• Summary
• Objective
• Background
• Sample JUP Research Topics
  • Massachusetts Institute of Technology
    • Runway Operations Planning and Control
    • Investigating Information Requirements in the Future NAS Information Architecture
    • Issues of Representing Aircraft Intent Information for Separation Analysi
    • Fast-time Simulation Modeling of Runway Incursions
  • Ohio University
    • Voice Recognition for CPDLC Transmissions
    • GPS Receiver Block Processing
    • GPS Attitude Determination of a UAV Using Ultra-short Baselines
    • Fading Multipath-Induced Range Biases in GPS Receiver Tracking Loops
  • Princeton University
    • Robust Flight Control Systems
    • Neural Network Flight Control Systems
    • Intelligent Aircraft/Airspace System
    • Coordinated Flight of Uninhabited Air Vehicles
• Footnotes

The FAA/NASA Joint University Program for Air Transportation Research (JUP) is a research partnership of three universities, which conducts scientific and engineering research on problems of a long term nature related to the ultimate improvement and development of the National Airspace System. This includes Massachusetts Institute of Technology, Ohio University, and Princeton University. JUP research covers a broad scope of technical disciplines which contribute to civil aviation, including but not limited to air traffic control theory, human factors, satellite navigation and communications, aircraft flight dynamics, avionics and meteorological hazards. The universities seek validation and suggestions for improvement of their research, as well as proposed new avenues for investigation via a series of quarterly reviews. These reviews are conducted at the campuses of the participating academic institutions, as well as at the sites of the governmental sponsors, the Federal Aviation Administration (FAA) and the National Aeronautics and Space Administration (NASA).

NASA and FAA are able to stretch available research funds to achieve high-priority goals by leveraging each other's support as well as by implicitly benefiting from the research that the principal investigators and their colleagues conduct on related projects. University researchers provide valuable feedback to NASA and FAA technologists regarding the goals and effectiveness of government programs. This leveraging of resources is exemplified best in the many instances where the JUP format and research it initiates has attracted additional agency sponsors for grants on narrower research topics. Following their example, agency program managers unfamiliar with the program are encouraged to participate and benefit as well.

Back to top

The FAA/NASA Joint University Program (JUP) is a long-term cooperative partnership between FAA and NASA to pursue common research goals by promoting research and education in selected aviation technologies. The program is dedicated to the principle that solutions to large-scale system problems in the National Airspace System (NAS) come only after the technological foundations have been laid through long-term basic and applied research. These concepts are directly built into the JUP's guiding Memorandum of Agreement [1], which states:

"The objective of this program is to provide access to, and dissemination of the long range, innovative research in civil aeronautics related technologies under development at American colleges and universities. A secondary benefit is the creation of a talented pool of graduates trained in engineering and scientific disciplines to advance the state-of-the-art in aviation technology in government, industry, and academic community..."

Back to top

Initiated in 1971, JUP was initially a NASA program, managed by the NASA Langley Research Center. FAA elected to participate in 1979, with FAA sponsorship and participation centered at FAA Headquarters through 1985, and at the FAA William J. Hughes Technical Center thereafter. An administrative framework consisting of two successive five year memoranda of understanding and interagency agreements between FAA and NASA, was employed to govern the conduct of the JUP from 1985 through 1995. Under these agreements FAA provided approximately 1/2 of the program financial support to NASA Langley for the program. By combining its funding with the FAA's, NASA Langley was able to award three grants annually to the participating universities for sponsorship of their research. These grants were NGL-022-009-640 to the Massachusetts Institute of Technology, NGR-36-009-017 to Ohio University, and NGL-31-001-252 to Princeton University. In 1995, the FAA assumed responsibility for awarding these grants, which became FAA grants 95-G-017 to M.I.T., 95-G-014 to Ohio and 95-G-011 to Princeton University. Funding was included in these grants to provide for hosting of participation in a series of quarterly review conferences, with one at each of the universities and one of the government sponsors. In 1996 the NASA Ames Research Laboratory replaced Langley as the sponsoring NASA center.

