Robust data fusion for 4D conflict-free optimal trajectories in a highly automated ATM system

Introduction

The improvement of air transport efficiency (in terms of economic and environmental impact) is one of the major drivers for research and development in the SESAR and NextGen programmes. New technologies and procedures for future ATM and on-board systems and operations are being investigated and proposed. For example, several research has been done in the integration of Continuous Descent Approaches (CDA) in dense TMAs [1,2]. The Oceanic Tailored Arrivals program, currently in place in San Francisco airport [3], is another relevant example. These arrivals are supported by the Efficient Descent Advisor (EDA) developed by NASA-AMES, which is able to compute conflict-free optimal descent trajectories and satisfy a given arrival fix metering [4]. Even if this research is indeed very promising and is setting the foundations for future applications, we are still far to fulfil SESAR objectives in terms of significantly improving the trajectory efficiency in terminal airspace. One requirement for efficient flow of air traffic is accurate knowledge of aircraft positions and adequate prediction of future aircraft movements. Without accurate information, maintaining safe separation between aircraft requires much more conservative, and therefore less efficient, methods [5,6]. Moreover, a large scale implementation of these optimised trajectories (i.e. not only for a small set of aircraft but for all those in a dense and complex TMA) remains an issue that will obviously require more levels of automation. Nowadays, focus is given to optimal arrival trajectories and optimised aircraft departures have hardly been investigated (see for instance [7]). However, their de-confliction with optimised arrivals is a problem to address as well. To depict that, the following figure shows the inclusion of a CDA (in green) as opposed to previous stepped descents (in red) and the impact into the climb corridor that was previously in place.

The computational burden may become an issue since real-time or quasi-real-time trajectories are required. Therefore, some research should be done in investigating new optimisation techniques, while assessing “where” this optimisation takes place: i.e., on-board the aircraft, where robust de- confliction with other traffic and coordination with ATC could pose a problem; on ground (like the in the EDA implementation), where the optimisation may suffer from the lack of important parameters (such as the actual mass of the aircraft or its trajectory intents) and communications delays should be considered in the algorithms; or a combination of both.
Other sources of inefficiency may come from the de-confliction algorithms, which at present only seek to ensure a conflict-free solution but hardly assess the optimality of this solution. Moreover, this solution is highly dependent to uncertainties such as winds at different altitudes and aircraft performances. Finally, it is widely known that the objectives for trajectory optimisation can be highly conflictive. For example, noise optimal trajectories are far from being optimal from a fuel or emissions point of view. Therefore, these trade-offs must be also taken into account in the generation of optimal flight trajectories.
The PhD research proposed in this document will address all these issues with the objective of implementing optimised 4D trajectories in high density TMAs, thus, increasing the aviation efficiency; coping with the expected increase of traffic; reducing the environmental impact of air transport; and, even if it is not the main concern of this research, contributing to improve the current levels of safety. This research is expected to mainly contribute to two Objectives of the HALA! Network: 4D Trajectory Management, by applying advanced techniques for trajectory modelling, optimisation and exchange; and advanced Decision Support Tools, since this project has a direct impact in AMAN and DMAN tools (Arrival and Departure MANagers). Indirectly, this work will also contribute to two other topics: network centric ATM services, since it will be necessary to specify needs for the communication network and involved protocols (in a SWIM environment); and, given the levels of automation proposed, the integration of UAS in controlled airspace studies will take advantage of the results of this research too.

Aims and Objectives of the Research

To achieve higher levels of automation within the Air Traffic Management (ATM) system in the context of SESAR, we propose to investigate information integration methods to improve the assessment of hazardous situations along 4D trajectories of manned and/or unmanned aircraft in non-segregated airspace and to perform online monitoring of the available and predicted performance in order to compute real-time optimal conflict-free trajectories. These methods will exploit both on-board sensors, data-linked information and information available through SWIM and from the ground infrastructure. Hazards will include traffic, weather, wake turbulence, terrain, and other obstacles. Performance based operations will be defined in terms of accuracy, integrity, availability, continuity and other applicable performance parameters. The specific objectives of this research are summarised as follows:

  • Develop an optimisation framework for aircraft departure, en-route and arrival trajectories.
  • Study the requirements in terms of trajectory modelling, description and data integration.
  • Study the possible data to be exchanged with other aircraft or ground facilities, and propose mechanisms for this purpose.
  • Study strategic trajectory de-confliction mechanisms while assessing optimality of the solution and considering the performance indicators of the navigational data.
  • Give a first assessment on the decision support tools needed on-board and on the ground
  • Evaluate two different case studies:
    • operations in a complex and dense TMA (heterogeneous flows and aircraft fleets)
    • oceanic (or remote) operations (parabola-like trajectories)

As commented before, accurate and reliable information on aircraft states is paramount when computing the 4D trajectories proposed in this research. Thus, an special emphasis will be given to data integration methods, which will:

  1. Allocate information sources to required functionality and operational regimes (i.e. conflict detection and avoidance function, navigation using Required Navigation Performance (RNP) or Area Navigation (RNAV), or continuous descent approaches with merging and spacing) and determine the required performance of these information sources as part of that function
  2. Monitor or evaluate the required performance of the individual information sources and perform consistency checks among various information sources if they are dissimilar
  3. Integrate the information to establish tracks of potential hazards that can be used for the conflict probes or conflict prediction for various time horizons including near-term (10 minutes) for decision support for ATM
  4. Detect and assess the class of the hazard, computation of near-term conflict-free trajectory for negotiation and/or coordination with ATM or nearby aircraft, or very near-term conflict resolutions
    The integration method will take into account that 4D trajectory constraints must be met including Required Time of Arrival (RTA) or Controlled Time of Arrival (CTA) at a Trajectory Control Point (TCP). This is especially important when establishing new conflict-free trajectory in coordination (implicit or explicit) with ATM or other aircraft.

