Ship autonomous collision avoidance: Improving sparrow search algorithm for complex multi-ship encounters

Autonomous collision avoidance decision-making is a fundamental challenge in maritime transportation. To improve the efficiency and safety of such decisions, particularly in complex multi-ship encounter scenarios, this study proposes an enhanced decision-making model based on an improved sparrow search algorithm, named the elite adaptive interpolation sparrow optimisation algorithm (EAISOA). This algorithm improves the initial population distribution using a tent map and elite strategy, optimises a discoverer location update mechanism through a quadratic interpolation coverage search, and integrates a differential evolution strategy with horizontal crossover to dynamically maintain population diversity. The experimental results show that these enhancements significantly improve global search capability and convergence speed. In addition, a validation framework is developed using historical ship trajectory data, which applies spatio-temporal constraints to identify ship clusters at risk of collision. Fuzzy logic is used to quantify the collision risk index (CRI) between ships, and a database of realistic scenarios is established to support extensive algorithm validation. The results demonstrate that the proposed model can efficiently generate optimal collision avoidance manoeuvres, providing strong theoretical support for increasing navigational safety in complex maritime environments. • A novel improved sparrow search algorithm for ship collision avoidance is developed. • A verification framework for AIS data-driven ship collision avoidance algorithm is proposed. • A Collision Risk Index is designed based on fuzzy logic theory. • The improved collision avoidance algorithm has a faster convergence speed.