弹道
颗粒过滤器
计算机科学
粒子(生态学)
控制理论(社会学)
航空航天工程
人工智能
卡尔曼滤波器
工程类
物理
地质学
天文
海洋学
控制(管理)
作者
Ioannis Lymperopoulos,John Lygeros
出处
期刊:AIAA Guidance, Navigation and Control Conference and Exhibit
日期:2008-06-15
被引量:8
摘要
Trajectory prediction constitutes a fundamental function in air traffic management systems as it can both enhance the capacity and increase the safety of the airspace. Inherent inaccuracies due to aircraft modeling discrepancies and weather forecasts uncertainty may result in large prediction errors. Moreover, current flight management systems don’t generally correct for along track deviations from the flight plan thus leading to substantial trajectory uncertainty. We formulate aircraft trajectory prediction as a Bayesian estimation problem. We employ a stochastic non-linear hybrid system to model the aircraft dynamics. Partial information about the system state is acquired sequentially through radar measurements. Such observations incorporate information about the missing parameters of the system and the effect of the stochastic part of the dynamics, representing weather uncertainty, on the system state. We can extract this information and exploit it in order to perform both parameter and weather identification. As this information builds-up we use it to improve the trajectory prediction accuracy. However the nonlinear, non-Gaussian nature of the model prevents the use of traditional filtering methods such as Kalman filtering, which may other-wise provide optimal solutions for linear systems with additive Gaussian noise. To confront this problem we employ sequential Monte Carlo methods (particle filters) in order to provide a numerical solution to the recursive Bayesian estimation problem, without compromising our aircraft model.
科研通智能强力驱动
Strongly Powered by AbleSci AI