Theory and Methodology

Our research approach will be incremental considering an aircraft-centric approach initially and slowly moving to a network-centric approach, where all parties involved will be exchanging and integrating information, i.e.: (1) the aircraft (or UAS) by itself tries to establish an accurate conflict-free/hazard- free path that stays as close to his originally negotiated trajectory with associated predicted Navigation and Surveillance performance parameters; near term/tactical decision making (0-3 minutes); (2) like (1) but now the aircraft negotiates a new conflict-free path with ATM (all automated); long-term decision making (2-20 minutes); and (3) like (1) and (2), but now also include aircraft-to-aircraft negotiations (0-3 minutes). The sensors can include all sensors that are already in existence (onboard avionics and ground infrastructure), sensors planned for SESAR, i.e. Automatic Dependent Surveillance (ADS) In and Out, or sensors for UAS that may come in future systems for deal with non-collaborative targets (cameras, radars etc.).
Kalman Filters (such as Multi-mode Kalman Filter or an Interacting Multiple Mode Kalman Filter) are proposed to be used for tracking potential hazards. In case of traffic, ownship state measurements are combined with traffic state measurements from TCAS II (Traffic Collision Avoidance System), ADS, TIS (Traffic Information Systems) or other sources. In case multiple sources provide the same information (independently), the Kalman Filter can be set up to take the multiple inputs and perform a consistency check as well [8]. The potential hazards can also be weather, terrain or other obstacles in which case the state vector must be augmented to include some model of extend (i.e. weather is modelled by one or more geometric shapes with associated position, velocity dimension, and severity attributes). Using the filtered results to obtain a more reliable estimate of velocity (and, in some cases acceleration) reduces the amount of uncertainty in the prediction for longer time horizons (> 1 minute for example).
Trajectory modelling and optimisation has been a subject widely researched in the last decades. Analytically, this optimisation problem can be formally written as a continuous optimal control problem and extensive research on its resolution can be found in the literature. However, realistic trajectories are hardly impossible to solve analytically and a wide variety of numerical solutions have arisen. One of the most relevant ones involves the direct transcription of the problem, leading to a Non-Linear Programming (NLP) problem with a finite set of decision variables [9]. This approach will set-up the basic theoretical background for the research proposed for this PhD. Nevertheless, it will be necessary to enhance and complement this optimisation methodology in order to cope with the difficulties that will arise from the fact that the optimisation may be conducted partially on-board, on ground or perhaps in a distributed manner; the (quasi-)real-time requirements for this problem; the high amount of constraints that conflict-free trajectories will generate; the multi-objective nature of the problem, etc. Therefore, other methodologies will also be explored in order to complement the optimisation framework, such as constrained programming [10], evolutionary algorithms or multi-objective optimisation strategies [11].

References

[1] Kuenz, A.; Mollwitz, V. & Korn, B. “Green Trajectories in high traffic TMAs” 26th Digital Avionics Systems Conference (DASC), 2007
[2] Ho, N.T. & Clarke, J-P. “Methodology for Optimizing Parameters of noise abatement approach procedures”. Journal of Aircraft, 44 (4) pp. 1168-1176. 2007.
[3] Coppenbarger, R.; Lanier, R.; Sweet D. & Susan Dorsky. “Design and Development of the En Route Descent Advisor (EDA) for Conflict-Free Arrival Metering”. AIAA Guidance, Navigation, and Control Conference, Providence, Rhode Island, 2004.
[4] Coppenbarger, R.; Mead, R. & Sweet, D. “Field Evaluation of the Tailored Arrivals Concept for Datalink-Enabled Continuous Descent Approach”. 7th AIAA Aviation Technology, Integration and Operations Conference (ATIO) Belfast, Northern Ireland, 2007.
[5] Roy, K.; Levy, B.& Tomlin, C. “Target tracking and Estimated Time of Arrival (ETA) Prediction for Arrival Aircraft”. AIAA Guidance, Navigation, and Control Conference, Keystone, Colorado, 2006
[6] Tadema, J.; Theunissen, E. ; Rademaker, R.M. & Uijt de Haag, M. “Evaluating the impact of sensor data uncertainty and maneuver uncertainty in a conflict probe,” Proceedings of the 29th Digital Avionics Systems Conference, Salt lake City, October 3-7, 2010.
[7] Heblij, J. & Visser, G. “Advanced noise abatement departure procedures: custom optimized departure profiles”. AIAA guidance, navigation and control conference. Honolulu, Hawaii, 2008.
[8] Bezawada R. & Uijt de Haag, M. “Evaluation of Hazard and Integrity Monitor Functions for Integrated Alerting and Notification using a Sensor Simulation Framework,” Proceedings of the SPIE Defense and Security Symposium, Enhanced and Synthetic Vision Conference, Vol. 7689, Orlando, FL, April 2010.
[9] Betts, J. “Practical methods for optimal control and estimation using nonlinear programming.” Advances in design and control, 19. Society for Industrial and Applied Mathematics, 2010
[10] Barnier, N. & Allignol, C. “Deconfliction with Constraint Programming” International Workshop on Constraint Programming for Air Traffic Control & Management, Brétigny sur Orge, 2008.
[11] Miettinen, K. “Nonlinear Multiobjective Optimization”. Kluwer Academic Publishers, 1999.