Captain, Brazilian Air Force (FAB) · Civil-Aeronautical Engineer · PhD Candidate and Researcher in Air Transport at the Aeronautics Institute of Technology (ITA)
Research on machine learning, reinforcement learning, and data-driven decision support for air traffic flow and capacity management under uncertainty, with applications to runway operations, demand-capacity balancing, Urban Air Mobility, aviation safety, and probabilistic prediction.
My current research develops operationally grounded machine-learning and reinforcement-learning methods for aviation systems in which demand, capacity, weather, and operational conditions evolve under uncertainty. A recurring objective is to connect data-driven methods with decisions that air traffic managers, airport operators, and future UAM service providers can interpret and evaluate.
SIGMA-Sky investigates new operational concepts, rules, and air traffic management services for the safe and efficient integration of Advanced Air Mobility (AAM) operations into the airspace. The project addresses airspace design, separation and conflict-management approaches, performance indicators, and computational simulation for quantitative evaluation of operational scenarios.
An interactive 3D airspace-corridor designer for Urban Air Mobility. The dashboard supports geographic and three-dimensional visualization of corridor networks, route and node exploration, CSV data import, and interactive route management for UAM airspace-design studies.
Authors: Lucas Orbolato Carvalho; Mayara Condé Rocha Murça.
An interactive performance-analysis environment for simulated UAM corridor scenarios in the São Paulo Metropolitan Region. Built from BlueSky simulation outputs, it organizes performance assessment across safety, efficiency, capacity, predictability, and equity, with comparative metrics, maps, conflict information, critical-waypoint analysis, and three-dimensional trajectory visualization.
Authors: Danilo de Farias Matos; Lucas Orbolato Carvalho; Mayara Condé Rocha Murça.
Model-free reinforcement learning for tactical demand-capacity balancing at São Paulo/Guarulhos, including miles-in-trail control and runway-capacity decisions under uncertainty.
A stochastic model-free reinforcement-learning framework for runway capacity management under operational and weather uncertainty.
Strategic ground-delay allocation under uncertainty, linking per-flight delay decisions to downstream tactical congestion in conventional ATM and UAM.
ResearchGate | Project page | Paper | Poster
Recognition: Selected as a Student Best Paper Award finalist at DASC 2026.
A minimax-regret runway-capacity allocation formulation for managing arrival and departure capacity under uncertain capacity constraints.
An earlier runway-capacity study using tabular reinforcement learning with eligibility traces to optimize capacity utilization under uncertain weather conditions for a fixed runway configuration.
UAM planning framework integrating demand estimation, socioeconomic information, vertiport siting, route design, existing aeronautical infrastructure, and operational constraints.
Simulation-based assessment of helicopter-to-eVTOL substitution, avoided emissions, and potential carbon-credit implications for UAM operations.
Reinforcement-learning approach to departure-delay repair of on-demand four-dimensional UAM trajectories under separation constraints.
DOI | ResearchGate | SSRN
Multiobjective UAM corridor-network design that evaluates route efficiency, population exposure, and distribution of overflight burden across social-vulnerability strata.
DOI | ResearchGate | SSRN
Multi-horizon runway-configuration forecasting at São Paulo/Guarulhos, testing when data-driven models improve on strong persistence and wind-rule baselines under strict temporal validation, with transition-focused warning metrics at 15, 30, and 60 minutes.
Bayesian neural networks for probabilistic prediction of arrival delays and flight-time deviations with uncertainty information for aviation decision support.
Open-source machine-learning models for estimating commercial-flight actual takeoff weight from trajectory, operational, and meteorological data.
Recognition: Based on the solution developed by Team Tiny Rainbow (Aeronautics Institute of Technology), 2nd place in the PRC Data Challenge.
Probabilistic modeling of critical ground-time variability in aeromedical organ-transport missions.
DOI | ResearchGate | Supplementary material / code | GitHub repository
Monte Carlo scenario analysis of thermal-runaway initiation risk during eVTOL battery fast charging, including sensitivity and model-form uncertainty.
Earlier work applied supervised, unsupervised, kernel, ensemble, and distance-based machine-learning methods to soil classification from cone penetration test data.
DOI | ResearchGate | GitHub repository (MMP) | GitHub repository (Trees)
Comparison of multilayer-perceptron artificial neural networks and support vector machines for CPT-based soil classification using a diverse international dataset.
Master's research on supervised and unsupervised machine-learning methods for soil classification from CPTU, connecting the early KNN/DWKNN, ANN/SVM, tree-based, multiple-model, and kernel-clustering research line.
When a published journal version is available, please cite the version of record through its DOI. Preprints are listed when they represent ongoing work without a journal version of record.
Researcher profiles: Google Scholar · ORCID 0000-0002-8402-5832 · Lattes CV · ResearchGate · GitHub.
Publication entries provide DOI and ResearchGate links when available, together with project, preprint, conference, code, and reproducibility links where relevant. This page uses full titles, venue names, semantic descriptions, and explicit topic terminology so scholarly and general web-search systems can associate each publication with the relevant research problem.