Basic Research for Air Force Network Systems and Communications
The Air Force is becoming a network-enabled paradigm, wherein many of its capabilities will be generated through, and dependent on, the integrated efforts of multiple components. This approach to operations is expected to result in greater agility and attendant tactical advantages. However, as is the case with any untested concept, there is a need for technology that enables the analysis and execution of the new paradigm. In this case, the scope of change is extremely large, requiring reanalysis of force structures, doctrine, acquisition options, command-and-control systems, training, and long-range planning, not to mention the considerable challenges involved in engineering, constructing, and managing the actual networks. Much of the planning and analysis will depend on the research described in this chapter, which will provide the necessary conceptual and technical foundation. Network-centric warfare also is critically dependent on software, the subject of Chapter 4, on the effective distribution and management of information throughout the network, which is the prime topic of Chapter 5, and on the effective employment of that information, which is considered in Chapter 6.
TYPES AND CHARACTERISTICS OF COMMUNICATION AND NETWORK SERVICES NEEDED IN THE FUTURE
Figure 3-1 shows the range of components that will be networked in the future Air Force. The following capabilities are key to achieving the desired functionality, but off-the-shelf technologies are not yet adequate:
The network must provide robust data and circuit services to tens or hundreds of thousands of fixed and mobile users with different service levels. Some of the service challenges include guaranteed rates, communication over difficult channels, hard time-deadlines, reliable message delivery over unreliable networks, security, and policy-driven resource allocation (explained below).
Sensors in the networks must be able to relay data while maintaining coherency, and there must be high compression of correlated sensors to reduce communication needs.
Laser communication must be available between satellites and aircraft and to terrestrial sites.
Satellite services must be available to mobile and fixed users. Those services must (1) maintain connectivity at all times and track important users, (2) have low probability of detection (LPD), low probability of intercept (LPI), and antijamming (AJ) capabilities, (3) be secure, and (4) support very small as well as large terminals.
Wireless radio services must be available in the field without pre-existing infrastructures, and they must be secure, power-efficient, have LPD and LPI, have AJ capabilities, and be able to maintain connectivity with important users.
The satellite, radio, and wireless networks must be integrated with a high-bandwidth, affordable, and secure terrestrial fiber network as one supernetwork.
The network must be an integrated one with multilevel security, data delivery within deadlines, and the capability for rapid reconfiguration and adaptation.
With the possible exception of power efficiency and some aspects of security, all of these characteristics are primarily driven by defense needs, and they will not be advanced by the commercial sector.
The remainder of this chapter describes some of the technical challenges of future communication modalities planned for the Air Force and the performance of networks when they are connected together in military settings. It concludes with recommendations of particularly important basic research areas that typically cut across multiple challenges.
TECHNICAL CHALLENGES POSED BY FUTURE AIR FORCE NETWORKS AND COMMUNICATIONS SYSTEMS
Challenges for Future Air Force Communications Systems
Defense applications have to contend with communication modalities that are not encountered in commercial and civilian settings. Satellite
channels have unusually long-delay data rates (sometimes known as delay-bandwidth products) and randomly fading dispersive channel characteristics. Radio channels, especially those associated with mobile platforms, have rapidly changing link capacities and connectivities, with disconnections or dropouts that can be on the order of minutes or more. In contrast to this dynamism, traditional layer 3 (Network Layer) and layer 4 (Transport Layer) protocols assume fairly stable underlying substrates that change, if at all, over the course of minutes—that is, much more slowly than most transmissions. Recent applications have shown these traditional protocols often yield low throughputs and poor quality of service when applied to DOD systems. In some cases, these protocols do not work at all despite valiant efforts to provide patches. Thus, the main challenge of Air Force communications is to provide assured connectivity between networks (albeit at varying rates) under difficult channel conditions, including during adversarial attacks.
A cost-effective investment strategy is to make maximum use of commercial technologies. However, not all the services listed above can be supported by commercial technologies. In many cases, commercial architectures only need to be modified, but in some cases completely new designs are needed.
