of other vehicles. These algorithms include long-term vessel prediction based on machine learning, fusion of radar and data from the automatic identification system (AIS), and pose estimation using cameras.