JUP provides an interdisciplinary team approach to research and education in aviation technologies. By bringing this multi-agency, multi-university approach to bear on large scale NAS management and technical problems, highly original and creative solutions emerge. An additional benefit is the creation of a talented cadre of engineers and scientists who presently form a core of advanced aeronautical expertise in industry, academia, and Government.

Back to top

This section summarizes selected topics currently being researched under the program. Listed simply in alphabetical order by university, each university's Principal Investigator followed by the current research project titles, graduate student researchers, and supplied abstracts of their work are presented [2].

Back to top

Back to top

It is a common observation that significant delays occur during departure operations at major European and US airports. The resulting environmental impact and economic inefficiencies generate a growing need for the reduction of such delays. There are ongoing research projects at the Institute of Flight Guidance, German Aerospace Research Establishment (DLR) and the Massachusetts Institute of Technology (MIT) towards the development of automated decision-aiding systems to assist air traffic controllers in handling departure traffic and mitigating the adverse effects of ground congestion and delays. After considering the operational context for such systems, the main outcomes of research work up to date are reported. A conceptual architecture with planning and control layers is described together with a detailed formulation of the runway operations planning problem. An algorithmic approach for planning aircraft pushback and takeoff release times is presented. During this development effort, system inherent uncertainty and incompleteness of information is always taken into account in studying the dynamics of the planning process.

Back to top

With the recent advances in technology, namely datalink and the World-Wide Web, there is an opportunity for the National Airspace System (NAS) to extend information-sharing between the ground and the air. We use the FAA-based "2005 Concept of Operations" as a central document for the discussion of the information requirements for this future NAS information architecture. A possible object-based information structure is proposed centering on the individual flight, schedule, and environment. Within these objects, there are both static and dynamic components, e.g., tail number and current status, respectively, that make up the Flight Information Object. The Flight Information Object (FIO) is a database of pertinent information about a flight available to all interested users and service providers across the NAS. We discuss the possible content of this FIO as well as who may be considered users. To determine the information content of this FIO, we take a human-centered approach and probe the users themselves for the desired information requirements through a survey. This survey queries what information the user would like to see in a future FIO as well as if the user is interested in issues of intent. It also probes whether this information-sharing would affect user performance. Currently, we are distributing the survey to a small group of pilots, air traffic controllers, and dispatchers to determine survey suitability. Eventually, we would like to distribute the survey across the Web through sites like AVweb, ATC webring, and the Airline Dispatchers Federation.

Back to top

Many proposals to increase airspace capacity involve reductions in the separation requirements that controllers must apply between aircraft. Before such proposals can be implemented safely, methods of modeling the effect of the procedural or technical changes are essential. One of the major factors affecting the ability to keep aircraft safely separated is surveillance, and particularly the quality and frequency of update of aircraft state information made available to the controllers. A surveillance state vector modeling approach is developed containing surveilled position, velocity, acceleration and intent state information at a given time. The uncertainty and errors that exist in these 'surveilled' states relative to the 'true' values are also represented. The intent state is of primary importance when estimating aircraft separation into the future. A notation is developed for the various types of intent which can exist, representing the future behavior of the aircraft from the viewpoint of the controller, pilot or aircraft systems. Separation safety issues can arise when any of these types of intent are incompatible with each other. Examples of levels of intent are provided depending on the timeframe of interest (current intent vs. future intent vs. higher level goals). The current intent state is particularly important as it may enable conformance monitoring (a key controller task) to be modeled, i.e. how well the pilot is adhering to the cleared path. This may in turn provide a basis for hypothesis testing of whether an aircraft is 'on' or 'off' the controller's intended path in order to maintain proper separation with other aircraft under their control.