Source Compression for Correlated Sensors
One communications challenge is the need for source compression methods when multiple sensors collect correlated observations, whether or not this was done in a coordinated manner. This is critical in order to fit the large volumes of sensed data in modest network capacities, especially in difficult communication environments. Particularly challenging is the identification and compression of highly correlated data in the absence of fine coordination among sensing platforms. This will require both lossless and information-lossy compression and new approaches to measuring signal fidelity tailored to battlefield signals. Another challenge lies in creating joint source-network coding: The separation theorem, which for point-to-point communication channels allows one to decouple source and channel coding, does not apply to networks of Air Force interest, where the data volume itself may cause packet loss due to congestion. It is important to explore the joint consideration of coding for both source compression and network transport (from layer 1 to layer 4).
Design of Communications Systems
A better understanding of capacity limits is needed in order to design the best architectures for Air Force communications systems, which
should perform at the theoretical limit for difficult channels and in the presence of adversarial attacks. The open-air satellite, radio, and optical channels planned for the Air Force exhibit fading, dispersion, and interference (benign and adversarial) not typically encountered in well-known classical channels, especially when the channel is broadband. For these channels, there are many different ways to design the communication system, and good system design will depend on understanding the channel effects and the fundamental limits of communication performance optimized over the class of possible architectures. Our current understanding of the fundamental limits of these channels is only rudimentary, and the space of communication designs has yet to be fully characterized and optimized. There are also open questions related to the capacity limits of multiple-access channels with or without coordination. Multiple-access techniques for open-air channels, such as satellite and battlefield radio channels, are extremely important for good communication performance. Classical techniques must be reconsidered in a context that includes adversarial jamming, interference, coupling, and little or no coordination with some key inputs such as the satellite channel. Classical information theory articulates the limits of communication capacities with no specific attention paid to the transaction delay of the session. In fact, most results are asymptotic, taken as the delay becomes large. In contrast, some critical Air Force applications are very delay sensitive, and a theory for the fundamental capacity limits of a communication system with delay constraints needs to be developed. In some applications power efficiency is also required.
Classical communication and information theories do not incorporate an element of adversarial attacks, although some results related to LPD, LPI, and AJ communications were developed 40 years ago. The technology assumptions underlying those results have changed, of course, and none of the results were obtained with networking in mind. For instance, networks using open-air interfaces present a new set of vulnerabilities that can be easily exploited by adversaries. It is important to understand the set of possible technologies and techniques available to adversaries and to design the network with those in mind rather than treating them as an afterthought. Game-theoretic and optimization techniques will likely be relevant approaches, as will methods for authentication and cryptography.
Reliable and Efficient Free-Space Optical Communications
Free-space optical communications in space is relatively well developed, though its capabilities for terminal spatial tracking systems and the weight and power demands of the systems can be improved using more
sensitive receivers. The current sensitivity of the best receivers (direct or heterodyne detection) available in prototype form requires 2-10 received photons per bit. With the advent of solid-state, single-photon detectors, a photon counting receiver can achieve 5-10 bits per received photon. A fundamental investigation into the best combination of modulation and coding under technology constraints could lead to large payoffs for Air Force space and aircraft systems.
More problematic is free-space optical communication over the turbulent atmospheric channel where fast fading (a few multiseconds) presents difficulties in the design of modulation/coding and receivers. These systems are important for aircraft and ground applications. For supersonic platforms, the bow shock presents even bigger challenges than atmospheric turbulence owing to its speed, which is three orders of magnitude higher. The current mainstream receiver uses adaptive optics to compensate for phase-front distortion, which gives rise to fading and other undesirable channel effects. While its performance is adequate, the solution is costly and heavy. The Air Force should pursue new system techniques, such as spatial-, temporal-, and frequency-diversity receivers in both coherent and incoherent forms, to mitigate the turbulence effect.
The Air Force needs more energy-efficient communications for two reasons. The first is that for operations in a hostile environment, keeping the transmitter power low is good for LPD and LPI. The second is that some operations may be conducted with ground mobile radios, where battery conservation is very important. The first problem can and should be addressed as part of the research outlined above aimed at developing capabilities for efficient communications for difficult channels and under attack. The second includes efficient signal processing and higher-layer network protocols, which will be addressed below in the discussion of sensor networks.
