Fine-Tuning Method of the GA-based Collision Avoidance System for AUVs

This article presents a fine-tuning methodology for an anti-collision system based on local path planning using Genetic Algorithm (GA). The anti-collision system has been implemented using the mathematical model of Autonomous Underwater Vehicle (AUV). Based on a map of the underwater environment with a high degree of complexity, depending on the input parameters of the simulation, the average number of collisions and the average deviation from the set trajectory were tested. The fine-tuning aimed to minimise the abovementioned parameters and determine the set of input parameters that would provide the best results. Fine-tuning was performed based on an exhaustive search method within specified limits. Compared to the average deviation value and the number of collisions calculated based on 45000 trials (all combinations of input parameters started 15 times) for each of the three selected sets, better results were obtained, which proves the purposefulness of this type of fine-tuning. Additionally, based on the determined impact of input parameter values change on simulations results, when fine-tuning the collision avoidance system, it is more effective to adjust the parameters related to the vehicle’s maneuvres first than the parameters related to the path planning method.