Using ADS-B big data pattern analysis to improve the quality of multivariate 4D trajectory optimization strategies

高效和安全的航空运输系统依赖于单个和全系统优化的飞机轨迹，其中优化功能通常集中在燃料，飞行时间最小化，运输质量和环境影响作为主要市场驱动因素，或其组合。考虑到这些多重标准和航空运输系统的规模，现代高效航空运输的轨迹构建过程计算成本很高，因为其优化依赖于准确的空气动力学和发动机特性，这些特性要反映在飞机性能模型中。此外，在寻找单独优化的高度、路径和时间条件的轨迹构建过程中，应统计考虑环境不确定性源。该项目旨在确定使用大数据分析和数据驱动方法进行高效轨迹和网络设计的好处。到目前为止，使用的轨迹优化工具大多是分析驱动的，基于模型/搜索的，并且很大程度上忽略了历史轨迹作为背景。随着高性能海量数据收集和存储能力的提高，对历史飞行数据进行数据挖掘是一种鲁棒和快速优化的新选择。鉴于国际民航组织要求在2020年前安装飞机ADS-B（自动相关监视-广播）设备，这是一种飞机跟踪监视技术，现在越来越多的商用飞机配备了这种能力。在整个项目中，我们的目标是更好地理解应用大数据驱动技术预测单个（局部最优）和多个（全局最优）飞机轨迹的适用性。这种方法的最大挑战是如何处理这些大量数据，在一年多的时间内每秒处理超过20万条消息。基于我们对压缩存储和索引技术的初步联合研究，我们打算首先基于大型实体的历史ADS-B轨迹数据和特定飞机-发动机组合的飞机性能模型，寻找双方优势在单次飞行飞机轨迹增强方面的有效耦合，然后将优化目标扩展到面向流的目标函数，特别是针对小型和大型航空公司网络。对所声称的模型的验证将在中国和欧洲的空域结构和市场战略中按照显著不同的风格和行为进行，指导两国不同的航空公司网络结构。基于我们的研究成果，将有可能将历史知识转化为基于场景的建议，从而为数据驱动、目标导向的飞机轨迹设计和优化提供前所未有的准确性和效率。

A highly efficient and safe air transport system relies on individually and system-wide optimized aircraft trajectories, where optimization functions typically focus on fuel, flight time minimization, transport quality, and environmental impact as major market drivers, or a combination thereof. Given these multiple criteria and the large-scale of the air transportation system, the trajectory building process for modern, efficient air transportation is highly computationally expensive, as its optimization relies on accurate aerodynamic and engine characteristics to be reflected in aircraft performance models. Moreover, environmental uncertainty sources should be statistically taken into consideration in the trajectory building process searching for individually optimized altitude, path and time conditions.This project aims to identify the benefits of using big data analysis and data driven methods for efficient trajectory and network design. So far used trajectory optimization tools are mostly analytically driven, model/search-based, and largely neglect historical trajectories as background. The rise of performance mass data gathering and storage capabilities, data mining over historical flight data is a novel candidate for robust and fast optimization. Given ICAO’s mandate on aircraft ADS-B (Automatic Dependent Surveillance - Broadcast) equipment installation by 2020, an aircraft tracking surveillance technology, increasing numbers of commercial aircraft are now being equipped with this capability. Throughout this project, we aim to better understand the suitability of applying big data-driven techniques for predicting single (local optimum) but also multiple (global optimum) aircraft trajectories. The biggest challenge in this approach is how to process these large quantities of data, more than 200,000 messages per second over a period of more than one year. Based on our initial joint research on compressed storage and indexing techniques, we intend to first search for efficient coupling of both partners’ strengths for single flight aircraft trajectory augmentation based on historic ADS-B track data in large entities and aircraft performance models for specific aircraft-engine combinations, then extending the optimization goal towards flow oriented objective functions reflecting particularly on small and large airline networks. Validation of the claimed model will be performed along significantly different styles and behavior in the Chinese and European airspace structure and market strategies guiding both airline network structures differently. Based on our research outcome, it will be possible to turn historical knowledge into scenario-based recommendations, which yield unprecedented accuracy and efficiency for the data-driven, goal-oriented design and optimization of aircraft trajectories.