Dynamic Information Theory for Network Applications
Recent developments in the new applications of communication networks often require that information be processed in a distributed fashion in real time. For example, in a sensor network, a relay node often needs to forward a received signal without being able to decode the embedded message reliably; when controlling a dynamic system, decisions often need to be made based on noisy observations. Conventional information theory, which is based on the concept of reliable communications and
thus assumes large delays and extensive coordination, is often not applicable to these new problems. The theoretical understanding of how to handle information without the error-free guarantees is therefore important for future developments of highly dynamic and distributed network applications. The study of communication systems with partial or list decoding using a layered coding structure could be the first concrete step in this direction.
Challenges for Future Air Force Networks
The theory of networks has not matured to a point where one can predict how well the protocols developed heuristically in one application setting will perform on a communication network built on radically different communication modalities. To deal with the new and complicated modalities of importance to the Air Force, fundamental tools must be developed to help understand how networks may perform in new environments and to optimize architectures. It is simply too costly to develop these architectures and protocols ad hoc and then experiment with the communication links in the field. In addition, there are many possible ways to configure a communication system. If a communication system is going to be used as part of a network, it should be designed jointly with the network and not independently. For example, end-to-end reliable data delivery can be a function of the communication system (using diversity receivers and error-correcting codes), or it can be a function of the network (using diversity path routing and automatic repeat requests at various layers such as the Data Link Control Layer and the Transport Layer).
High-Bandwidth, Affordable, and Secure Global Fiber Network
The Air Force needs a global broadband fiber network to interconnect its multiple modalities and act as the high-speed backbone. Current networks, such as the Global Information Grid (GIG), use commercial technologies with wavelength-division multiplexed optical transport and electronic packet switching. This technology is both costly (it does not scale well to very high data rates) and insecure. For terrestrial networks there is a burning need to create economical new optical transport mechanisms such as optical flow switching (dynamically set up with short-duration connections on demand) with lightweight protocols and also a security architecture for both the Physical Layer (the transport mechanisms) and the higher layer protocols.
Network Layer (Routing) Design
Routing over the Internet today is purposely set to change rather slowly to prevent oscillations and overreactions to sudden rate surges. In contrast, most of the channels used by the Air Force can experience rapid changes in connectivity and link capacities due to random channel effects, mobility, and adversarial attacks. Thus, most routing protocols in use today cannot keep up with such changes nor can they provide effective congestion and flow control. To create better protocols it is important first to understand the fundamental effects on the Network Layer of these fast dynamics. For example, mobile networks (especially those that support high data rates and time-deadline services) are not well served by conventional commercial protocols, which are mostly designed for static or quasi-static network topologies. Many of these dynamics appear first in the Physical Layer and permeate up the protocol stack. The mobility aspect calls for the joint design of the Physical Layer and the higher layers, including routing and the Transport Layer. The development of fundamental mathematical tools is a prerequisite to the understanding and solution of this very important problem.
Transport Layer Design
The task of providing error-free, end-to-end delivery of messages in a network is jointly shared by the Data Link Control Layer (layer 2) and the Transport Layer (layer 4). For random channels (especially those in mobile networks), these two layers interact in ways that are stochastic and not yet well understood; the interactions often result in drastically reduced throughput. Researchers have not yet succeeded in predicting the performance of specific implementations and determining the fundamental limits of these protocols when coupled with random channels. The answer might come from control theory and stochastic system analysis.
Current Transport Layer protocols assume a stable Physical Layer communication infrastructure, so packet losses are typically interpreted as buffer overflow at routers due to congestion. In the random channels with which the Air Force has to deal, packet losses can also be due to path fades or intentional interference by an adversary. Some of these effects can be very fast owing to the mobility and agility of electronic attacks. The Transport Layer Protocol will react poorly to these effects, resulting in very low network throughput. This is plaguing many current DOD programs. For the difficult channels DOD and the Air Force deal with, fundamental research is needed to address this large and critical performance gap. Joint design of the Physical Layer and upper Network Layers must be done in these cases.
Network Coding and Diversity Routing for Difficult Channels
For networks using difficult channels, such as channels that change quickly compared with network coordination time, it may be impossible to solve the network routing problem with traditional techniques. Network coding is a new technique that uses almost no buffering, no route computation, and no flow control. The simplicity of this technique could be the answer to these problems (especially for mobile networks that have no infrastructure and dynamic topologies that change too quickly). However, the area of network coding is just beginning to be defined, and there are many theoretical problems in throughput, delays, and congestion control that need to be solved. This topic is explained further later in this chapter.
