Elsevier

Neurocomputing

Volume 391, 28 May 2020, Pages 167-176
Neurocomputing

LSTM-Cubic A*-based auxiliary decision support system in air traffic management

https://doi.org/10.1016/j.neucom.2019.12.062Get rights and content

Abstract

Airports, pilots and geographical environment contribute a lot to stable and safe air traffic, especially during the climbing and descending phases. Accidents occur frequently in these two phases, even small errors can lead to serious consequences. Modern equipment has made great progress in monitoring the airspace. However, when encountering strong external interference or blind zones, potential risks will be raised. With the assistance of newly developed technologies and equipment, flight trajectories can be recorded at high frequency, which enhances capabilities of Air Traffic Management systems. A four-dimensional flight trajectory prediction model is put forward in this paper. Combined with sliding windows, Long Short-Term Memory network maintains the long-term features and manages to predict accurate trajectories. Massive data contributes a lot for forecasting, but cannot guarantee better decisions. Decision-making is closely related to threat assessment. Taking the geographical environment into consideration, the static threat is involved in the current work. The experimental scenario is set to be an aircraft takes off from Hong Kong International Airport, which is located close to mountains. Characteristics of terrains are introduced to our system by digital elevation model. The proposed cubic A* search algorithm provides reasonable path by considering the ultimate dynamic performance of an aircraft. Threat is assessed during the path planning process. Finally, the auxiliary decision support system is developed based on ArcGIS 10.0, to graphically provide the intuitive and quick assistance. Multiple sets of experimental results indicate that this system is able to provide timely decision support.

Introduction

Accidents occur frequently when the aircraft take off and land. Even small errors can lead to serious consequences. MH370 crashed in March, 2014, which shocked the whole world but does not stop worldwide aviation accidents. ET302 disappeared from radar and dropped in six minutes after departure in March, 2019. Once a civil aircraft crashes, it will inevitably cause loss of lives and properties. Driving skills and experience of pilots contribute a lot to our safe travel. The most famous example happened on the plane 3U8633 in 2018, whose windshield was broken and felled off, the pilot managed an emergency and safe landing without loss of properties and lives. In addition to the factors of skilled pilots, infrastructures of airports are also crucial, including the construction sites, surrounding environment and corresponding emergency facilities. Therefore, most of the world’s airports are well-designed to provide the most comprehensive protection for the aircraft, except for some special cases. There are such airports around the world, some of which are located in the mountains, some are located in the middle of a big city, some runways extend straight into the ocean. Most famous “dangerous” airports distribute in Kansai International Airport in Japan, Qamdo Bangda Airport in China, Funchal Airport in Portugal and to name but a few. Once an aircraft took off from Hong Kong International Airport (HKG) with abnormal heading, where mountains located just in front of it. The pilot managed to fly over the mountain after receiving a warning from the ground tower. Though no accident occurred, we should raise enough attention to trajectory management and decision-making.

Furthermore, frequent aviation activities brought by increasing aircraft in a fixed amount of airspace also result in certain burdens and potential risks, as they can cause severe air traffic jam. According to International Civil Aviation Organization (ICAO) [1], the growth rate of air traffic has expanded twice every 15 years since the mid-1970s. In the year 2014, the number of passengers reached 330 million. While the air freight value exceeded $ 6.4 trillion annually [2]. Any single or combination of the above-mentioned aspects will bring unprecedented challenges to the timeliness and accuracy of Air Traffic Management/Control (ATM/ATC).

Trajectories should be predicted in the ground-based ATM systems to better plan traffic flows, reduce delays, operating costs and minimize adverse environmental impact, according to the instructions of Base of Aircraft Data (BADA) in EUROCONTROL [3]. Modern equipment has been greatly improved in terms of type and data capacity, thereby enhancing their capabilities in monitoring and managing trajectories. A large number of multi-modal data is generated and accumulated in data processing centers. However, massive data contributes a lot for forecasting, but cannot guarantee better decisions. Decision-making is closely related to threat assessment. Geographic information system (GIS) manages data by indexing spatial-temporal location. Both related and unrelated information can be integrated by using location as the key index variable in GIS, which makes it able to provide auxiliary decision-making in wide-area applications.

In our early work [4], Long Short-Term Memory (LSTM) network embedded with sliding windows was proposed for four-dimensional (4-D) flight trajectory prediction. Current work can be carried out on this basis, the framework of this research is shown in Fig. 1. Our proposed auxiliary decision support system (ADSS) consists of four functional parts. Specifically, they are four-dimensional trajectory prediction (4-DTP) model, path planning model, threat assessment (TA) model and visual display platform, respectively. We take the unique geographical environment of HKG as an experimental example, which is surrounded by mountains and sea. 4-DTP model is trained by massive historical flight trajectories of climbing phase recorded by multiple ADS-B ground stations. Geographical environment around HKG is modeled as a discrete gridded space by digital elevation model (DEM). A cubic A* algorithm is proposed for aircraft path planning based on the predicted state in 4-D space. Threats between aircraft and the mountains are assessed along the optimally planned path. Overall, the above-mentioned models are embedded to our auxiliary decision support system, which is built based on ArcGIS 10.0 so that to provide Geo-environment information and spatial-temporal mapping. Our system has been proven to provide stable and effective assistance in practical circumstance by multiple tests. This work mainly contributes in following aspects:

  • 1.