Back to top

Runway incursion incidents have been increasing steadily over the last 5 years, and are one of the top safety concerns for the FAA. This, coupled with the importance of human performance and surveillance technology level, made it a suitable test problem for a fast-time simulation study. Flight crew and controller behavior modules are being developed and integrated with ATAC's Simmod-Pro! Model. The initial modeling effort has concentrated on incursions caused by taxiing aircraft which violate hold short instructions when crossing an active runway, thus interfering with aircraft landing or taking off. Transgressions are injected into the simulation to model these violations, which are detected via surveillance logic in the human modules. The human performance is modeled by defining probability density functions representing the time delays introduced by the surveillance, detection and response functions for each of the human and communication elements. Different curves can be defined reflecting the different technology levels which may be available to the different users (e.g. representing AMASS or LVLASO surveillance technologies). The pilot and controller response logic dictates subsequent behavior. The model outputs closest point of approach data for the aircraft involved in the incursion. This enables incident severity, human intervention and technology impact analyses to be undertaken. Preliminary results from the first phase of incursion modeling at LAX airport are presented.

Back to top

Back to top

C-Cast (Controller's - Communication and situational Awareness Terminal) uses voice recognition to translate verbal messages into text for Controller/ Pilot Data Link Communications (CPDLC). The message is displayed on a monitor along with a map of the runway and its traffic. This serves as a back up for the pilot, reducing any miscommunication with the controller. The pilot is also able to see the other planes on the runway and their communications with the controllers. By communicating routine clearance and frequency information on the data link, this system will reduce congestion on a given channel and will free the voice channel for critical messages. There are, however, some issues that need to be addressed using voice recognition technology. The recognizer must be trained to understand an individual's voice. This can become a long process since the user must pronounce every possible word and then combine those words into valid command statements. Non-standard phraseology can be a problem along with an infinite number of different aircraft call signs. C-Cast is scheduled for installation at Dallas Fort Worth in March of 2000 and testing is to begin in early July. The TCP/IP interface is still under development along with the serial interface with the VDL Mode 2 transmitters. Map and frequency specifications also must be updated for the Dallas Fort Worth installation.

Back to top

Another GPS-based landing issue is signal quality monitoring. When a GPS receiver reports a bad measurement, it is often difficult to analyze the anomaly due to a lack of detailed data surrounding the event. A system that samples the GPS radio frequency signal at a rate that preserves all the pertinent signal information allows for off-line processing to determine the cause of the anomaly. The GPS Anomalous Event Monitor (GAEM) is designed to capture GPS data at a rate of 5 mega samples per second for several seconds surrounding an event. New block processing techniques are used on the 12-bit samples to assess the quality of the received GPS signal. Block processing provides estimates of the GPS signal parameters without the use of tracking loops. The sampled data are processed in 1-millisecond blocks. Code-phase, frequency, carrier-phase, and signal-to-noise ratio are estimated for each block of data. It has been implemented and tested in a software GPS receiver, and the results are consistent with a NovAtel GPS receiver. Further, the satellite signal quality is analyzed in terms of carrier-to-noise ratio, shape of the correlation function, false acquisition peaks, code-carrier divergence, sudden change in code and/or carrier, and change in the satellite code. Based on the statistical distributions of the detection parameters, thresholds can be calculated to provide a specified probability of a false detection, and to calculate a minimum detectable bias with a specified probability of missed detection. A detection statistic is introduced for the shape of the correlation function.

Back to top

A navigation system designed specifically for use in an unmanned aerial vehicle (UAV) is described. The UAV platform applies many constraints on the design of a navigation system. Particularly, the size of an unmanned aerial vehicle directly affects the achievable accuracy in an attitude determination system because the antenna separation is confined to the maximum dimension of the vehicle. With a brief background on the use of GPS carrier phase measurements, the results from a recent preflight data collection are presented and analyzed. Specifically, a heading accuracy of approximately one degree is obtained by an attitude determination system with antenna separation distances (baselines) near half of an L1 wavelength, or 96mm.