Because some packets on Air Force networks have hard time-deadlines, those networks must be designed with an understanding of the limits and trade-offs of network throughput, delay, and available resources. Diversity routing, for instance, is one possible technique for increasing performance, albeit at the expense of more network resources. Some network topologies are better for this purpose than others, so there is a need to optimize network performance as a function of topologies and protocol design at the higher layers.
Policy-Driven Efficient Network Resource Allocation
There may be a misconception that with modern fiber and radio technologies, bandwidth is plentiful and should not be a bottleneck to network performance. This is far from true in a tactical battlespace, where there are several limitations to bandwidths and quality of service. First, the amount of asset that can be deployed is limited. Second, difficult channel conditions can drastically reduce rates; in some situations, expensive relay assets must be put in place to provide minimal but critical connectivity. Finally, in adversarial attacks, the number of surviving routes may be few and signaling bandwidths may have to be used to provide AJ via band spreading, reducing the actual data rates. These factors all point to the reality that high-performance tactical network assets will always be precious and in demand. To allocate these precious resources fairly and provide the best operational support to our forces, the network management system must be able to take into account external policy on priorities, which would change from time to time depending on the important missions at the moment and requirements that arise in the field. The network management system should be able to translate these high-level guiding principles into network actions such as routing and media access control priorities in a timely fashion without a human in
the loop. Currently, assets are managed manually, and the process is far from responsive and optimal.
There are risks associated with a network management system that is agile and responsive. To guard against the network going into undesirable states, it needs to be closely monitored for unusual behavior, and fallback procedures and network states must be implemented to ensure minimal critical network service performance. The fundamental research that must be done to support such a vision includes multicommodity resource allocation in a competitive environment, game-theoretic approaches to deal with adversarial attacks, and inference techniques to assess network states in the presence of noise and intentional masking.
Networks Designed to Function Under Adversarial Attacks
There will be many nodes in a tactical network and the span can be wide, with connectivities back to the continental United States from anywhere on the globe. Since it will not always be possible to ensure that no nodes are compromised, the network should be designed to sense dead or malfunctioning network elements and route around them. In addition, network failures often result from operator errors, so the network should have an architecture that confines such damage to a local area and does not allow it to propagate across the network. When the network senses outside attacks, it should have the capability of first locating the real entry points and defending and removing the attacks. Because sensing and other protection techniques can fail, the network should be designed to recognize such failures and be able to continue to function, probably at a lower performance level. This feature is necessary because it may be impossible to avoid rogue nodes or network operator errors. Techniques such as Byzantine robust networking can provide sabotage-resistant routing and defend against compromised trusted network elements. There is some parallel here to fault-tolerant computing, except in this case the adversarial factor must be included in the analysis. Of particular importance are the techniques for refreshing and distributing cryptographic keys for dynamic narrowcast groups. This is especially difficult when a central authority is not readily connected to the population. A provably secure consensus agreement and key exchange mechanism must be devised.
In particular, jamming of open-air communication systems (e.g., satellite links and wireless radios) is a standard military technique to bring about denial of services. Only a few military radio communication systems today have been designed with any AJ capability. Of these, only the satellite system Milstar has adequate AJ capability to deal with jammers that employ modern technologies. Milstar succeeds by deploying spread spectrum and antenna nulling techniques. However, Milstar was designed
for voice communications. When data networking protocols are put on top on this physical layer medium to provide data services in new DOD networks, such as those described by the Transformational Communication Architecture, new vulnerabilities have surfaced that significantly weaken the system’s AJ capability. For example, the adversary need only jam a few bits per Internet Protocol (IP) packet to fool the Transport Control Protocol (TCP) into believing that there is congestion at some routers downstream. In response, TCP will begin closing its transmission window (reducing the number of packets released in flight), reducing the effective throughput of the system to less than 1 percent. Counteracting this new attack requires a combination of techniques from the Physical Layer to the Transport Layer: spread spectrum, nulling, rerouting and diversity routing, and changes to TCP. There are several other known network weaknesses, some of which would appear to be correctable, although history tells us that ad hoc changes to overcome vulnerabilities often open (or overlook) other vulnerabilities. What is needed is a systematic fundamental look at networking, perhaps with a solid mathematical foundation, to provide some assurance of protection. In particular, vulnerabilities to cross-layer attacks (which might be more effective than traditional within-layer attacks) should be examined and addressed.
