Cognitive Radio

problem statement

It is well known that radio spectrum (~30MHz ... 30GHz) is inefficiently used, because of our complicated radio regulation. Cognitive radios attempt to overcome this problem. Often, "agile radio" is used as synonym for "cognitive radio." Cognitive radio is not only a new radio technology, it also includes a revolutionary change in how the radio spectrum is regulated.

Today, access to radio spectrum is frustratingly difficult. The access is restricted by an old radio regulatory regime that emerged over the last 100 years. Large parts of our radio spectrum are allocated to licensed radio services, in a way that is referred to as command-and-control. Open access to most of the radio spectrum is only permitted with very low transmission powers, in a so-called underlay sharing approach, as for example used by Ultra Wideband (UWB). The overlay sharing approach, i.e. the free access to open spectrum, is generally not permitted.

Only some small fractions of the radio spectrum, the unlicensed frequency bands, are more or less openly available. Unlicensed frequency bands build a tiny fraction of the entire radio spectrum, where overlay sharing is commonly used. Excitingly, over the past decades, this lead to a wide variety of new wireless technologies and services, among many others the popular Wireless Local Area Network (LAN) IEEE 802.11™ and Bluetooth™ for Wireless Personal Area Network (PAN).

However, spectrum access remains to be a restricting bottleneck that may even slow down the development of new radio services that can substantially improve our health, safety, work environment, education of people, and quality of leisure time.

Changing the status of licensed radio spectrum can be perilous and painfully slow. It takes a concerted effort among government regulatory agencies, technology developers, and service providers to achieve efficient and timely deployment. This is one of the reasons why, paradoxically, 90-95% of the licensed radio spectrum is not in use at any location at any given time. The existing radio regulatory regime is simply too complex to handle the increasingly dynamic nature of emerging wireless applications. As a result, we waste precious spectrum.


On the contrary, the commercial success of wireless applications in the unlicensed bands, and the many radio systems utilizing this fraction of the radio spectrum, indicate that it may be helpful to change our so much established and manifested radio regulatory regime, towards a more flexible, open spectrum access. We may just want to let radio systems coordinate their usage of radio spectrum themselves, without involving regulation. Self-organizing radio systems would then autonomously regulate in a technology-based approach: the machines would make the decisions, not humans. It can only be imagined how our economies would gain from such a flexible, technology-based approach. If successful, this approach would support new emerging wireless applications, at the same time allowing existing incumbent radio services to continue operating without considerable  quality-of-service  degradation. Such an open approach would potentially increase the usage of our radio spectrum significantly.

Cognitive radios are radio systems that autonomously coordinate the usage of spectrum. They utilize radio spectrum when it is not used by incumbent (primary) radio systems. Such unused radio spectrum is called "spectrum opportunity," also referred to as "white space." Figure 1 illustrates such white space in an artificial scenario. The figure indicates idle and busy (=used) frequency spectrum over time/frequency.

Figure 1: White space (unused spectrum opportunities) in our radio spectrum

To use spectrum opportunities with overlay sharing, cognitive radios morph their transmission schemes such that they fit into the identified spectrum usage patterns, as illustrated in Figure 2. The figure indicates how a multi-carrier wideband system such as proposed for Ultra Wideband can change transmission powers, sub-carrier spaces, and sub-carrier bandwidth to optimize spectrum usage. For this, spectrum opportunities have to be identified with high reliability, and their usage needs to be managed in a distributed and coordinated way. A spectrum opportunity is defined by frequency, time, and location. It is a radio resource that is either not used by licensed radio devices, or used with predictable patterns, such that idle intervals can be detected. The predictability and the dynamic nature of spectrum usage contributes to the challenge of identifying spectrum opportunities accurately: the less frequent and more predictable the spectrum usage by primary radio devices occurs, the higher the success of identification, and efficiency of opportunistic usage by cognitive radios.

Figure 2: Opportunistic usage (overlay sharing) with cognitive radio

Spectrum opportunity identification can be further improved by taking characteristic patterns of spectrum usage into account, as well as characteristic signal features of the radio signals transmitted by incumbent radio systems. Different spectrum usage patterns are shown in Figure 3. In this figure, time is progressing from bottom to top, frequency increases from right to left. Here we see patterns of random IEEE 802.11a/e™ spectrum usage (in four channels of the unlicensed 5GHz band), in parallel to a predictable, deterministic spectrum usage (e.g., TDMA/TDD for illustration purposes). The cognitive radio technology, and the ideas discussed here, are inspired by the Defense Advanced Research Projects Agency (DARPA) Next Generation Communication (XG) program. A new standard IEEE 802.22™ is currently being defined. This standard will enable what is referred to as regional area networks. Such networks operate in the VHF/UHF bands, and include mechanisms to protect incumbent radio systems (terrestrial TV broadcast) from harmful interference.

