Desenvolvimento e avaliação de desempenho do software CAPLAN para previsão de colisões no espaço aéreo

Abstract

This paper presents an evaluation of the performance of collision detection software between aircraft that will be part of the air monitoring system of the Radar Livre project, a project of the Federal University of Ceará at its Quixadá campus. This project is based on ADS-B technology, a modern method of sharing information between aircraft, to feed their data to diverse systems. The proposed collision prediction software, CAPLAN, will work in a centralized database and requires a fast and accurate collision prediction algorithm. In addition, the implementation of such a system requires optimization techniques that allow it to process a large amount of aircraft data obeying memory consumption and processing time constraints that fit the needs of a real-time collision prediction system. Through research in existing works, a collision prediction algorithm was selected that analyzes the imminence of a collision between only two aircraft at a time. This algorithm was integrated to the CAPLAN module, leaving the task of applying it efficiently to the data on the thousands of aircraft information received daily by the system. The optimization techniques implemented in the software are: Symmetric Multiprocessing, which promotes a parallelization of the total processing of the software, dividing it among the processors available in the system; Division of Airspace in Areas, a technique that classifies aircraft into regions and verifies the collision imminence only in neighboring areas, preventing very distant aircraft from being analyzed; And Minimum Verification Radius, which, similar to the previous technique, proposes a minimum distance for two aircraft to be verified among themselves in the search for future collisions. Each of these techniques is detailed in section 4 of this paper. The performance of the developed software was evaluated by the Trace-based Simulation technique, which is widely used in Computer Science to make simulations and observations during the execution of a program and thus measure the resources spent, look for possible faults and evaluate their performance. The results of the experiment showed that the software is feasible to compose the Radar Livre system since it works with a maximum quantity of 33000 aircraft simultaneously, which does not represent an obstacle because the system currently operates covering the area of the state of Ceará and does not receive enough aircraft data to exceed that limit.