Spectrum Agile Radio: Radio Resource Measurements for Opportunistic Spectrum Usage

Spectrum Agile Radio: Radio Resource Measurements for Opportunistic Spectrum Usage Stefan Mangold, Zhun Zhong, Kiran Challapali Wireless Communication and Networking Department Philips Research, 45 Scarborough Rd. Briarcliff Manor NY 5, USA {stefan.mangold, zhun.zhong, kiran.challapali}@philips.com Abstract Radio spectrum allocation is undergoing radical rethinking. Regulators, government agencies, industry and the research community recently established many initiatives for new spectrum policies and seek approaches to more efficiently manage the radio spectrum. In this paper, we are examining new approaches, namely, spectrum agile radios, for opportunistic spectrum usage. Spectrum agile radios use parts of the radio spectrum that were originally licensed to other radio services. A spectrum agile radio device seeks opportunities, i.e. unused radio resources. Devices communicate using the identified opportunities, without interfering with the operation of licensed radio devices. The identification of spectrum opportunities is coordinated by policies, which are defined by, and under the control of, the radio regulator. Our approach is motivated by the publications of the Next Generation Communications, XG, research project of the US-based Defense Advanced Research Projects Agency, DARPA. We focus on IEEE 8.k for radio resource measurements as an approach to facilitate the development of spectrum agile radios. Keywords spectrum agile radio, opportunistic spectrum usage, radio resource management, IEEE 8.k I. INTRODUCTION W ireless communication became increasingly popular over the last decades. The overall acceptance of wearable, hand-held computing and communicating radio devices, as well as consumer electronics, will continue to result in an ever-increasing demand for radio communication networks providing high capacity communication. However, this results in an increasing demand for radio spectrum, which is scarce. Such scarcity is due to existing licensing regiment, i.e., radio spectrum allocation. Radio spectrum is traditionally licensed through spectrum policy in a non-flexible way. Today, spectrum policy in the US is undergoing radical rethinking, which motivates us to discuss innovative new approaches for opening the radio spectrum by allowing opportunistic usage of licensed but unused radio resources. A radio resource is defined as follows. Radio resources are frequency bands that can be used/occupied for certain duration, in a certain area. The larger the area, the broader the frequency band, and the longer the duration of allocation, the more radio resources are occupied. We describe in this paper a new approach for radio regulation to improve the utilization of radio resources that we refer to as Spectrum Agile Radio (or, Agile Radio). The agile radio approach is motivated by a US-government project, which is coordinated by the Defense Advanced Research Projects Chun-Ting Chou Real-Time Computing Laboratory The University of Michigan Ann Arbor, MI 489-, U.S.A. choujt@umich.edu Agency (DARPA), and referred to as Next Generation Communication (XG) Program []-[]. The problem statement and the agile radio approach are discussed in the rest of this introduction, see Section A and B. In Section II, we describe a typical agile radio usage scenario to highlight the XG idea with focus on commercial application. The DARPA XG project, and what we learned from the XG publications, is briefly summarized in Section III. A new type of radio measurement that is developed at standardization of IEEE 8.k (the letter k indicates the task group, developing mechanisms for new radio resource measurements, see [4]) is discussed in Section IV. We evaluate this measurement and its usage for agile radios, and other spectrum analysis purposes. Simulation results indicate that the measurement is indeed helpful for radio resource control in IEEE 8. networks, but additionally has the potential to provide information about spectrum usage patterns that are generated by other, non-8. radio devices. The paper ends with a conclusion in Section VI. A Problem Statement Considering the increase in demand for freely available, i.e., unlicensed, radio resources, it is clear that the necessary radio spectrum will not be available in the future, due to the limited nature of radio resources in the current unlicensed frequency bands. Radio communication systems that support consumer electronics operate mainly in unlicensed frequency bands. Radio resources of the unlicensed frequency bands are generally considered as being efficiently used, because of the high penetration of unlicensed radio devices. Only a small fraction of the entire radio spectrum is regulated with the unlicensed approach, mainly because many services require protection against interference from other radio devices (for example TV-broadcasting, as currently discussed at IEEE 8.8). Protection against interference can be guaranteed by licensing of radio spectrum, and exclusive usage. Most of the radio spectrum is for this reason allocated to traditional licensed radio services, which results in inefficiencies: radio spectrum is not efficiently used if licensed radio services are commercially not successful in the market. Another reason for inefficient usage of radio spectrum in the licensed bands are radio services that only occasionally require radio spectrum, for example emergency calling and safety services/disaster relief communication services. Hence, with the traditional regulation of radio spectrum we have today, most radio resources are not efficiently used. B Technical Approach The described problem is approached by Spectrum Agile Radio (Agile Radio). To improve the efficiency of the spectrum usage, Globecom 4 467-78-8794-5/4/$. 4 IEEE

