A game theory-based fusion algorithm for autonomous navigation of smart ships

Recent technological advancements facilitate the autonomous navigation of smart ships. Modern navigation systems and a range of various sensors provide real-time data on ship movements. Nevertheless, the disturbance of the marine environment and the occasional failure of the sensors affect the accuracy and reliability of such data. Hence, it is essential to develop methods that accurately and reliably estimate the ship's motions. This paper proposes a data fusion method based on game theory and investigates its accuracy and reliability. In this method, we consider a model-driven discrete-time Kalman fusion, and a data-driven adaptive weighted fusion as the game objects estimating the ship's position and its heading in real-time. We then design the corresponding game strategy based on local and historical estimations. Following the designed game strategy, the local estimations are then converted to the final fusion results. To verify the validity and effectiveness of the proposed method, we carry out extensive simulations and model experiments. The results confirm the accuracy of the estimations provided by this method and demonstrate its fault-tolerance performance. It is also shown that the proposed method meets the actual engineering requirements in real-time.