ELSA extension

From ComplexWorldWiki
Jump to: navigation, search

Contents

Introduction

The general objective of the ELSA Phase I was to analyse, describe and model the dynamics of the ATM system, especially those concerning the propagation of disturbances (performance and safety related). The analysis was carried out in the current scenario (based on real data). Methods and techniques were selected to be suitable also for the analysis of disturbances in a SESAR scenario. The project had three strands of activity:

  • an extensive statistical analysis of data of the ATM system with Complex Systems theory techniques in order to characterize statistical regularities. Example include variation in predictability during the flight execution phase, correlation between traffic and safety metrics, seasonal fluctuations, etc.;
  • the development of an Agent Based Model of increasing complexity and degree of realism, to simulate the current scenario and possibly the SESAR scenario;
  • the design of a prototype of a decision support tool, based on and informed by the results of the two other strands. The prototype mostly focused on visualising and interacting with some of the phenomena analysed by ELSA, e.g. disturbances to predictability.

In particular, the objective of the Agent Based Model (ABM) activities was to understand which are the main mechanisms underlying the ATM system behavior. The expected outcome was a tool capable of:

  1. replicating some of the behaviors observed in the real world system and identified in the data analysis part of the project. For example, the same statistical regularities observed in real data (e.g. most critical navigation points, spatial distribution of controller’s actions) should be observed in the model outcomes.
  2. supporting the analysis of the ATM working mechanisms and new concepts from SESAR

While ABM modeling process was completed to an adequate extent, the model outcomes could not be fully analysed during the project duration and the model did not reach a sufficient maturity in the current scenario to be extended to the SESAR one. However, preliminary results on the current scenario had shown a good potential for this tool to be successfully extended and to possibly be used as a research tool by the ATM community. The ELSA project extension moved from the above considerations with the following objectives:

  • improving the ELSA ABM in order to make it capable of effectively simulating different ATM operational concepts in current and SESAR scenarios on a larger spatial scale;
  • developing a Portable version of the ABM intended as a “Generic ATM simulator” to give the research community a tool to experiment innovative concepts in the ATM environment.

These objectives have been achieved by extending the model to multi-sector level, covering an area as big as an entire FIR. The strategic and tactical phases were also fully integrated to properly simulate a scenario consistent with some of the features foreseen by the SESAR Step 1 Time Based Operations [1]. Moreover, the project has delivered a modular, cross-platform simulator of air traffic management. This simulator was built upon the ELSA ABM and is designed to be used as a complex exploratory tool to study new concepts from SESAR, new organization of the airspace and/or new rules of air traffic management. The whole work has been supported by continuous validation activities. This helped in defining the right inputs to the modeling process and supported the research activities with interviews with experts and stakeholders in order to improve the effectiveness of the outcomes, in terms of the ability to replicate the SESAR scenario and the usefulness of the produced Portable ABM.

SESAR Agent-Based Model

One of the main objective of the ELSA project extension was to extend the ABM and to use it to address research questions in the current and in the SESAR scenario. The aim was to investigate the issues that affect the predictability of the last filed flight-plan within the ATM system, and what are the changes brought by SESAR in terms of airspace management and controllers' workload. The specific scientific questions the project investigated are:

  • What are the issues that affect the predictability of the last filed flight-plan within the ATM system? How is the predictability affected by these issues?
  • Can sectors capacity be improved by a more efficient management of conflicts?
  • What are the impacts of some of the changes foreseen by SESAR on the airspace management and on the controllers' workload?
  • Are these changes able to accommodate efficiently the foreseen traffic increase?

The model fully integrated the tactical and strategic layers. The tactical layer of the Agent Based Model aims at describing the interactions between flights and controllers while the strategic layer simulates the submission of flight plans by the air companies to the Network Manager, who computes the sector loads and rejects flight plans if the sector capacity has been exceeded. As in the first phase of the project, the ABM development was driven by a criterion of simplicity, in order to model only the parameters that are strictly necessary to replicate selected real features. These have been selected in order to effectively mimic the main characteristics of the reference SESAR scenario. In particular the ABM development was targeted at modelling some of the features that are foreseen in SESAR Step1 as described in the Concept of Operations[1]. In particular the consortium modelled:

  • The implementation of business trajectories by making requested trajectories from airlines progressively straighter and straighter across sectors and FIRs;
  • The increase of the traffic load according to the foreseen forecasts [2] by testing the model outcomes when varying the capacity of the airspace up to the maximum foreseen traffic load;
  • The improved coordination, information sharing and trajectory prediction available through the implementation of SWIM by simulating conflict-free planned trajectories;
  • The new ATM roles with extended look ahead and in particular the role of the multi-sector planner by modelling an extended controller look-ahead time up to 40 minutes.

