Report

Teaser, summary, work performed and final results

Periodic Reporting for period 4 - DART (Data-driven AiRcraft Trajectory prediction research)

Teaser

DART delivers understanding on the suitability of applying data-driven techniques for predicting single aircraft trajectories without considering traffic, as well as agent-based methods for assessing the impact of traffic to individual aircraft trajectories, thus accounting...

Summary

DART delivers understanding on the suitability of applying data-driven techniques for predicting single aircraft trajectories without considering traffic, as well as agent-based methods for assessing the impact of traffic to individual aircraft trajectories, thus accounting for complex phenomena due to co-occurring multiple trajectories.

As part of this objective DART emphasizes the role modern visualization techniques can have in facilitating trajectory predictions.

Towards this high-level research objective, the following specific research objectives have been accomplished:
• Definition of requirements for the input datasets towards increasing trajectory prediction accuracy.
• Study of big-data techniques to trajectory related data gathering, filtering, storing, prioritization, indexing or segmentation to support the generation of reliable and homogenous input datasets.
• Study of data-driven learning techniques to single trajectory prediction.
• Formal description of traffic and interacting trajectories for resolving DCB problems at the pre-tactical stage of operations.
• Study of agent-based models to assess the impact of traffic to individual trajectories, in the context of DCB problems.
• Description of visualization techniques to enhance trajectory data management capabilities.
• Exploration of advanced visualization processes for algorithms formulation, tuning and validation, in the context of 4D trajectories.

The overall DART concept is shown in the DART concept figure, while the DART work structure is depicted in the DART work package structure.

Innovative results produced in DART aim to show contribution to the SESAR Programme strategic objectives, the most crucial of which is predictability: Indeed, TBO is a key element in the SESAR Programme. One of the cornerstones of this paradigm is the improved predictability of the ATM system (both at traffic and individual trajectory levels). Additionally, improvements in trajectory prediction are fully aligned with FlightPlan 2050 goals, for instance weather-independent arrival punctuality.
Improved forecasts at pre-tactical phase are a matter of interest of all actors of the system, as would allow better resource allocation and reduce extra costs (both delay and fuel). Viability depends on the effective accessibility of information in pre-tactical phase.

Work performed

During this fourth and final period, the project performed according to planned work the following tasks:
WP1:
- Supported the linkage and exploitation of data provided.
- Extended datasets to include full year 2017 data.
- Provided data accessibility over 99.99% of the time though Data Transaction Pipeline.
- Refined visualization and visual analytics methods with capabilities for exploration and comparison of input data.

WP2:
- Optimized the model parameters for trajectory prediction methods.
- Performed experiments with the data-driven trajectory prediction methods proposed, exploiting various trajectory features and also, in some cases, in combination with model-based trajectory prediction approaches.
- Performed experiments with robust but much more complex data-driven architectures.

WP3:
- Devised an extensive set of experimental cases of various complexity/difficulty and traffic conditions.
- Performed an extensive set of experiments towards tuning methods’ parameters, and comparing the potential of the four proposed multi-agent reinforcement learning methods to achieve qualitative solutions for solving real-world DCB problems of increased complexity.
- Proposed visualizations of solutions’ overview in time and space, providing a comprehensive way of assessing and comparing solutions quality.

WP4:
- Project management activities.
- Organization and coordination of dissemination activities, including scientific publications and and DART newsletters
- Organization of two workshops in conjunction to WAC 2018 & ICRAT 2018
- Organized two DART Working Group meetings, to present progress achieved and discuss project results.

Final results

Two important drawbacks of state-of-the-art methods for predicting aircraft trajectories are that (a) they are limited to single trajectory predictions, and (b) their prediction horizon is a short time one. Consequently, the network effect resulting from the interactions of multiple trajectories is not considered at all, which may lead to huge prediction inaccuracies due to several reasons. This is due to the complex nature of the ATM system, which impacts the trajectory predictions in many different ways.

The design of the new data-driven methods for single trajectory prediction are focusing on increasing predictability and improving scalability, in order to be able to handle extended airspaces and large volumes of data simultaneously.

Multiple algorithms implemented for single trajectory prediction in DART are exploring different directions: Either by ingesting raw data, or enriched surveillance datasets with additional variables, or derived datasets such as enriched trajectories and AIDL datasets. In doing so, DART provides a comprehensive evaluation of state of the art machine learning algorithms for single trajectory prediction, also combined with clustering and model-based approaches, exploiting varying features concerning 4D trajectories.
Algorithms benchmarking activities included comparison between data-driven predictions versus flown trajectories, and data driven predictions versus Eurocontrol Network Manager pre-flight prediction. Results show that data-driven methods can achieve high accuracy in predicting trajectories, also when they exploit information about flight plans.
Validation results help to understand that the combination of models and data-driven approaches is the correct way to evolve current operational systems towards the implementation of Trajectory Based Operations (TBO).

Towards delivering an understanding on the suitability of applying agent-based modelling techniques taking into account multiple trajectories and towards assessing the impact of traffic to individual trajectory predictions, the focus is on the DCB problem in Air Traffic Management, whose solution takes place at the pre-tactical stage: our objective is to predict delays that are applied to the flights.

To this end, DART makes the following contributions:
DART provided two formulations of the demand-capacity balance (DCB) problem using multi-agent Markov Decision Processes (MDP), modelling flights as agents whose decisions range in the space of their preferred/allowed delays: A “flat” and a hierarchical model.
DART designed and devised four multi-agent reinforcement learning methods towards assessing the impact of traffic to individual trajectories for resolving the DCB problems at the planning phase.
These methods provide a shift-of-paradigm for regulating flights: Each agent, corresponding to a trajectory decides its own delay w.r.t to operational constraints, own constraints on delays, and of course, according to the cost of strategic delay imposed.
Evaluation results for all methods show that the proposed methods are capable to resolve hotspots, while keeping the average delay for flights at low levels, also compared to CFMU regulations, with fairness. All methods can incorporate stakeholders’ constraints on flight delays.

Website & more info

More info: http://dart-research.eu/.