    Different from existing systems, which apply conventional models for trajectory prediction, for example, Markov Models. Our system relies on Neural Networks for accurate trajectory prediction and applies Information Fusion technologies to provide foundations for decision-making:

    • LSTM is the core technology in 4-DTP model, which mines dependency between spatial and temporary characteristics to make an accurate prediction on 4-D flight trajectory;

    • Information Fusion technologies defined in JDL model (situation awareness) are involved to assess the impact of geographical environment.

  • 2.

    In order to shorten the time required for data analysis and provide a reliable path for emergencies. Airspace is divided into discrete adjacent cubes according to the Rate of Climb (RoC) of specified type of aircraft.

    • Cubic A* search algorithm is proposed considering the dynamic performance of aircraft. In the climbing phase, maximum yaw angle and RoC limits the spatial span of the aircraft, which makes our model able to find the optimal 3-D path in a very short time;

    • Threat factors consist of real-time speed, altitude and track angle of the aircraft along with the planned path. Threats caused by mountains and no-fly zones are assessed to ensure reliability of the optimal path.

The rest of our research is organized as follows: Related works are introduced in Section 2. Section 3 presents the methods on prediction, path planning and threat assessment. Experimental results and construction of the auxiliary decision support system are addressed in Section 4. Section 5 concludes our work and provides some possible extensions.

Section snippets

Related works

Air traffic management systems based on 4-D trajectory-based operations (TBO) were proposed successively in the United States and Europe, which were named as the Next Generation Air Transportation System (NextGen) [5] and Single European Sky ATM Research (SESAR) [6], respectively. Modern technologies and equipment, such as Automatic Dependent Surveillance-Broadcast (ADS-B) [7], made the aircraft visible and provided a more accurate report of their positions. The cooperation between ADS-B and

Construction of ADSS

In this Section, we first introduce the preliminaries on our proposed models, which is followed by the descriptions on the main models. ArcEngine provides the GIS component library for secondary development. On this basis, we embed functional modules to develop the ADSS.

Experimental results

We start this Section with data collection. According to the data processing flow, 4-DTP model is first performed to generate an accurate prediction of the aircraft. Path planning and threat assessment models are triggered simultaneously based on the current and predicted states of the aircraft. ADSS visually display real-time scenarios and the outputs of each module to provide auxiliary services.

Conclusion

An auxiliary decision support system based on ArcGIS 10.0 for 4-D trajectory management was presented in this paper. We aim to provide auxiliary decisions or options for the operators in ATM systems. The fundamental function of our system is to display static geographic elements and dynamic flight trajectories simultaneously. Three additional functions are embedded in ADSS to provide auxiliary information. Firstly, the optimized LSTM network combined with sliding windows was applied in 4-D

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgment

Our research is partly supported by the National Natural Science Foundation of China under Grant 61790552.

Zhiyuan Shi received the B.E. degree in measurement and control technology and instrument from Anhui University, and the M.E. degree in control theory and control engineering from Northwestern Polytechnical University (NPU) in 2012 and 2015, respectively. He is currently working toward the dual Ph.D. degree in control theory and control engineering in NPU and computer science in faculty of engineering and information technology in University of Technology Sydney (UTS). His research interests

References (68)

  • L. Snidaro et al.

    Fusing uncertain knowledge and evidence for maritime situational awareness via Markov logic networks

    Inf. Fus.

    (2015)
  • Y. Han et al.

    An evidential fractal analytic hierarchy process target recognition method

    Defence Sci. J.

    (2018)
  • A.T. McCartt et al.

    Evaluating group decision support system effectiveness: A performance study of decision conferencing

    Decis. Support Syst.

    (1989)
  • S. Biswas et al.

    Anomaly detection via short local trajectories

    Neurocomputing

    (2017)
  • S.J. Landry et al.

    A decision support methodology for dynamic taxiway and runway conflict prevention

    Decision Support Syst.

    (2013)
  • M. Liang et al.

    Conflict-free arrival and departure trajectory planning for parallel runway with advanced point-merge system

    Transp. Res. Part C Emerg. Technol.

    (2018)
  • C. Di Ciccio et al.

    Detecting flight trajectory anomalies and predicting diversions in freight transportation

    Decision Support Syst.

    (2016)
  • E. Koksalmis et al.

    Deriving decision makers weights in group decision making: an overview of objective methods

    Inf. Fus.

    (2019)
  • B. Khaleghi et al.

    Multisensor data fusion: a review of the state-of-the-art

    Inf. Fus.

    (2013)
  • U. specialized agency, Administration and governance of the convention on international civil aviation (chicago...
  • C. Montreal, The trade association for the world’s airlines, [Online]. Available: https://www.iata.org...
  • E.E. Center

    User manual for the base of aircraft data (bada), revision 3.6

    EEC Note

    (2011)
  • Z. Shi et al.

