Deepak Kulkarni

Modeling Weather Impact on Ground Delay Programs

Scheduled arriving aircraft demand may exceed airport arrival capacity when there is abnormal weather at an airport. In such situations, Federal Aviation Administration (FAA) institutes ground-delay programs (GDP) to delay flights before they depart from their originating airports. Efficient GDP planning depends on the accuracy of prediction of airport capacity and demand in the presence of uncertainties in weather forecast. This paper presents a study of the impact of dynamic airport surface weather on GDPs. Using the National Traffic Management Log, effect of weather conditions on the characteristics of GDP events at selected busy airports is investigated. Two machine learning methods are used to generate models that map the airport operational conditions and weather information to issued GDP parameters and results of validation tests are described.

Data mining for understanding and improving decision-making affecting ground delay programs

Difficulty of deciding on control action depends on the weather and traffic conditions. Weather signature on different days can categorize days into days with little decision difficulty, days with moderate decision difficulty and days with high decision difficulty. This paper examines performance of different data mining methods in the three regions of decision difficulty. Not surprisingly, data mining methods have the best performance in the region of little decision difficulty and have the poorest performance in the region of most decision difficulty. In applications where data mining methods have differing performance in differing regions, it would be more useful to characterize the region specific performance instead of characterizing performance by a single parameter.

Analysis of Airport Ground Delay Program Decisions Using Data Mining Techniques

Air traffic service providers have to make decision s regarding changes to air traffic flow in the event of major weather dist urbances and traffic congestions to maintain safety of the system. The b ehavior of the air traffic management system will be more predictable if consistent decisions are made under similar traffic and weather conditions. Consi stency of deciding on control action depends on the weather and traffic c onditions as well as accuracy in predicting these conditions. Weather parameters (defined in terms of forecast and actual weather and traffic co nditions) on different days can be used to categorize these into days with litt le decision consistency, days with moderate decision consistency and days with high decision consistency. Five years of traffic, weather and ground delay pro gram decisions data at major airports in the United States are used in the analysis. This paper examines performance of different data mining methods in the three regions of decision consistency. Not surprisingly, data mi ning methods have the best performance in the region of most decision consiste ncy and have the poorest performance in the region of little decision consis tency. In applications where data mining methods have differing performance in differing regions, it would be more useful to characterize region specifi c performance instead of characterizing performance by a single parameter. Finally, the results show no significant variation in the performance of diff erent data mining methods for this particular problem. The fact that differe nt mining methods show no significant variation also provides further confide nce in the results of data mining methods. This paper also discusses how prediction errors impact regions of decision consistency.

Impact of uncertainty on the prediction of airspace complexity of congested sectors

The ability of traffic controllers to separate aircraft determines the capacity of the region of airspace under their control, referred to as a sector. Complexity metrics, specifically dynamic density, is used as an estimate for controller workload. The prediction of dynamic density is required for the development of efficient long-term air traffic plans. This paper explores the influence of trajectory errors on the prediction of dynamic density and uses a worst-case analysis to describe the conditions under which forecast errors may lead to excessive complexity. Although the approach has general applicability, it is described using one definition of complexity. Depending on the sector and the complexity function, when a sector is highly congested, the method identifies aircraft entering the sector at certain locations, boundaries and altitudes, whose errors in prediction contribute significantly to the increase in workload. If these errors cannot be reduced, it may be necessary to limit the traffic approaching the sector from these altitudes and boundaries.

Aviation Data Integration System

During the analysis of flight data and safety reports done in ASAP and FOQA programs, airline personnel are not able to assess relevant aviation data for a variety of reasons. This report discusses the Aviation Data Integration System (ADIS), a software system that provides integrated heterogeneous data to support safety analysis. Types of data available on ADIS include weather, D-ATIS, RVR, radar data, and Jeppesen charts, and flight data. Three versions of ADIS were developed to support airlines. The first version has been developed to support ASAP teams. A second version supports FOQA teams, and it integrates aviation data with flight data while keeping identification information inaccessible. Finally, a prototype was developed to demonstrate the integration of aviation data into flight data analysis programs. The initial feedback from airlines is that ADIS is very useful in FOQA and ASAP analysis.

Data Analysis using Scale-space filtering and Bayesian Probabilistic Reasoning

Abstract This paper describes a program for the analysis of output curves from a differential thermal analyzer (DTA). The program first extracts probabilistic qualitative features from a DTA curve of a soil sample, and then uses Bayesian probabilistic reasoning to infer what minerals are present in the soil. It consists of a qualifier module and a classifier module. The qualifier employs a simple and efficient extension of scale-space filtering DTA data. Ordinarily when filtering operations are not highly accurate, points can vanish from contours in the scale-space image. To handle the problem of vanishing points, our algorithm uses perceptual organization heuristics to group the points into lines. It then groups these lines into contours by using additional heuristics. Probabilities are associated with these contours using domain-specific correlations. A Bayes tree classifier processes probabilistic features to infer the presence of different minerals in the soil. We show experimentally that using domain-specific correlations to infer qualitative features, this algorithm outperforms a domain-independent algorithm that does not.