Cross-Layer Network Design and Optimization
More generally, there would be value in rethinking the network architecture across the layering boundary for military networks, because the difficult channels encountered by DOD are a reality that cannot be avoided. It is easy to state the obvious—namely, that when network layering structures are broken and the architecture is optimized without boundary constraints, the network will perform better. However, the unstructured problem may become so unwieldy that any insight and hope of arriving at a good architecture is lost. There should be a wholesale reevaluation of the functions of each traditional layer to see if there is a better way to group these functions in view of new communication modalities. Even if it is highly unlikely that one could find a layering structure that totally decouples the network architecture, breaking down the architecture problem into weaker interacting subsystems would be a great advance over today’s architecture.
Optimization is a fundamental discipline very important to network design and multiagent control. It relies on tools such as mathematical programming for optimization in static contexts, dynamic programming for optimizing systems that evolve over time, and game theory for optimizing in the presence of competing interests. A key characteristic of the Air Force’s communication networks is their decentralized nature. They
are being constructed from the interconnection of different systems and users with different objectives and performance measures, and their operation relies on varying degrees of collaboration and competition among these entities. Thus, there is a need to understand the optimal allocation of resources in the presence of users with heterogeneous service requirements, which would seem to necessitate the use of game-theoretic and economic market models for resource allocation and network control.
Challenges in this area might include (1) developing game-theoretic and artificial intelligence models for resource allocation problems in different types of networks, (2) studying the optimality/efficiency properties of such multiagent control schemes, (3) developing new theoretical tools in optimization and game theory to analyze convex and nonconvex optimization problems and games that arise in these settings, (4) devising efficient computational methods with established convergence/rate of convergence properties, and (5) developing tractable approximate dynamic programming methods that generate suboptimal control strategies.
Some of the difficult network design problems can benefit from new mathematical analysis tools. In the past few years, ideas developed in statistical physics seem to have had a strong impact in many areas of electrical engineering and computer science. In particular, “replica” and “cavity” methods arising from physics have led to better (heuristic) understanding as well as the development of novel algorithms for hard questions arising in combinatorial optimization, error-correcting codes, statistical inference, and the like. It would be profitable to study these connections rigorously as well as to identify their strengths and weaknesses. This would also lead to a healthy shift in the engineering thinking of large-scale systems such as a globally connected defense network.
Traditionally, network schemes have considered transmission of information in networks in terms of flows—in essence, networks are used in a manner akin to any transportation system. Recently, a radically different approach has been proposed. Rather than considering information as a good to be distributed, one could actively make use of the fact that information is composed of algebraic entities, such as bits, that can be manipulated. One could then consider combining the information of different packets in the interior of the network in order to increase throughput, even in the case of a lossless network. This type of network routing has significant potential application in military networks, where links are unreliable to begin with and there are also adversarial attacks. The main theoretical underpinnings of this area should be developed by merging algebraic theory, networks, and stochastic algorithms to estab-
lish the fundamental performance characteristics of network coding. A strong focus should also be maintained on randomized distributed algorithms for multicast network coding, which has shown promise for network operation in preliminary industry demonstrations.
Some specific basic research challenges in network coding would include the following:
Concurrently explore the theoretical foundation of network coding and applications of that theory.
Bring network coding to the area of network security by developing the ability of network coding to perform algebraic manipulations in the interior of the network.
Consider the economic implications of network coding. In particular, the pricing of resources in network coding differs significantly from traditional network and other pricing mechanisms based on additive resources. Network coding builds on the use of entropy rather than volume as a pricing unit.
Develop the theory and implementation of network coding in lossy networks for data dissemination and for creating altogether new means of performing distributed storage.
Bring network coding to wireless applications. The dynamic and lossy nature of wireless networks renders the use of network coding, particularly for robustness, very effective.