Figure 3: Spectrum usage patterns of IEEE 802.11a™ @5GHz, which provides limited opportunities due to its random access. A more predictable spectrum usage pattern allows more reliable opportunity identification.

research challenges

Algorithms are needed that would enable cognitive radios to utilize the licensed radio spectrum in an overlay sharing approach.

Two main challenges exist:

(1) protection of the incumbents - vertical sharing: A first challenge for cognitive radio is to protect the operation of incumbent, licensed radio services. The additional opportunistic usage of radio spectrum should not harm the operation of already established radio services.

(2) coexistence - horizontal sharing: Another challenge for cognitive radios is to overcome the problem of coexistence and potential free riders in an openly accessible spectrum. When radio systems share spectrum, it is difficult to guarantee that spectrum usage is not monopolized by some radio systems, while at the same time other radio systems would not be able to operate. In fact, this problem is today discussed for the overlay sharing approach in unlicensed frequency bands. a spectrum etiquette is a concept that offers promising solutions to this problem. See below for further details on spectrum etiquette.

See Figure 4 for an illustration of vertical and horizontal sharing. The helpful notion of vertical/horizontal sharing was invented by Jan Kruys/Cisco.

How should radio systems be designed to efficiently use and share radio spectrum, and at the same time not cause significant interference on incumbent radio systems? How can we make sure that our radio spectrum is more efficiently used? In an overlay sharing of open spectrum, how do cognitive radios coexist? How can radio regulatory bodies remain their control, and trace the behaviors of cognitive radios?

Figure 4: Vertical and horizontal sharing.

Spectrum Etiquette

problem statement

Cognitive radios coordinate radio resources autonomously while operating. With this approach, fair and efficient resource sharing between radio systems is difficult to achieve. This horizontal sharing problem is often referred to as coexistence problem. Horizontal sharing is a challenge for cognitive radio as well as traditional radio systems operating in unlicensed frequency bands. Figure 5 illustrates an example for a simple frequency channel allocation problem, where different decisions on what channel to use, will lead to inefficient or efficient spectrum usage. It is the objective of spectrum etiquette rules to provide incentives that will motivate radio systems to allocate spectrum efficiently.


Figure 5: Frequency channel allocation of radio systems with different channel bandwidths. There are inefficient (left), and efficient allocations (right). An etiquette offers incentives to allocate spectrum efficiently.


The problem can be approached with spectrum etiquette rules that are voluntarily used.

A spectrum etiquette is a set of rules for radio resource management to be followed by cognitive radio systems that share the radio spectrum. Spectrum etiquette often helps to establish fair access to the available radio resources, in addition to a more efficient usage of radio spectrum. Spectrum etiquette rules are typical based on actions like dynamic channel selection, transmission power control, adaptive duty-cycles, and carrier sensing (listen-before-talk).

A spectrum etiquette defines the rules for the behavior of radio systems mainly in order to achieve two goals. First, if all radio systems follow the spectrum etiquette, fairness in access to the shared radio resources is maintained, and second, the frequency band is more efficiently used. A spectrum etiquette is independent of any radio transmission scheme and protocol: it does not define a protocol and is not restricted to one radio standard. Further, a spectrum etiquette is not an algorithm that describes the entire radio resource management of all radio systems. Each radio system can apply its own algorithms within the constraints of the spectrum etiquette. The spectrum etiquette provides a framework for behaviors, which may restrict the degrees of freedom in radio resource management of the individual radio systems, but leaves room for innovations and differentiations between devices from different vendors.

research challenges

In order to motivate manufacturers to implement spectrum etiquette rules, convincing sets of rules are needed that provide incentives for radio systems to operate spectrum efficiently, for the benefit of all radio systems. What rules are acceptable for radio systems? How can they be defined independently of existing radio standards? How to approach the tragedy of commons: if today's radio systems operate spectrum efficiently, other competing radio systems benefit from their nice behavior. So why should a manufacturer actually build such efficient devices?