4 time [ms] 5 45 Licensed spectrum, not used at all: 5 5 5 5 5 Unlicensed spectrum, heavily used (4 channels, MHz each) Unlicensed (4 channels) 58 56 54 5 agile radio devices operate in the licensed radio spectrum in an opportunistic way. An agile radio device seeks opportunities, i.e. unused radio resources. See Figure for an illustration of spectrum opportunities. Different types of opportunities are indicated. Spectrum opportunities occur if spectrum is not used at all, or used with deterministic pattern. We discuss later in this paper how to detect such spectrum opportunities. Agile radio devices communicate by using only the identified opportunities, without interfering with the operation of licensed radio networks. Identifying spectrum opportunities is regulated by policies. In agile radios, policies are understood as rules for operation that are phrased in a machine-understandable form, based on the popular Extensible Markup Language (XML). Policies are made available to agile radio devices and networks for example with the help of memory devices such as flash cards, or by downloading them from a server. One of the main objectives of agile radio is the interference preservation of primary radio services of incumbent radio systems, such as TV broadcast networks in the TV bands. With the help of the machine-understandable policies, radio regulators can carefully, and eventually stepwise, open the spectrum for more flexible usage. The policies will allow regulators to minimize the imposed restrictions that typically come with regulation, and at the same time protect non-agile, incumbent radio services. Agile radio devices will adapt to the policies and modify their radio resource management, for example by selecting other frequencies and transmission powers. Once a core set of agile radio policies is designed and made available to the agile radio devices, this core set enables then the devices to determine how and when to make use of radio resources, hence to manage the spectrum usage autonomously. With such an adaptive approach, the authority on radio regulation and spectrum management remains at the existing regula- This approach is not new. The DARPA XG program ([],[]) released the XG policy language in [], which will facilitate developing agile radios. In contrast to XG, our vision of agile radio technology focuses on consumer applications, which are build on top of existing radio standards, for example IEEE 8. for wireless local area network, or the emerging Ultra Wideband (UWB) IEEE 8.5. 5 58 Licensed spectrum, used with deterministic pattern: frequency [MHz] Figure : Spectrum usage pattern in four 8.a channels in the unlicensed 5GHz frequency band, and deterministic pattern of primary (simulation). Time progresses from bottom to top. The gray fields indicate busy channels. 56 54 5 5 tory body, which is for example the Federal Communications Commission (FCC) in the US. An example usage scenario that illustrates the employment of agile radios is given in Section II, to highlight the characteristic of this new way of managing the spectrum usage. II. SPECTRUM AGILE RADIO USAGE SCENARIO The technical approach presented in the last Section I.B allows us to outline a typical expected (imaginary) agile radio usage scenario. Let us assume that the penetration of radio devices (example: Wi-Fi ) that operate in the unlicensed frequency bands (example: Unlicensed-National Information Infrastructure (U-NII)) in a given regulatory domain (example: New York State) is very high, too high to meet the growing demands of consumers. For the existing and the emerging radio services (example: wireless data-, video-, and medical applications, with throughput demands greater than Gb/s (Gbps)), the unlicensed U-NII frequency bands are likely to become saturated. The radio regulation authority (example: FCC) for the regulatory domain recognizes this scarcity of radio resources as potential barrier for social/economic development. The authority therefore identifies a licensed band of the radio spectrum where the penetration is not high, because corresponding assigned licensed radio devices are not used due to the nature of the radio service (example: terrestrial TV broadcast of some unused TV channels). The corresponding licensed band is identified as potentially available for use by agile radio devices. The authority assigns a set of policies that provide rules and constraints of how to use this band. The set of policies are published in a machine-understandable form. The policies are published for download from servers of