Back to top

Over the next generation it is envisioned that satellite-based communication, navigation, landing and surveillance systems will largely displace the current ground-based systems. GPS-based landing systems, for example, hold the promise of service to all runway ends at a given airport with a single ground installation. There are obstacles, however, which must be addressed. Multipath is the largest source of error in high accuracy GPS applications. A GPS receiver employing a noncoherent delay lock loop (NCDLL) in its code tracking loop has been shown, theoretically, to yield a range bias under certain multipath conditions. When the multipath signal has a large relative Doppler shift, with respect to the code tracking loop bandwidth, the NCDLL based receiver will produce a biased code phase estimate. In contrast, a coherent delay lock loop (CDLL) will tend toward an unbiased estimate. A bench test has been constructed using two GPS single channel, C/A code, L1 only emulators. One is triggered later than the other and attenuated as well as frequency-shifted such that the combined signal appears to be one corrupted with a single Doppler-shifted specular multipath reflection. A NovAtel GPSCard receiver using an NCDLL architecture makes pseudorange and carrier phase measurements on the combined RF signal. In post-processing, the bias caused by the multipath is isolated and compared with theoretical and previously simulated (in software) results. The actual worst case errors were on the order of a few meters, or about half of what theory predicted. This difference has been attributed to a hybridization of loop structures, meaning that the NCDLL has operation characteristics that are more indicative of a CDLL architecture, resulting in a bias that is in between the expected performance for each loop alone.

Later Results

Multipath is now the foremost obstacle in moving towards higher accuracy applications of the Global Positioning System such as aircraft precision approach and landing. An abundance of past research has focused on the theoretical and practical effects of slowly varying multipath but little has been done to explore the behavior of real receivers under conditions of quickly changing multipath. The purpose of this work is to provide insight into the range errors produced by multipath of relative Doppler shift greater than the tracking loop bandwidth in a GPS receiver. In previous theory, a substantial advantage is expected with respect to multipath when moving to a coherent delay lock loop (CDLL) receiver from a noncoherent delay lock loop (NCDLL). The results obtained by taking measurements from a modified NovAtel Beeline GPS receiver showed that the range errors produced by each loop type are actually quite similar. After designing a higher fidelity model to take carrier reference distortion and discriminator normalization into account, it is indeed seen that the prediction of improved multipath performance in the CDLL was a result of oversimplification of the model. Future research in the area includes finding optimal discriminator algorithms for multipath mitigation and working with a software radio for receiver parameter studies.

Back to top

Back to top

We are developing a stochastic approach to designing practical flight control systems, which provide satisfactory stability and performance in the presence of parametric uncertainties. Solving this problem is important for the design of future transport aircraft because they will rely on fly-by-wire systems for safe, efficient control. Stochastic robustness of the control system is defined by the probabilities of violating system design requirements, and a numerical search is used to minimize these probabilities. The design metrics may include flying qualities goals, limitations on control deflection and rate, structural limits, and ride quality targets.

The robust control design problem is "Non-Polynomial (NP)-Complete," that is, the number of computations that are required grows exponentially with the number of uncertain parameters and conditions. This computational complexity is inherent in the problem; it can be avoided only by finding precise solutions for simplified dynamic models or by finding approximate solutions to high-fidelity models. We take the latter approach, using Monte Carlo simulation to estimate the probabilities of violating design goals and Genetic Algorithms to optimize control system parameters. The number of calculations is a polynomial function of the number of uncertain parameters; hence, the computations required for control system design are tractable.

Unlike previous attempts to design robust flight control systems, our approach is not limited to linear dynamic models and linear control system structures. We are currently exploring the direct design of nonlinear flight control logic, in which our numerical methods are applied to a flight controller that inverts the aircraft's nonlinear equations of motion. A substantial benefit of this approach is that the final design can account for all significant nonlinearities in the aircraft's high-fidelity simulation model. Because evaluation of this model is an inherent feature of the design process, limiters, delays, aerodynamic phenomena like separation and stall, and even aeroelastic effects can be considered in the design process.

Back to top

The need for improved aircraft safety and performance, driven by both cost and operational requirements, is placing increasingly complex demands on practical flight control systems. These systems operate in noticeably nonlinear regimes and must be capable of adapting to changing flight conditions while still preserving stability and satisfactory performance. The ability of neural networks to approximate unknown nonlinear mappings with high-dimensional input spaces and their potential for on-line learning make them excellent candidates for use in full-envelope flight control systems. To address the aforementioned requirements, a global nonlinear controller containing neural networks as subsystems has been created. Its basic structure is motivated by the proportional-integral (PI) controller, a well-known linear control structure that enjoys widespread usage in many practical applications.