Social Science and Game Theory

problem statement

A contemporary society is a group of socially acting individuals (i.e., "actors") where each individual acts according to classified motivations. An action is for example the selection of radio transmission parameters such as transmission powers and the frequency of operation. Actor’s interests can be expressed through their application requirements (for example, the throughput in Mb/s that an application would require), as it is typically applied when games are used to analyze communication protocols. Figure 6 gives a flavor about how games are applied: actors are implemented with decision-making algorithms that are part of a discrete-time systems. Actor's information covers their own requirements, and the observation of outcomes from the history of previous interactions. This is a model for repeated interaction in multi-stage spectrum sharing games.

Figure 6: Decision making in spectrum sharing scenarios by actors.

However, we are not only interested in the economic self-interests of actors, but also the actor’s value-orientations. Value-orientation is a concept in contemporary societies with social awareness. Sociological models known from Max Weber's early work may be applied to build a framework for a society of machines. These models allow designing new types of radio systems that are not only self-aware, but also socially aware. This approach for autonomous spectrum sharing supports our goal to achieve technology-based radio regulation with cognitive radios that manage and regulate spectrum without human interaction. By introducing value-orientation, cognitive radios can be designed so that they are capable of coordinating their spectrum usage. Whereas until today, cognitive radio systems are thought of as pure technocratic decision-makers, value-oriented cognitive radios will support the introduction of true cognition, for example for the purpose of higher efficiency in spectrum usage.


Cognitive radios that share the radio spectrum can be interpreted as forming a virtual society. This virtual society faces many challenges that have been already analyzed for real-life societies as for example interacting nations, or groups of animals such as ant colonies. The theory of games is an efficient means for analyzing such societies. Cognitive radios will be required to act environment and interference aware when operating with shared radio resources. They have to operate with high flexibility, by also considering the implications of their decisions on other radio systems. It may be helpful to extend the traditional way cognitive radios are defined by introducing social awareness. Cognitive radios that are aware of their society, i.e., aware of the existence and demands of other radio systems, can benefit by supporting not only their own interests, but also the interests of the other radio devices.

Cognitive radios make decisions about what action to take; hence the reference to actor. There is a predefined set of valid actions that an actor can take; this set spans the so-called action space. According to Weber, an action is usually associated with a subjective meaning. Two classes of action exist: social action and economic action. An action is social, if "its subjective meaning takes account of the behavior of others and is thereby oriented in its course." An action that intends to improve the actor’s outcome is an economic action. However, the subjective orientation of an actor is what we are particularly interested in when discussing the concept of social action.

Weber's classification of social action distinguishes between four types:

  (1) Technocratic social action ("zweckrational")
  (2) Value-oriented social action ("wertrational")
  (3) Affective social action (based on emotional state)
  (4) Traditional social action (guided by custom)

Here we focus on the first two types of social action. It is highlighting to interpret the existing way of radio regulation and spectrum sharing as technocratic social action: whatever other individuals observe, each individual attempts to optimize it’s own spectrum usage only. This often leads into chaotic and unpredictable scenarios of spectrum usage if there had been no regulation: the purely technocratic social actions of today’s radio systems leverage the existence of radio regulation. However, we are interested in opening the spectrum for free usage, with a minimum amount of radio regulatory constraints. Therefore, motivated by the findings of Max Weber, it may be helpful to apply value-oriented social action - the same social concepts that allowed human beings to live without dominant regulation of their daily interactions. For the spectrum-sharing problem, the decision-making processes in cognitive radios is then modified for supporting the social concepts of value-orientation, and hence to implement voluntary rules.

Game theory provides the means to analyze spectrum sharing. Coexistence and open spectrum sharing problems can be modeled as games of interacting decision makers. A game is based on a set of entities that are referred to as players, or actors. Results of interaction can be estimated for single stage games including the concepts of action, utility, preference and behavior. A Nash analysis allows predicting the outcomes of rational play. A multi stage game comprises a set of actors that choose their actions in each of the consecutive stages of the game to maximize their own utility in the current and future game stages, given their assessment of their opponent’s actions in that particular stage. The game is called a dynamic game if actors adapt their action to the environment from stage to stage.

As an example, consider the normalized results shown in Figure 7. In this example, actor 1 interacts with two other actors and can only control its own protocol parameter, the contention window (CW), a parameter used in contention-based medium access protocols like the listen-before-talk in IEEE 802.11™. In general, the smaller the CW, the more aggressive the medium access. The more aggressive the medium access, the higher the outcome when opponents cooperate (i.e., when other actors do not select a small CW as well, as indicated in the left subfigure of Figure 7). As shown in the right subfigure of Figure 7, if all players select a small CW, the overall results are very poor, as seen in the marked area (3). This is a classic game problem.