Using an Automated Air Traffic Simulation Capability for a Parametric Study in Traffic Flow Management [STUB]

Flight delays occur when demand for capacity-constrained airspace or airports exceeds predicted capacity. Demand for capacity-constrained airspace or airports can be controlled by a series of Traffic Management Initiatives (TMIs), which use departure and airborne delays, as well as pre-departure and airborne reroutes, to manage access to the constrained resources. Two systems exist in current and planned future operations to address imbalances between demand and capacity. The Collaborative Trajectory Options Program (CTOP) reduces demand to constrained resources by assigning strategic departure delay and predeparture reroutes. Reroutes are selected from Trajectory Options Sets (TOSs) submitted by airlines. As flights approach the constrained resource, the Time-Based Flow Management System (TBFM) is used to assign tactical delay to satisfy constraints. This paper describes experiments performed to study the impact of varying levels of airline participation in CTOP via submission of TOSs on ground delay and flight time, and the impact of departure uncertainty on TBFM delays. Results suggest that as CTOP participation increases, average ground delays decrease for all airlines, but to the greatest extent for airlines participating in CTOP. A threshold in CTOP participation, which varies with the constraint capacity, is identified beyond which there is relatively little further reduction in average ground delays. Similarly, given the likely level of CTOP participation, the capacity reduction for which CTOP would be an appropriate TMI is also identified. Results also suggest that high average departure errors and high variability in departure error can make the prioritization of TBFM internal departures in TBFM metering and scheduling infeasible. Departure errors at current levels are, however, acceptable.

Agent Architecture for Aviation Data Integration System

This paper describes the proposed agent-based architecture of the Aviation Data Integration System (ADIS). ADIS is a software system that provides integrated heterogeneous data to support aviation problem-solving activities. Examples of aviation problem-solving activities include engineering troubleshooting, incident and accident investigation, routine flight operations monitoring, safety assessment, maintenance procedure debugging, and training assessment. A wide variety of information is typically referenced when engaging in these activities. Some of this information includes flight recorder data, Automatic Terminal Information Service (ATIS) reports, Jeppesen charts, weather data, air traffic control information, safety reports, and runway visual range data. Such wide-ranging information cannot be found in any single unified information source. Therefore, this information must be actively collected, assembled, and presented in a manner that supports the users problem-solving activities. This information integration task is non-trivial and presents a variety of technical challenges. ADIS has been developed to do this task and it permits integration of weather, RVR, radar data, and Jeppesen charts with flight data. ADIS has been implemented and used by several airlines FOQA teams. The initial feedback from airlines is that such a system is very useful in FOQA analysis. Based on the feedback from the initial deployment, we are developing a new version of the system that would make further progress in achieving following goals of our project.

Models of maximum flows in airspace sectors in the presence of multiple constraints

Recently, the ATM community has made important progress in collaborative trajectory management through the introduction of a new FAA traffic management initiative called a Collaborative Trajectory Options Program (CTOP). FAA can use CTOPs to manage air traffic under multiple constraints (manifested as flow constrained areas or FCAs) in the system, and it allows flight operators to indicate their preferences for routing and delay options. CTOPs also permit better management of the overall trajectory of flights by considering both routing and departure delay options simultaneously. However, adoption of CTOPs in airspace has been hampered by many factors that include challenges in how to identify constrained areas and how to set rates for the FCAs. Decision support tools (DST) providing assistance would be particularly helpful in effective use of CTOPs. Such tools would need models of demand and capacity in the presence of multiple constraints. This study examines different approaches to using historical data to create and validate models of aircraft counts in sectors and other airspace regions in the presence of multiple constraints. A challenge in creating an empirical model of aircraft counts under multiple constraints is a lack of sufficient historical data that captures diverse situations involving combinations of multiple constraints especially those with severe weather. The approach taken here to deal with this is two-fold. First, we create a generalized sector model encompassing multiple sectors rather than individual sectors in order to increase the amount of data used for creating the model by an order of magnitude. Secondly, we decompose the problem so that the amount of data needed is reduced. This involves creating a baseline demand model plus a separate weather constrained sector count reduction model and then composing these into a single integrated model. A nominal demand model is a sector aircraft count model (gdem) in the presence of clear local weather. This defines the flow as a function of weather constraints in neighboring regions, airport constraints and weather in locations that can cause re-routes to a location of interest. A weather constrained flow reduction model (fwx-red) is a model of reduction in baseline counts as a function of local weather. Because the number of independent variables associated with each of the two decomposed models is smaller than that with a single model, need for amount of data is reduced. Finally, a composite model that combines these two can be represented as fwx-red (gdem(e), l) where e represents non-local constraints and l represents local weather. The approaches studied to developing these models are divided into three categories: (1) Point estimation models (2) Empirical models (3) Theoretical models. Errors in predictions of these different types of models have been estimated. In situations when there is abundant data, point estimation models tend to be very accurate. Also, empirical models do better than theoretical models when there is sufficient data available. The biggest benefit of theoretical models is their general applicability in wider range situations once the degree of accuracy of these has been established. Quantile regression methods are used to create models of different quantiles of aircraft counts as well as probability distribution functions. Such models can be used in CTOP DSTs in providing assistance with recommendations about CTOP parameters and in supporting what-if reasoning about consequences of potential decisions.

Compressing Aviation Data in XML Format

Design, operations and maintenance activities in aviation involve analysis of variety of aviation data. This data is typically in disparate formats making it difficult to use with different software packages. Use of a self-describing and extensible standard called XML provides a solution to this interoperability problem. While self-describing nature of XML makes it easy to reuse, it also increases the size of data significantly. A natural solution to the problem is to compress the data using suitable algorithm and transfer it in the compressed form. We found that XML-specific compressors such as Xmill and XMLPPM generally outperform traditional compressors. However, optimal use of Xmill requires of discovery of optimal options to use while running Xmill. Manual discovery of optimal setting can require an engineer to experiment for weeks. We have devised an XML compression advisory tool that can analyze sample data files and recommend what compression tool would work the best for this data and what are the optimal settings to be used with a XML compression tool.