Dynamic Collision Avoidance for Maritime Navigation Using UKF Based Predictive Probability and Velocity Obstacle Framework

Abstract This study proposes a collision avoidance algorithm based on predictive probability using the Unscented Kalman Filter (UKF) to address the challenges of dynamic and uncertain maritime navigation. Previous UKF-based approaches relied on fixed time steps to extrapolate future positions from a single application, leading to progressively increasing uncertainties. In contrast, the proposed algorithm integrates ship dynamics into the UKF framework, allowing real-time recalibration and tailoring prediction horizon based on the ship’s physical characteristics and speed, ensuring a relevant and physically meaningful duration for enhanced accuracy and reliability regarding future positions. The algorithm integrates predictive probabilities into a Velocity Obstacle (VO) framework, ensuring adherence to maritime navigation rules (COLREGs) while enabling optimal path planning via a Nonlinear Model Predictive Controller (NMPC). By accurately calculating the Time to Closest Point of Approach (TCPA) and Distance to Closest Point of Approach (DCPA), it effectively prevents unexpected maneuvers, facilitating smoother, safer, and more reliable navigation in dynamic maritime environments. The effectiveness of the proposed method is demonstrated through comprehensive simulations, showcasing its ability to plan optimal paths and accurately track trajectories across various collision scenarios, such as head-on, overtaking, and crossing. The results emphasize the Unscented Kalman Filter’s robustness in estimating ship positions with precision, maintaining computational stability, and effectively mitigating collision risks ensuring enhanced safety and reliability in dynamic maritime environments.