the radio regulation authority. Agile radio devices repeatedly seek for updates of policies (example: once a day) that are relevant for their regulatory domain. The devices that are located in the regulatory domain for which new policies have been published, download the machineunderstandable policies, and update their local information bases. Alternatively, policies are made available through memory devices such as flash cards, to allow agile radio devices that do not have access to servers to update their information bases. After the local information base has been updated, agile radio devices systematically adapt their behaviors (example: set of candidate frequency channels in the licensed bands that are eventually now permitted to be used, constraints on spreading factors and transmission powers), and disseminate the new policies. The dissemination enables other agile radio devices to learn about the policies from each other. Some agile radio devices may not be capable of using the new radio resources, and ignore the updated policies. Other devices learn about the radio environment in the licensed band, attempt to identify the existence of licensed radio devices, and derive behaviors from the given policies. A behavior can be for example the usage of listen-before talk, or dynamic frequency selection. The agile radio devices identify spectrum opportunities, and disseminate the characteristics of the identified spectrum opportunities within their networks. The process of deriving behaviors from policies and the process of modifying the radio access of an agile radio device must be traceable. This means that the impact of the opportunistic spectrum usage on existing licensed services has to remain under the control of the radio regulation Globecom 4 468-78-8794-5/4/$. 4 IEEE

SARA Constraints -sharing technology -policy restrictions -waveform regulation..*..* «metaclass» SARA Policies SARA Abstract Behaviors «uses»..* «utility» SARA Policy Meta Language SARA Core Behaviors SARA radio communication system SARA Protocols «implementation class» SARA Real Life Implementation SARA abstract behaviors are composed by set of core behaviors Figure : The agile radio approach illustrated with the universal modeling language. SARA=Spectrum Agile Radio. authority. Agile radio devices belonging to different networks may not directly communicate with each other, which makes information exchange difficult. In this case, devices that belong to different networks, but attempt to operate with the same radio resources, may coordinate sharing of radio resources with the help of spectrum etiquette, as discussed in [5], [6]. As a resulting step in our example scenario, the agile radio devices make use of the new spectrum opportunities in accordance with the updated radio regulation, by at the same time not interfering radio transmissions of existing primary, i.e., licensed radio devices. This resolves the problem of spectrum scarcity in the given regulatory domain in our example. III. DARPA S XG PROJECT An outline of the Defense Advanced Research Projects Agency (DARPA) Next Generation Communication (XG) project is given in this section. Note that our work is independent of the DARPA XG project. The structure, objectives, and the planned schedule of the DARPA XG project as described in [], [] are reviewed in the following. The XG project operates with socalled Request For Comments (RFCs), and seeks feedback from the research community on the selected approaches. At the time this paper is written, three RFCs have been released (see []- 7 Density Vector (m=8) thresholds 7 6 6 5 p() 4 5 5 4 5 5 4 4 Density Reported densities are Received Power Indicators (s) [..55] for probability p of occurrence: p() Density(), =...7 55-57dBm -87dBm Figure : IEEE 8.k Noise Histogram report with relative probabilities of the measured received powers. =Received Signal Indicator, m=number of densities. A density [-55] corresponds to a probability of occurrence [.-.]. The numbers in this figure are only examples, and may vary from report to report. time []). Our work on agile radios may be understood as the extension of IEEE 8. or UWB under consideration of the XG way of radio resource regulation, and management. The agile radio approach is in principle suitable for any type of radio network, including wide area networks. It is not restricted to local or personal area networks, with their typical short communication distances. The concept of the XG project is based on so-called abstract behaviors, protocols, and a policy language []. The reasons for this approach are mainly flexibility, long-term impact, and the need for regulatory approval []. In other words, behaviors are used instead of detailed descriptions of a standardized protocol, or a set of different standardized protocols, to allow regulators and industry to dynamically align future regulatory requirements and rules for spectrum usage with existing and emerging technologies for future radio systems. It is planned in the XG project to develop proprietary solutions based on this existing radio standard. However, it is mentioned in [], that the final behavior definitions and the policy-based approach in general should