Collision avoidance is integrated into a Model Predictive Control (MPC) framework by incorporating track constraints into the MPC's objective function. These constraints are governed by COLREGs to make rule-compliant decisions and actions for resolving track conflicts. The predictive model employed in MPC also allows for more accurate collision risk assessments, with safety distances calculated at each prediction step to avoid ship domain conflicts. Collision scenarios are classified into 16 categories (4 by 4 matrix), considering both the relative bearings of the own ship and the target ship, to ensure consistency in the decisions made by the involved ships. A series of experimental model tests were conducted at the Towing Tank for Manoeuvres in Confined Water, Antwerp, Belgium to validate the proposed controller. Initially, scenarios involving collision avoidance with static obstacles in combination with various ship speeds were examined, with varying safety margin parameters to impose different safety requirements. The obstacles were set both physical and virtual to investigate the influence of ship-obstacle interactions. Further tests involved a dynamic obstacle, creating overtaking and meeting scenarios with different speed combinations for the own ship and the dynamic obstacle. Simulations were performed to evaluate scenarios that are impractical to replicate in the towing tank, thereby complementing experimental testing. The test results confirmed that collision avoidance can be successfully accomplished within a confined environment.
All data in the Integrated Marine Information System (IMIS) is subject to the VLIZ privacy policy