Energy-Efficient Sensor Networks
This is a topic of considerable interest in the private sector, so there are opportunities for leveraging developments there. However, there remain significant research challenges in the area of networked sensor-based embedded systems (for instance, to control swarms of UAVs). Within sensor networks of importance to the Air Force, the following characteristics are common: (1) sensors vary widely in size, computing resources, and sensing information, ranging from very simple sensors like temperature sensors to acoustic sensors and surveillance cameras capturing and distributing video signals, (2) sensors are deployed in very large numbers to ensure adequate coverage, (3) sensors generally perform only a single task (such as sensing temperature or other environment characteristics), but they operate at very high sensing rates to ensure accuracy in environment monitoring, object identification, coordinate location, or other tasks, (4) sensors are deployed in difficult environments where energy sources are not easily accessible and where external forces can damage them, and (5) sensors need to communicate with the environment (other sensors or other IT devices) to distribute information.
The embedded sensor area is relatively new, although some work has been done in the development of operating systems for sensors and in protocols for the Physical, Media Access Control (MAC), Network, and Transport Layers. Related work exists in vehicular sensor networks, applied to sensor information such as Global Positioning System (GPS) positions, video, and others. Efforts are starting in programming and debugging large-scale sensor networks. Many relevant results exist in the use of sensors for location identification, but their integration with other sensing platforms has still not happened. Energy is an issue, and extensive work has been done on minimizing the consumption of ad hoc sensor networks. However, significant research challenges specific to Air Force applications remain in the area of sensor networks:
Development of scalable large-scale sensor network topologies and energy-efficient architectures, including algorithms, operating systems, protocols, and internetworking management for the collection of information, taking into account constrained resources such as energy, communication bandwidth, and processing capacity.
Design/development of efficient algorithms and protocols when sensors are part of a vehicular area network and need to communicate with one another under high-mobility scenarios.
Development of efficient programming support and tools for programming large-scale systems—that is, developing efficient languages, compilers, and debuggers for programming large-scale networks of sensor nodes.
Development of efficient memory systems and caches on sensor nodes to capture the state of the environment.
Development of energy-efficient encoding schemes to distribute information at low data rates.
Development of location identification sensors (e.g., radio-frequency identification), their associated processing algorithms, and their integration into the overall computing/communication environment.
Investigation of security and privacy to protect sensing information.
Incorporation of fault tolerance to ensure reliability of processing at sensing nodes and during information dissemination.
RECOMMENDED BASIC RESEARCH AREAS IN SUPPORT OF AIR FORCE NETWORKS AND COMMUNICATIONS
Some of the challenges described above need more investment by the Air Force than others. In this section the committee recommends a few particularly important basic research areas that typically cut across mul-
tiple challenges. The committee believes that the most valuable results will be achieved when communications and network challenges are considered together, so it has formulated recommendations that target these topics jointly. Because these recommendations are for basic research, the research programs should be tailored to the abstractions that arise from these applications rather than the specific applications themselves.
Satellite Communications and Data Networking
Satellite communications will be an important means of providing linkages between disconnected pockets of radio clusters and theaters to the outside world. Very high frequencies (>20 GHz) are necessary to support high data rates in benign situations and provide good AJ capability with wide-bandwidth, spread-spectrum modulation. Thus far, all military satellite systems have been designed for circuit operations having relatively few users (~1,000s), and the systems are highly inefficient for the transmission of bursty data traffic. These inefficiencies are exacerbated by the unique characteristics of the satellite channels. For example, satellite systems often have longer propagation delays and higher bit error rates than their terrestrial wired counterparts, while the open-air interface for satellite channels (and the high sunk cost of those channels) argues for dynamic sharing of resources. To accommodate data communication for many (~100,000) bursty computer users, the architectures of the Air Force’s satellite communications systems will have to undergo a radical change. At the Physical Layer, the communication system should adapt to weather-induced impairments with power and variable rate control; over 10 dB of efficiency gain could be realized. To serve a large population of bursty data communication users, an efficient and rapid-response MAC Protocol needs to be adopted for the unique properties of the satellite channel. Rapidly changing channel conditions, especially under open-air attacks, will require a fresh look at the routing and flow/congestion control algorithms at the Network and Transport Layers. Since this satellite network will be interconnected to other networks, issues of routing at the Network Layer, reliable packet delivery at the Transport Layer, and network management and control will also need to be addressed. The corresponding research efforts should center on communication theory, experimentation, and practice, with emphasis on heterogeneous networks involving satellites, aircraft, UAVs, and terrestrial wireless and optical systems. Examples of some high-priority research topics include the following:
Design of architectures and protocols for satellite communications systems, including design of satellite constellations, efficient
resource allocation algorithms, Link Layer Protocols for dealing with special properties of space systems, and efficient channel-sharing mechanisms (i.e., MAC Protocols).
