PUBLISHED PAPERS
2023 |
Comparative Performance Analysis of Two Multi-Aerial Threat Evaluation Algorithms Proceedings Article Humberto Baldessarini Pires; Lamartine Nogueira Frutuoso Guimarães Resumo | Links | BibTeX | Tags: C2 - Inteligência Artificial e Aprendizado de Máquina @inproceedings{234765,In modern aerial defense operation, the evaluation of potential threats is of paramount importance for effective response strategies, particularly when such assessment is performed in real-time. This study presents a comparative analysis of an algorithm developed by the authors, and referred to as DM, and a Markov chain-based approach (MC) in terms of prediction accuracy, execution time, and processing capacity. Notably, DM consistently achieved higher accuracy until simulation time 1350, despite both methods utilizing the same Artificial Neural Network architecture. Additionally, DM exhibited superior execution time and processing capacity, handling a maximum of 89 threats within a one-second timeframe, while MC processed 10 threats. Based on this, it can be asserted that DM meets the requirements for real-time threat evaluation. The results can be attributed to DM's simplified methodology, enabling more accurate and distinct predictions. |
A brief review of concepts and technological solutions to perfom DAA functions in the context of SIMUA Project Proceedings Article João Raphael Cioffi; Christopher Shneider Cerqueira; Jeanne Samara dos Santos Lima Resumo | Links | BibTeX | Tags: C2 - Inteligência Artificial e Aprendizado de Máquina @inproceedings{235508,The Unmanned Aircraft System Traffic Management (UTM) is a “traffic management ecosystem” for uncontrolled operations that is separate from, but complementary to, the Air Traffic Management (ATM). In this scope, the operations’ risk management is crucial to keep a safe integration of this manned and unmanned aircraft into a non-segregated airspace domain. Even more, “Detect and Avoid” functions and approaches must be established while dealing with multiple entities on this complex scenario. This paper provides a brief review of concepts and technological solutions to perfom DAA functions in the context of the project SIMUA - Safe Integration of Manned and Unmaned Aircrafts in non-Segregated Airspace, in regarding the main methods to be in compliance with the established research questions that scope project. |
Exploration and Rescue of Shipwreck Survivors using Reinforcement Learning-Empowered Drone Swarms Proceedings Article Leonardo D. M. de Abreu; Luis F. S. Carrete; Manuel Castanares; Enrico F. Damiani; José Fernando B Brancalion; Fabrício J. Barth Resumo | Links | BibTeX | Tags: C2 - Inteligência Artificial e Aprendizado de Máquina @inproceedings{235050,The goal of this project is to create a reinforcement learning algorithm that locates shipwrecked individuals using a swarm of drones. A simulated environment was developed to train and visualize the outcome of the trained algorithm considering the ocean’s dynamic circumstances. This project does not discuss image recognition of shipwrecked people, since the true focus of this project is to optimize the search routine of a drone to find the target in the most efficient way possible. The implemented Reinforce algorithm takes into account a dynamic map of probabilities, representing the chances of a person being found, as well as the position of other agents. Outcomes include an open-source Python package for the environment and the implementation of the reinforcement learning algorithm. The algorithm demonstrates superiority over the predefined approach, proving the advantages of reinforcement learning in efficiency and effectiveness. |
2022 |
Defense-related Object Detection in Aerial Images Proceedings Article Luciano Bittencourt; Paulo Andre Castro Resumo | Links | BibTeX | Tags: C2 - Inteligência Artificial e Aprendizado de Máquina @inproceedings{226575,Object detection in aerial images (ODAI) is an important task in computer vision and has applications in several areas. Recently, researchers directed their efforts to the ODAI, which requires detectors capable of dealing with arbitrary orientations, large variations in aspect ratios, densely clustered objects, multiple classes and instances per image. Our work is focused on objects that are relevant to defense systems. Such defense-related objects may present special challenges for detection and a reliable detector may be very useful as information source for defense systems. We have used publicly available aerial images and implemented some detectors based on Rotation-equivariant Detector - ReDet, which presented a very good performance for a broad class of objects. We tested such detectors using only defense-related objects. Our tests included dataset with and without data augmentation. The results achieved are consistent with the results published in some previous competitions. |