The next generation air transportation system (NGATS) will benefit from a variety of system improvements that will enhance capacity and alleviate congestion. These include improvements in the flight deck and avionics, vehicle performance, communications, navigation, flight planning, and air traffic control and management service provider capabilities. In the area of traffic flow management, several tools and concepts are available or under development. These tools attempt to solve the problem of how to guide flights in capacity-constrained scenarios. Pre departure gate holding, and pre- and post-departure rerouting are typical means of dealing with congestion.
NGATS will also benefit from improved data. These improvements include automation, standardization, reduced latency, and improved reliability and accuracy. Aviation weather data are fundamental and crucial to the NGATS. Historically, air traffic management has sometimes segregated the decision making process associated with aviation weather data. The NGATS architecture, on the other hand, likely will integrate these data at a more fundamental level.
Here we present ProbTFM, an automated, traffic flow management (TFM) experimental platform and prototype demonstration which integrates automated meteorological forecasts and observations with automated traffic and system information. ProbTFM is a real-time system with an automated "machine-to-machine" architecture that incorporates aviation weather forecasts and observations at the fundamental, problem formulation, level.
The ProbTFM experimental platform is available to researchers in both the air traffic control / management, and aviation weather communities. Its modular architecture allows researchers to develop and evaluate their own data and algorithms, for application to the difficult problem of air traffic control and management in congested environments.
Dealing with broadly congested scenarios is a difficult problem. During the daytime hours many thousands of flights are departing or in the air at any one time. They are flying between hundreds of different airports and are using hundreds of different aircraft types and equipment. Because demand sometimes exceeds airspace or surface capacity, flight delays are inevitable. This makes for a high-dimensional nonlinear traffic planning problem.
The problem is complicated by several additional factors. First, there is significant uncertainty in the takeoff times for pending departures and in the meteorological forecasts. Therefore, both demand and capacity predictions are uncertain. Also, key aspects of the problem are difficult to generalize. Weather patterns may be categorized, but only to a limited extent. Even patterns that appear similar often have important differences. Likewise, while traffic patterns may be categorized into a set of traffic flows, many flights do not easily fit within such an abstraction.
Furthermore, modeling the NAS is complicated because it is a highly adaptive system, with thousands of human operators in the loop. And there are crucial factors and constraints that may be difficult to quantify. These include procedural and operational constraints, limited equipage, and issues of economic feasibility and equity.
For these reasons traffic flow management in the presence of uncertainties is a difficult problem and a topic of current research. One challenge in probabilistic traffic flow management research is constructing the simulation platform required to evaluate meteorological data and to investigate candidate traffic flow management algorithms and concepts. A NAS-wide simulation with approximately a day of traffic is required to evaluate new concepts. The platform needs to simulate the trajectories of individual flights with realistic traffic schedules, model airspace and airport capacities, represent the various system uncertainties, and model realistic meteorological data. Also, a large number of experiments must be performed to better understand this complicated problem, so the simulation run time should not be excessive. The right level of modeling fidelity is needed to capture critical aspects without over modeling the problem.
ProbTFM can evaluate a wide variety of traffic flow management concepts using real-time system and environmental data. The data may be historical or current. Advanced traffic flow management concepts are often developed and tested in a non realtime environment. But as the concept matures there are several advantages to real-time testing. For instance, a real-time evaluation platform forces timing issues into the forefront. These issues are easily ignored in off-line testing where the input data are simply used when the algorithm is ready.
The use of real-time input data also forces robustness issues into the forefront because the input data cannot be manually checked over. It is easy in the development phase to use input data that have been examined and perhaps edited or otherwise cleaned prior to use in the algorithm. With a real-time evaluation platform the raw input data must be consumed and any checking, editing or filtering must be done automatically, in real-time. Our real-time evaluation platform also provides the advantages that it saves costly resources for developers, and it provides a common, independently developed, platform from which to evaluate the various traffic flow management concepts under development.
ProbTFM also tests concepts in hypothetical future scenarios with, for example escalated traffic demand levels. The platform architecture accommodates future traffic schedules and trajectories, airspace and surface geometries and capacities, historical or current weather data and forecasts, user preferences, and tactical and strategic air traffic control decision making.
In this paper we describe our real-time evaluation platform. We show example scenarios with automated meteorological forecasts and observations, used in our probabilistic traffic flow management decision support algorithm.