Figure 7: Actors outcome in a spectrum sharing scenario. One actor (left) and the cumulative results of all actors (right) are shown. By selecting an action, actor 1 can only control its own parameter (CW=contention window), and hence depends on its opponents to improve its own outcomes.

research challenges

What multi stage strategies are robust and lead to predictable outcomes? What strategies lead to cooperation between interacting actors? In what scenario? How to build a strong and sustainable reputation? How to model it? Are coalitions of cooperating actors stable against intruders that attempt to break an established cooperation??

Ontology Engineering

problem statement

True cognition requires semantic concepts that can be implemented with ontology representations. Sets of ontology representations can be used to describe knowledge about the radio domain and further to describe algorithms about how to utilize radio resources - enabling cognition. With our goals of opening our radio spectrum with the help of the cognitive radio technology and the concepts we may adopt from social science and game theory, we may want to explore new engineering methods to develop the radio systems of the future. Cognitive radios have to make decisions that are traceable: When decisions are made, a cognitive radio may be required to formulate its objectives (goal of reasoning), and how it came to the final decision.


This idea is again inspired by the Defense Advanced Research Projects Agency (DARPA) Next Generation Communication (XG) program. We want to explore research opportunities to describe knowledge about wireless communication in a machine-understandable way: To engineer machine-understandable radio semantics using ontologies, as discussed in the following.

A taxonomy classifies information entities, so that they can be browsed or navigated for information. A taxonomy is a (semantically weak) hierarchy in which information entities are related to each other. An information entity is unique and identifiable by its distinguishing properties (also referred to as attribute, quality) that make it unique. An example for taxonomy is "unused spectrum is a spectrum opportunity." Such expressions can be used to express semantic content (meaning) with relationships such as for example "synonym for," or "associated with."

An ontology extends the concept of taxonomy: ontology is a "vocabulary of terms and the precise relationships between them." It is a collection of classes and properties grouped together: ontology extends the concept of taxonomy towards logical theory. Ontology can be taxonomy when it has weak semantic contents. Ontology can be a logical theory when it has strong semantic content. Ontology tries to capture the meaning of a particular subject area or area of knowledge that corresponds to what a human being knows about that domain. Ontology makes the radio semantics domain knowledge usable. Ontology includes

   - Classes (general things) and instances (particular things) in the domains of interest
   - Relationships among things
   - Properties (= attributes) of things
   - Functions of and processes involving things
   - Constraints on and rules involving things

Ontologies encode the description of knowledge that enables a cognitive radio to use the description. Ontology tools such as Stanford's Java Theorem Prover (JTP) can perform automated reasoning about the knowledge. The language of choice for representing ontology is the Web Ontology Language (OWL).

Ontologies represented as logical theories are directly semantically interpretable by the opportunity manager software engines. Logical theories are built on axioms and inference rules. Axioms are statements that are declared as true. Inference rules are rules that, given assumptions, provide valid conclusions. Axioms together with inference rules can be usedto prove theorems about the knowledge domain. The set of axioms, inference rules, and theorems together constitute the logical theory to express radio semantics.

With an ontology set describing a domain of knowledge it is argued that first order predicate logic enables actors to make decisions: A predicate (propositional function) is used to make a statement about something, e.g. to attribute a property. If it is said "radio resource A is an opportunity," then there is the predicate "is an opportunity" applied to a radio resource. We also might say that we have predicated "being opportunity" of radio resource A, or attributed "opportunity" to radio resource A.

A predicate applies to an instance (in our example radio resource A). It hence yields to a proposition (for example "A is an opportunity"). "First-order" refers to the fact that quantification is only permitted over instances, and not over predicates. It is claimed that first-order logic permits reasoning. Figure 8 illustrates a graphical user interface (GUI) that allows designing an opportunity manager for cognitive radios.

Figure 8: Opportunity manager graphical user interface. Knowledge can be imported with input statements. Automated reasoning allows extracting knowledge.

research challenges

Can we describe radio communication, protocols, algorithms, and regulatory constraints with the help of semantic concepts based on ontology? How efficient would be a reasoning engine included into cognitive radios? What are the new algorithms for radio resource management that would enable traceable decision-making? Is it too complex?

  I am an employee of Swisscom. The comments made here are my own
and do not necessarily reflect the official position of Swisscom.