    Lstm-based flight trajectory prediction

    Proceedings of the International Joint Conference on Neural Networks (IJCNN)

    (2018)
  • D. Washington, Federal aviation administration nextgen implementation plan, [Online]. Available:...
  • E. European Union, European ATM master plan, [Online]. Available: https://www.sesarju.eu/...
  • Airservices

    How ADS-B works

    Australia

    (Nov. 2012)
  • T. Yong et al.

    ADS-B and SSR data fusion and application

    Proceedings of the IEEE International Conference on Computer Science and Automation Engineering (CSAE)

    (2012)
  • J.A. Besada et al.

    Data fusion algorithms based on radar and ads measurements for atc application

    Proceedings of the Record of the IEEE International Radar Conference [Cat. No. 00CH37037]

    (2000)
  • D. Jeon et al.

    Estimation fusion with radar and ADS-B for air traffic surveillance

    Int. J. Control Automat. Syst.

    (2015)
  • L. Song et al.

    Evaluating location predictors with extensive Wi-Fi mobility data.

    Proceedings of the IEEE INFOCOM

    (2004)
  • D. Ashbrook et al.

    Using gps to learn significant locations and predict movement across multiple users

    Personal Ubiquitous Comput.

    (2003)
  • P. Lingras et al.

    Prediction of recreational travel using genetically designed regression and time-delay neural network models

    Transp. Res. Record

    (2002)
  • M. Elhoushi et al.

    A survey on approaches of motion mode recognition using sensors

    IEEE Trans. Intell. Transp. Syst.

    (2017)
  • J. Roskam

    Airplane flight dynamics and automatic flight controls

    (1998)
  • Cited by (9)

    • A hybrid optimization algorithm and its application in flight trajectory prediction

      2023, Expert Systems with Applications
      Citation Excerpt :

      A method based on covariance bidirectional extreme learning machine was proposed for trajectory prediction and fuel consumption estimation (Khan et al., 2021). The Long-Short Term Memory (LSTM) network (Hochreiter & Schmidhuber, 1997) was combined with a sliding window to predict flight trajectory (Shi et al., 2020). LSTM and the deep feedforward neural network (DNN) were merged to form a hybrid flight trajectory prediction model (Zhang & Mahadevan, 2020).

    • Data fusion for ITS: A systematic literature review

      2023, Information Fusion
      Citation Excerpt :

      The mentioned traffic data are broadly applied for various traffic applications. First, the historically recorded information supports predictions [15,16,55,57,67,70–91], planning [92–94], behavior reporting and analysis [95–99]. Second, streaming or (near) real-time traffic data can be useful for automated driving cars [56,100], incident detection [11–13], advanced travel information systems (ATIS) (e.g., V2X communication [101], positioning/tracking [102–105], parking assistance [106]), and traffic signal priorities [107], which require the most up-to-date data so that applications can take the best decision.

    • Flight Trajectory Prediction Method Based on Attentional Graph Convolutional Network

      2023, 2023 3rd International Symposium on Computer Technology and Information Science, ISCTIS 2023
    View all citing articles on Scopus

    Zhiyuan Shi received the B.E. degree in measurement and control technology and instrument from Anhui University, and the M.E. degree in control theory and control engineering from Northwestern Polytechnical University (NPU) in 2012 and 2015, respectively. He is currently working toward the dual Ph.D. degree in control theory and control engineering in NPU and computer science in faculty of engineering and information technology in University of Technology Sydney (UTS). His research interests include time series prediction, decision and registration.

    Quan Pan received the B.E. degree from Huazhong Institute of Technology in 1991, and the M.E. and Ph.D. degrees from the School of Automation at NPU in 1991 and 1997, respectively. He is now a professor in School of Automation of NPU and the head of Key Laboratory of Information Fusion Technology (LIFT), Ministry of Education, China. His research interests include information fusion, hybrid system estimation theory, multi-scale estimation theory, deep neural networks, machine learning and image processing. He has published 11 books, and almost 400 international journal/conference papers, including in IEEE Transactions on Automatic Control, Automatica, and IEEE Transactions on Signal Processing. He obtained the 6th Chinese National Youth Award for Outstanding Contribution to Science and Technology in 1998 and the Chinese National New Century Excellent Professional Talent in 2000.

    Min Xu received the B.E. degree from University of Science and Technology of China, M.S. degree from National University of Singapore and Ph.D. degree from University of Newcastle, Australia. Dr. Xu was a full-time researcher in Nanyang University of Technology, Singapore. She was awarded UTS Chancellor Postdoctoral Fellowship. Currently, she holds a full-time academic position (Associate Professor) in the School of Electrical and Data Engineering, Faculty of Engineering and IT, UTS. She has made significant contributions to various areas in multimedia data (video, audio and text) analytics and computer vision. Recently, she is focusing on applying machine learning algorithms (e.g. deep neural networks) for multimedia applications, including affective computing, image caption and action recognition. She has published over 150 research papers in high quality international journals and conferences. Over 1500 citations of her research papers show her reputation in her research field.

    View full text