Iterative Genetic Fuzzy Logic System for Solving the Aircraft Conflict Resolution Problem
Résumé
It is extremely important to have an efficient mechanism to determine optimum conflict-free paths for aircraft in order to improve airspace safety. This research focuses on developing a genetic fuzzy logic based approach for solving the aircraft conflict resolution problem where the objective is to obtain conflict-free trajectories for aircraft in a circular airspace while minimizing the cost of maneuver. Uncertainties in the velocity and the maneuver parameters are also considered which causes each aircraft's position at any instant to be within a region of uncertainty represented by a convex hull. A new and unique architecture for Fuzzy Logic System (FLS) is used that consists of a hidden layer of neurons and a layer of decoupled Fuzzy Inference Systems (FISs) which is capable of iteratively traversing the search space to find a near optimal solution. For this purpose, an artificial intelligence (AI) called EVE is used to tune the parameters of the system and once it is trained, its capability is evaluated on a set of test scenarios. The results obtained for the five and ten aircraft problem for different levels of uncertainty are compared to those obtained by directly applying Genetic Algorithm (GA). The FLS is able to obtain near-optimal solutions comparable to those of GA at a fraction of the computational cost.