Enhancing Collision Avoidance at Sea: A Fuzzy Logic Decision Support System

By Samudrapom DamReviewed by Susha Cheriyedath, M.Sc.Sep 20 2023

In an article recently published in the Journal of Marine Science and Engineering, researchers investigated the feasibility of using a novel decision support system for collision avoidance in multi-vessel situations at sea.

Background

Currently, decision support systems harnessing advanced computer technology are used to ensure greater safety at sea by improving decision-making and situational awareness both ashore and onboard the ship. The key functions of these systems include ship course prediction, possible ship collision warnings, approach and groundings to a guard zone, and collision avoidance maneuver planning based on the Convention on the International Regulations for Preventing Collisions at Sea, 1972 (COLREG) rules.

However, avoiding collisions in multi-vessel situations is increasingly becoming difficult for navigators due to the growing traffic density at sea. Additionally, navigators have to make decisions in these situations as the existing COLREG rules that govern the right-of-way in close-quarter situations between vessels do not effectively address multi-vessel situations. However, decision-making can be challenging at sea, specifically in busy areas, where vessels lack adequate time to coordinate and communicate collision avoidance measures.

The COLREG rules regulating different types of encounters between two vessels can be potentially applied to multi-vessel situations if the navigational situation is effectively classified and determined. The classification process can be made efficient and simple using decision-making tools/integrating these tools into a routinely used navigation system. Identifying the most dangerous vessel is the first crucial step for navigators to avoid a collision in a multi-vessel situation, followed by a maneuver that meets COLREG regulation safety requirements.

The proposed model

In this paper, researchers proposed a novel decision support system that utilizes fuzzy logic to enhance situational awareness and assist navigators in avoiding collisions during multi-vessel encounters.

The decision model/system was based on the integration of artificial intelligence (AI) techniques and COLREG rules and consisted of two main modules, including MODULE 1 and MODULE 2, to determine the initial encounter conditions for target vessels, evaluate navigational situation and collision risk based on COLREG rules, sort target vessels, and identify the most dangerous vessel.

The authors developed algorithms to calculate collision avoidance maneuvers and collision risk in multi-ship encounters. Different parameters, including distance to the closest point of approach (DCPA), time to the closest point of approach (TCPA), navigation area, relative bearing (RB), ship types, and relative speed, were considered to determine the right-of-way between vessels and evaluate the collision risk.

In the collision avoidance system, fuzzy logic was employed as a decision-making tool. Researchers implemented fuzzy inference systems with trapezoidal or triangular membership functions to calculate the degree to which the inputs belonged to various fuzzy sets. The collision avoidance maneuver for the selected most dangerous ship was calculated based on the ship’s own speed, RB, and the closest point of approach (CPA) using fuzzy logic.

In this decision model, MODULE 1 contained initial parameters of the own ship and target ships and involved the calculation of initial conditions based on initial parameters. MODULE 2 consisted of two components, including Component 1 based on the sorting algorithm operation and use of COLREG rules, and Component 2, which included fuzzy reasoning and COLREG rules to simulate expert knowledge.

Model operation and simulation

In a multi-vessel collision avoidance situation, the relative course and speed, RB, CPA point position, TCPA, and DCPA were initially calculated by the proposed decision model using the initial data for every vessel.

The right-of-way was then determined according to the COLREG rules, and the most dangerous vessel was selected. Subsequently, the collision avoidance course was calculated based on the most dangerous vessel parameters for every two-minute time delay. Eventually, a time interval was calculated by the model in which collision avoidance was recommended considering every safety parameter and then a collision avoidance maneuver was recommended. 

A Monte-Carlo class of simulations involving several runs was performed to assess the proposed decision model performance for collision avoidance in multi-ship scenarios. A simplified ship dynamics simulator combined with the fuzzy collision avoidance system was used to perform the simulations. Several initial positions, orientations, and speeds were used for every ship to evaluate collision avoidance performance in different scenarios.

Significance of the study

The proposed model demonstrated excellent performance during reference simulation with only two vessels/two-ship simulations, as no cases of possible collisions were observed. However, the number of collisions increased significantly with the increasing number of ships in the simulation, as the avoidance of one target resulted in a critical condition with another target.

This finding indicated that the sole reliance on COLREGS and avoiding a target with the highest risk/lowest TCPA can lead to a crash with a third vessel that had not been identified as a dangerous vessel before performing the collision avoidance maneuver. The minimum initial distance between two vessels severely impacted the maneuver performance, with the collisions increased from one to nine when the minimum initial distance was lowered from one nautical mile (NM) to 0.5 NM in three-ship simulations.

Although the average minimum distance during simulation runs did not reduce significantly when the minimum initial distance was lowered to 0.8 NM, a major decrease in the average minimum distance was observed when the minimum initial distances were further reduced to 0.5 NM. Moreover, the maximum average risk obtained during simulation runs also did not increase significantly.

To summarize, the findings of this study demonstrated that the fuzzy algorithm can effectively prevent collisions in scenarios with only two ships. Thus, more research is required to develop and implement a cooperative collision avoidance algorithm that can recommend mandatory maneuvers for all vessels/ships in high-traffic areas.