be applicable for a wide variety of radio standards, including third generation wide-area cellular networks, and future emerging standards. Figure illustrates the different levels of abstraction. This figure is a modified version of an illustration in []. It illustrates (from left to right) policies, behaviors, protocols, and the real life implementation. Policies use a policy meta language as utility. There is a direct association between policies and technical constraints. Abstract behaviors are derived from policies. A behavior is composed by core behaviors. Protocols are derived from behaviors, realized by the real implementation. IV. IEEE 8.K AS STEP TOWARDS SPECTRUM AGILE RADIO The agile radio technology can be built on top of existing radio communication standards such as IEEE 8. with its extensions for radio resource measurements. The IEEE 8. Task Group k (TGk) develops radio resource measurements as an extension to the IEEE 8. standard for wireless local area networks. This extension will specify the types of radio resource information to measure and the request/report mechanism through which the measurement demands and results are communicated among stations. We discuss in the following the application of the existing measurements of IEEE 8.k for Globecom 4 469-78-8794-5/4/$. 4 IEEE

Reported densities are indicators [..55] for probability p of occurrence: Density: Bin: p(bin) Density(bin), bin=...6 55 p(bin) bin Density Vector (m=var.) 89 6 4 5 6 SIFS DIFS PIFS aslottime Figure 4: Medium Sensing Time Histogram report with six bins. Slot access probabilities allow deriving information about ongoing medium accesses from other radio devices. A bin corresponds to a probability of occurrence of a certain idle or busy duration. identification. The goal of this new extension is to provide tools by which a radio device can measure and assess the radio environment and take corresponding actions. To fulfill this goal, the current IEEE 8.k draft defines different types of measurements (see [4] for details). Note that radio resource measurements are in general not required to be standardized for a communication protocol. However, what need to be standardized are the frame formats of a request for a specific measurement (from one radio device to another), and the respective report of such a measurement, for example as response to the request. Among other measurement reports, with the Channel Load report, a measuring device reports the fractional duration over which the carrier sensing process, i.e., Clear Channel Assessment (CCA), indicates the medium is busy during the measurement period. In the Noise Histogram report, a measuring device reports non-8. energy by sampling the medium only when CCA indicates that no 8. signal is present. This report is illustrated in Figure. Each of the fields in the left hand side of the figure is associated with the probability of a certain detected energy level. The key measurement for agile radios is the Medium Sensing Time Histogram report, which was developed by the authors, density CCA idle time histogram Bin Interval = aslottime Bin Offset = SIFS Number of Bins = 6 4 5 6 Figure 5: Medium Sensing Time Histogram for scenario with low traffic load, CSMA based medium access. Idle durations are often longer than.5ms, busy durations are distributed equivalently to packet sizes. Two stations operate in parallel. time prob (idle dur.)....5..5..5..5.4.45.5.5.4.......4.5.6.7.8.9 Figure 6: Medium Sensing Time Histogram for scenario with high traffic load. Now, only short idle durations occur. The idle durations are geometrically distributed. Two stations operate. see [9]. See Figure 4 for an illustration of the report. A measuring station reports the histogram of medium busy and idle time observed during the measurement period. The states busy and idle are typically defined by CCA. This measurement is evaluated in the following for assessing the spectrum usage pattern of other radio devices. V. SIMULATION ANALYSIS We analyze how the Medium Sensing Time Histogram report may help to determine the spectrum utilization. A Simulation Environment Event driven simulation of IEEE 8.a/e/k are used to illustrate the potential of the existing 8.k measurements to determine the spectrum utilization for a given frequency channel. The entire simulation environment, including the eventhandling, is implemented in an object-oriented design, using MATLAB. With this simulator, we are able to model the medium sensing time histogram measurement in detail, and evaluate the measurement results for a wide variety of scenarios. B Scenarios and Results In our model, the radio channel is error-free, traffic generators create uncorrelated packet arrivals with neg.-exponential distribution, if not stated otherwise. Transmission rate is 6 Mb/s (Mbps). We investigate multiple scenarios of spectrum usage, with different numbers of stations, and different offered traffic per station. One second of spectrum usage is measured in all scenarios. Measurements Figure 5 and Figure 6 show the results for medium sensing time histogram measurements, for CCA idle (top) and CCA busy (bottom). Comparing the two figures, it can be seen that the busy durations are similarly distributed. The busy durations only depend on the packet sizes, and therefore are similar in each result. The idle durations, however, indicate clearly the difference of the spectrum usage in the two scenarios. As expected, with low offered traffic, idle durations are typically longer, whereas with high offered load, idle The model includes a for the 5 GHz unlicensed band, e for quality of service support with the contention-based medium access (Enhanced Distributed Coordination Function, EDCA), and k for radio resource measurement. Globecom 4 47-78-8794-5/4/$. 4 IEEE