Our approach consists of replacing the linear gains of a PI flight controller with nonlinear, structured neural networks. A forward neural network produces desired trim control settings and responds to pilot commands. A feedback neural network provides stability augmentation, rejects disturbance effects, and minimizes the effects of changes in light condition and aircraft configuration. A command-integral neural network enhances low-frequency disturbance rejection, provides precise steady-state response to commands, and further reduces the effects of parameter variation. Thus, the neural network flight controller is motivated by well-established classical control structures while adding the broader capabilities of nonlinear control.

The neural network design approach has been developed with flight control certification issues in mind. The networks are pre-trained to mimic a satisfactory gain-scheduled controller. Thus, the control system's small-amplitude response at flight conditions throughout the flight envelope is identical to the response of a gain-scheduled controller. Subsequently, the networks' parameters can be fine-tuned to improve system performance, either on-line, for fully adaptive operation, or off-line, for adaptation between flights or at times of regularly scheduled maintenance. Operation of the neural network flight controller in all of these modes is completely transparent, allowing certification engineers to assess the controller's operation in both normal and emergency flight conditions.

Back to top

Inter-airport flights are the result of the actions of many different people and organizations, operating in different fashions and over different time scales. Each organization can be regarded as an agent in the aircraft/airspace system, as each can change some facet of the air traffic process. The arrangement of the agents is broadly hierarchical in nature. Aircraft operate within the constraints on flight path set by ground controllers. Airlines can schedule flights only if they have the appropriate slots and gates. An air traffic controller works within the ground traffic control framework set by the FAA, and implements the restrictions imposed by flow control.

The Aircraft/Airspace System can be viewed as a collection of Intelligent Agents, each with differing beliefs and interests. When two agents are involved in the process of making a decision, the process can be represented as negotiation, with each agent having different interests. Future advances in air transportation technology will increase the overlap in capabilities of the agents, increasing the need for negotiation. Principled Negotiation is being pursued as a means of improving the decision-making process.

This research is developing means by which the negotiation paradigm can be used to create an intelligent aircraft/airspace system. It will provide a framework within which both air- and ground-based systems will cooperate to increase the levels of safety, reliability, and robustness consistent with the technologies that will exist in future decades.

Back to top

A new system architecture for the coordinated flight of uninhabited air vehicles is being developed. The system architecture takes advantage of and builds upon prior projects sponsored by the JUP, including concepts for an AUTOCREW, intelligent aircraft/airspace systems, Bayesian belief networks and expert systems, and hierarchical flight control. This research subsumes formation flight of swarms of uninhabited air vehicles (UAVs) but goes far beyond the notion of close-formation flying. The principal focus is on the guidance, navigation, control, and communications requirements of UAV teams that are tasked with one or more joint mission goals. In the process, the individual vehicles within a team can be expected to fly on dissimilar paths, to provide support for each other as needed to achieve primary goals, to sense and evaluate changing scenarios, situations, and environments, and to automatically re-plan missions or interchange leader-follower roles when damage or failure occurs. The project will be pursued through flight testing.

The experimental UAVs will exhibit intelligent behavior in several ways:

• They are fully autonomous, subject to safety backup systems. Expert systems organize the functions of route planning and revision, path optimization, traffic alert and avoidance, ground proximity warning, communication, and system monitoring and control.
• They are failure-tolerant. Physical and analytical redundancy provide increased reliability, maintainability, and survivability. In addition to navigation, guidance, and control, the on-board computers execute logic for system failure detection, identification, and reconfiguration.
• They learn complex control behavior. UAV systems may learn performance-improving features, such as adapted control gains for reduced tracking error and increased disturbance rejection, and they will learn to perform specific tasks and maneuvers, such as aerobatic routines, landing in a crosswind, delivering a payload, or retrieving an object.

Back to top

[1] Memorandum of Agreement between FAA and NASA Concerning Joint University Research in Air Transportation, FNA/05-97-02, 9/25/97.

[2] The projects listed here are not necessarily solely funded by JUP, which has a very modest budget, but quite often the projects are initiated by the program and continue under other agency or industry sponsored grants.

Back to top