The current scenario, in which all model parameters are defined from real data analysis, has been used as the baseline. For SESAR different scenarios have been considered, each of them including some of the above features. The aim was to determine the effect of each single feature (straight trajectories, conflict-free planned trajectories, increased look ahead) on the model outcomes. The results of the analyses carried out in the current and SESAR scenarios address, implicitly or explicitly, most of the four main SESAR KPA: Environment, Cost-Effectiveness, Capacity and Safety. In particular the following main results and related operational benefits can be identified:

  1. Safety improvement from the reduction of total number of conflicts and reduction of controller workload. According to the results, controller's activity in the SESAR scenario will change, moving from a situation where he has to give attention to a high number of conflicts concentrated in specific points to a situation where he will have to manage less conflicts spread in a much larger portion of the airspace. This will imply a quantitative and qualitative change of controller’s workload. Quantitative in a sense that total workload will be reduced while qualitative in the sense that nature of controllers tasks will be changed (shifting from mainly conflict resolution to mainly traffic monitoring tasks). In fact as reported in Figure 1 and Figure 2 when increasing the efficiency of the airspace structure (i.e making trajectories straighter keeping the same controller’s look-ahead) the average number of conflicts (and thus of controller’s actions) diminishes and they are spread on a wider area. Moreover, this behavior is not dependent on the number of flights thus implying that SESAR scenario will provide the same benefits also in a situation with increased traffic load. This analysis has been conducted on different ACC evidencing the same general behavior in all of them.
    Figure 1: Average number of conflicts detected in the flight trajectories of the LIRR ACC, for different values of efficiency (horizontal axis) and for different values of the aircraft present in the ACC (Nf).
    Figure 2: Density map of the conflicts (PSE) detected when considering three different levels of efficiency in the LIRR ACC. When increasing the efficiency (moving from left to right) conflicts are spread on a wider area.
  2. Capacity definition, predictability and system costs reduction. From most of the analyses carried out (see point 1) it emerged that the definition of capacity in the SESAR scenario will have to take into account the shifting of the controller workload from mainly conflict resolution to mainly monitoring tasks. Moreover, results showed that rectifying trajectories is beneficial for the general efficiency of the system. By looking at Figure 3 it can be seen that in SESAR scenario sector throughput (traffic load) is mostly lower than sector capacity due to the fact that traffic is spread across the airspace leading to more balanced traffic load among sectors. This result has been obtained in a condition where planned trajectories are perturbed not only by conflicts with other flights but also with shocks (forbidden areas representing weather events or military areas.
    Figure 3: Sectors occupancy over the 24 hours for the current (left) and the SESAR scenario (right) for three sectors in the LIRR FIR. When red is visible this indicates that planned occupancy is larger than the actual one. The opposite when yellow is visible. Orange indicates that occupancy in the planned trajectories equals capacity in the simulated trajectories.
  3. Improved airspace management. Results related to the scenario where controllers have an increased look-ahead showed that in SESAR scenario (i.e. with straight trajectories) this increase will have the consequence of reducing the number of actions shifting workload from Executive air traffic controller (EC) to Planning air traffic controller (PC), i.e. EC will have lower workload due to more monitoring tasks and less conflict resolution and separation assurance tasks. Given that controller have a tool to monitor the traffic (e.g. Medium Term Conflict Detection), all the actions could be taken at the entrance of the airspace. This could lead to a standard procedure where pilots systematically receive instructions at the entrance of a controlled area and then fly free. In fact, as reported in Figure 4, with a higher look-ahead controller’s actions will reduce in number (the highest number of actions reduces from around 18 to 8) and will be located in some specific areas thus potentially reducing the overall workload in presence of a Monitoring Tool to support subsequent tasks.
    Figure 4: Spatial location of the actions taken by the controller with low (15 min, left panel) and high (40 min, right panel) look-ahead.

Portable Agent-Based Model

The other output of the ELSA project extension is the Portable version of the Agent-Based Model named “ELSA Air Traffic simulator”. The objective of this activity was to make the ABM a modular, cross-platform simulator of air traffic management. The structure of the ABM and its functionalities were kept and effort was invested in developing a more user-friendly interface and all the required documentation and support material. The ELSA Air Traffic Simulator has therefore been designed to be used as a tool to study new concepts from SESAR, new organization of the airspace and/or new rules of air traffic management. The primary users are from academy, because the model does not provide a sufficiently realistic description of the traffic as it is for more operational users. As such, it is meant to be used as a scenario generator for synthetic data generation.

