Probabilistic conflict detection for commercial aircraft near airports

[摘要] ENGLISH ABSTRACT: Increasing air traffic and urbanisation has led to a cluttered airspace, particularly nearairports, where both complex terrain and multiple moving obstacles are frequent. Accuratelyand efficiently predicting violations in safe separation criteria for commercialaircraft, a process called conflict detection, is therefore crucial in assessing risk associatedwith threats of collision. Existing avoidance systems in operation such as TCAS, EGPWSand ATC exhibit shortcomings, leaving room for uncertainty and possible conflict scenarios.A single on-board system capable of minimising errors in prediction would informconflict resolution decisions more accurately as well as support the notion of free flight, anobjective of next-generation air traffic management systems.This thesis investigates the viability of a modern algorithm, probability flow, as a method ofprobabilistic conflict detection for commercial aircraft in airport environments. Simulationresults for realistic flight scenarios are presented in comparison with a ground-truth resultobtained through Monte Carlo simulation. Observations are made regarding the suitabilityof probability flow for real-world application. It is found that probability flow is capableof calculating a tight upper bound to the probability of conflict quickly and accuratelyfor most conflict scenarios. However, unreasonably large overestimates on the probabilityof conflict are obtained when flying parallel to an obstacle conflict region. This problemcould lead to a high frequency of false alerts, particularly in aborted landing scenariosand at airports operating parallel runways. It is therefore advised that further researchbe conducted to resolve this problem before probability flow can be reliably implementedin an airport environment.

[发布机构] Stellenbosch University