prob (idle dur.)... prob (idle dur.)... peak indicates that different stations use different carrier sensing parameters.5..5..5..5.4.45.5.5.4.......4.5.6.7.8.9 Figure 7: Medium Sensing Time Histogram, for a scenario of ten stations operating in parallel. Only very short idle durations occur, no opportunities for an agile radio device. durations are short. Note the characteristic geometric distribution of the idle durations, confirming the results in [5]. Figure 7 indicates how short the idle durations become if the number of stations is increased from two to ten. Two more interesting results are shown in Figure 8 and Figure 9. Figure 8 clearly indicates how deterministic, periodic spectrum usage can be identified by the medium sensing time histogram. Figure 9 illustrates that once stations use different contention parameters, such a heterogeneous scenario is detected by the measurement. The peak in the reported histogram of idle durations in Figure 9 indicates that some stations access the medium later than others, which is typical for an IEEE 8.e environment. VI. CONCLUSION The US policy for spectrum licensing is undergoing fundamental rethinking, which may impact research in wireless communications. Spectrum Agile Radio shows the potential to provide a solution for dynamic and flexible spectrum licensing and dynamic radio resource management, to migrate into a highly flexible way of radio regulation. We have shown usage scenarios, and illustrated our approach for spectrum agile radio. We Deterministic spectrum usage Figure 8: Medium Sensing Time Histogram for deterministic, periodic medium access. The idle and the busy durations are fixed, and future idle durations can be predicted..5..5..5..5.4.45.5.5.4.......4.5.6.7.8.9 Figure 9: Medium Sensing Time Histogram for 8.e medium access with multiple different contention parameters. The local maximum indicates that some stations operate with low access priority. This is a potential opportunity. discuss existing radio resource measurements of the existing IEEE 8. protocol, which may provide first steps towards spectrum agile radios. However, we focus on spectrum access with contention-based protocols. Other measurements may be needed for time/frequency division. Further, we inherently assume that idle spectrum is unused spectrum, which may not be the case in real life: an incumbent device may wish to remain idle in order to perform measurements. It is often claimed that with a successful deployment of spectrum agile radios, an increase of availability of radio resources in the order of ten can be anticipated. This is in the interest of regulators, service providers, and consumers. REFERENCES [] DARPA XG WORKING GROUP (a) The XG Vision. Request for Comments, version.. Prepared by: BBN Technologies, Cambridge, Massachusetts, USA. July. [] DARPA XG WORKING GROUP (b) The XG Architectural Framework. Request for Comments, version.. Prepared by: BBN Technologies, Cambridge, Massachusetts, USA. July. [] DARPA XG WORKING GROUP (4) XG Policy Language Framework. Request for Comments, version.. Prepared by: BBN Technologies, Cambridge, Massachusetts, USA. April 4. [4] IEEE 8. WORKING GROUP () Draft Supplement to STANDARD FOR Telecommunications and Information Exchange Between Systems - LAN/MAN Specific Requirements - Part : Wireless Medium Access Control (MAC) and Physical Layer (PHY) specifications: Specification for Radio Resource Measurement, IEEE 8.k/D.7. New York USA: The Institute of Electrical and Electronics Engineers, Inc. [5] MANGOLD, S., AND HIERTZ, AND WALKE, B. (a) IEEE 8.e Wireless LAN - Resource Sharing with Contention Based Medium Access. In: IEEE Personal Indoor Mobile Radio Conference Beijing P. R. China 7- September. [6] MANGOLD, S. () Analysis of IEEE 8.e and Application of Game Models for Support of Quality-of-Service in Coexisting Wireless Networks, PhD Thesis, ComNets, Aachen University,. [7] MANGOLD, S. AND CHALLAPALI, K. () Coexistence of Wireless Networks in Unlicensed Frequency Bands. Wireless World Research Forum #9 Zurich Switzerland, July. [8] CHALLAPALI, K. AND MANGOLD, S. AND ZHONG, Z. () Spectrum Agile Radio: Detecting Spectrum Opportunities. International Symposium on Advanced Radio Technologies 4 Boulder Colorado USA, Mar 4. [9] ZHONG, Z. AND MANGOLD, S. AND SOOMRO, A. () Proposed Text for Medium Sensing Measurement Requests and Reports. IEEE Working Document 8.-/4r. May-,. [] MANGOLD, S. AND CHOI, S. AND HIERTZ, G. AND KLEIN, O. AND WALKE, B. () Analysis of IEEE 8.e for QoS Support in Wireless LANs. IEEE Wireless Communications, pp. 4-5. Dec. Globecom 4 47-78-8794-5/4/$. 4 IEEE