The code has been written in Python and C. The choice of non-proprietary languages ensures the continuity of the development process and allows to use heavily tested and highly optimized libraries. Python, which is a scripting language, has been chosen for its simplicity, its portability, and the presence of many scientific libraries. It has been preferred to low-level languages for the parts of the model featuring dedicated agents with limited access to memory because of its object-oriented characteristics, fitting the idea of agents. Parts of the code which required less agentification and more optimization have been written in C, a very wide-spread low-level language for which many libraries are also available. Python has also been used to provide simple interfaces to the C code, to help the user using these parts more easily.

The code has been released under the General Public License version 3, which means that it is open-source and freely downloadable. It is hosted on Github at the address https://github.com/ELSA-project/ELSA-ABM. Github provides free hosting as well as handy tools for distributed development, like a wiki, an issue tracker, etc. In other words, it allows anyone to download the code, modify it, submit the changes, discuss about them and so on. It is one of the most popular repositories for open-source software and hosts some of the most famous ones. The release has been accompanied by basic documentation, an install guide and a set of three tutorials to help the user in understanding the potential use of the simulator. The tutorials included are the following:

  1. How to adapt the current shocks module (used to model weather events and highly congested areas) and how to plug in a customized one. The user is guided through the steps needed to test its own weather models or airspace occupancy models and their relationship with ATM performances. It describes also how to include the possibility of forecasting these shocks (how they move and when and where they appear and disappear).
  2. How to define different behaviours of the controllers in each different sector. This tutorial explains how to modify the current conflict resolution module or how to plug in a customized one. The user is guided through the steps needed to fully customize the controlling parameters in each area of the airspace at different granularity levels (sector, FIR, FAB). These parameters include: directs probability, angle of re-routing, velocity change range.
  3. How to define a different structure of the airspace. This tutorial explains how to modify the current airspace generator module or how to plug in a customized one. The user is guided through the steps needed to generate the airspace structure according to a set of predefined parameters such as: average size of sectors, density of sectors in each FIR. It also describes how to define manually the FIR boundaries along with the sectors inside it and each single navpoint.

Potential users have been involved from the early phases of the project to ensure that their needs, wants, and limitations were given adequate attention at each stage of the design process. The main target users identified included the ATM research community and in particular experienced researchers, PhD students and anyone involved in ATM and more specifically in SESAR related research activities from various Universities and Research Centres all over Europe (NLR, Innaxis, ENAC, TU-Delft, Universities of Rome, Belgrade, Bologna, Hannover). The requirements gathered by the target users have been taken into consideration during the development process.

Conclusions

The results achieved and the foreseen benefits for the SESAR community coming from the work carried out during the ELSA project extension are the following:

  • Measuring in quantitative terms the degree of optimisation brought by SESAR concepts in normal conditions, including also aspects like the number and type (e.g. tactical vs. planning) of ATCOs’ actions,
  • Providing a tool with the capability of testing different future scenarios, for instance by simulating the transition period, when a SESAR concept is being implemented, but not fully deployed,
  • Providing a common platform to be used by the SESAR research community, to make it easier to compare different research programs, or to provide synthetic data for data mining algorithms testing.
  • Building a community of researchers interested in exploiting the potential of Agent-Based Modelling in ATM research.

Future research based on ELSA project results

Two main results of the whole ELSA project can be exploited for future research. First, methods and tools developed within ELSA to numerically describe the current ATM system at the EU level could be used to measure the benefits brought by the implementation of SESAR concepts, e.g. integration of sectors in FABs, user-preferred trajectories. They can also be used to identify proxies of behavioural patterns in the ATM system at large. Analysis of these can allow to dig into the complexity of the ATM system and identify non-trivial deviations. Performance improvements could then be gained by addressing research questions such as:

  • What are the drivers behind these deviations? What are their precursors or early warning signals?
  • How can they be replicated/eliminated?

Secondly, the ELSA simulator itself can be potentially used to run different scenarios, applying the ELSA analyses to measure with rigorous quantitative metrics the achieved optimisation, compare different solutions, carry out stress tests to see how various configurations cope with shocks like strikes, bad weather, large volcanic ashes, and so on. Future research can also build on the current version of the simulator by extending its scope to make it capable of investigating new concepts such as the impact of information sharing on the airlines business choices, or the system resilience in presence of major disruptive events.

Refrences and published papers

  1. 1.0 1.1 SESAR, “SESAR Concept of Operations Step 1,” 2012.
  2. EUROCONTROL, “Challenges of Growth – Task 4: European Air Traffic in 2035”, 2013
Personal tools
Namespaces

Variants
Actions
Navigation
Toolbox
Share