Development of hybrid space-terrestrial network architectures, including space-ground network architectures and interfaces, and the design of protocols for internetworking over heterogeneous networks that are efficient, reliable, and able to assure quality of service.
Network architectures and protocols for air vehicle systems, including architectures for reliable communications between autonomous air vehicles (e.g., UAVs) for the purpose of delivering time-critical control information.
Development of analytically tractable models of wideband channels and exploitation of channel characteristics to increase transmission availability, reliability, and/or spectral efficiency.
Development of new lossless and lossy techniques for joint compression and aggregation of correlated sensor signals.
Efficient design and performance analysis of transmission systems, receiver algorithms, diversity methods, and impairment mitigation techniques for air-ground, space, and inter- and intravehicle communications.
Reliable and robust ultrawideband communications for high-speed wireless access, inter- and intra-airplane and -spacecraft communications, UAV communications, and special operations in harsh environments.
Radio Communications and Networking
The development of wireless applications and wireless networks in recent years has motivated some significant improvements in communication theory and information theory. However, maintaining good performance at the Physical Layer for difficult radio channels in the battlefield will be very challenging. Channel conditions and topologies can change rapidly, and the network must efficiently manage its communications over these shared, dynamic, unreliable resources. High-priority research goals in this area include the following:
Development of a methodology that unifies the rich collection of new results in the wireless field. While numerous new techniques have been introduced in recent years to address different aspects of wireless applications (e.g., multiple input, multiple output (MIMO), beam-forming, energy-efficient ad hoc networks), translating these results into fundamental principles requires that we
first merge them into a unifying framework. Designs for modulation and coding, antenna shaping, and processing should be optimized across multiple nodes and with resource scheduling and routing and adversarial attacks all incorporated as key elements in the problem.
Extension of the basic setup of information theory to allow more flexible communications. The primary goal of information theory is to enable reliable communication over long blocks at a prescribed rate. Wireless applications, however, call for breaking such operational scenarios. Many of these applications require communication over dynamic links, with varying qualities and performance requirements. The goal is thus to extend information theory to address the dynamics of information transmission and thereby develop rateless, blockless, self-adaptive communication schemes with delay as a prominent constraint.
Development of cooperation in wireless environments and the architecture of wireless networks. While most of the current wireless networks adopt the cellular architecture, new self-organizing networks promise much better flexibilities, making many more applications possible. A basic research goal in this area would be to study cooperation between wireless nodes with limited information exchange between them and, further, to develop new wireless network architecture that takes advantage of such local cooperation.
The emergence of new applications with diverse requirements has made traditional wireless systems suboptimal, often leading to inefficient use of spectrum, energy, and other resources. It is important to understand how the communications paradigm is evolving and to develop methodologies for the design of flexible, efficient networks. Of particular interest is how decentralized systems—where decisions are made with local information only—can achieve globally optimal use of resources. Also, the research efforts should incorporate factors that are traditionally overlooked, such as the competitive equilibrium and mechanism design for the resource allocation problem, in order to allow the design of fundamentally different systems based on the paradigms of cognitive/software-defined radio and opportunistic communication.
Free-Space Optical Networks
Free-space optical networks are important for Air Force applications connecting satellites, aircraft, and ground terminals. The Physical Layer properties of fading and phase distortion for optical links passing through atmospheric and aircraft turbulence create serious problems for protocols
for higher network layers, often resulting in very low throughputs and long delays for data delivery. It is important to consider optics as a network system rather than an isolated communication link. As described in the section above on Transport Layer Design, the Physical Layer and the network architecture must be codesigned. High-priority areas of research include the following:
Understand the interaction of network protocols with the optical link and quantify the performance shortcomings.
Jointly design the Physical Layer and higher network layers, making judicious use of spatial, temporal, frequency, and route diversity.
Internetworking with other modalities, including using them as backup for guaranteed message delivery.