A Novel Method to Represent Spectrum Management Policies

4
A Novel Method to Represent Spectrum Management Policies Li Da-li Nanjing Telecommunication Technology Institute Nanjing, China [email protected] Cao Long Institute of Communications Engineering PLA University of Science and Technology, Nanjing, China [email protected] Abstract—Policy Reasoner (PR) plays a key role in the implementation of the policy-based approach. This paper introduces a novel method to express spectrum management policies, which can facilitate the design of the reasoning algorithms. We compare it with the traditional way and give the simulation results. Finally its advantages and applicable condition are summarized. Keywords-spectrum management; policy-based; policy reasoner; decision tree; binary decision diagrams I. INTRODUCTION The static spectrum allocation policy, where governmental agencies assign wireless spectrum to licensed holders on a long-term basis for large geographical regions, leads to two significant problems: spectrum scarcity and deployment delays. Assigning frequencies statically has left little of the spectrum available for growth of new technologies. Moreover, studies have shown that most of the assigned spectrum goes unused for a majority of the time, which is a waste of resource. As the solution to these problems, a policy-based [1] approach has been proposed by the DARPA’s neXt Generation (XG) Communications Program to manage spectrum resources more effectively and efficiently. In such a policy-based mode, regulators develop and distribute spectrum policies dynamically, while cognitive radios adjust their own transmission activities in order to conform to the policies. To implement the policy-based spectrum management, radios must understand and reason policies from regulators, which can be realized by the Policy Reasoner (PR). This paper is outlined as follows: After introducing the XG radio architecture and the reasoning process in Section II, we propose a novel way to represent spectrum policies using decision tree in Section III. Section IV gives the reasoning process of policies represented by decision tree and compares it with the traditional binary decision diagram method (BDD) way. The paper is concluded with a discussion on advantages and drawbacks of the two policy representation methods in Section V. II. XG RADIO ARCHITECTURE AND POLICY REASONER A. XG Architecture DARPA’s XG Program presents a radio architecture which ensures that the transmission activities conform to spectrum policies. As Fig.1 indicates, an XG radio [2] contains four main components. Figure 1. XG architecture 1) Sensors. XG radios need sensors to discover available spectrum and transmission opportunities 2) Radio frequency (RF). The RF component transmits and receives. 3) System Strategy Reasoner (SSR). The SSR controls the radio’s transmissions. It builds transmission requests based on sensor data received from the sensors and its current strategies. Replies to its requests from the PR may affect strategy. 4) Policy Reasoner (PR). The PR accepts transmission requests from the SSR and checks policy conformance. The PR has all active policies loaded. In general, the PR oversees every transmission activity by checking the transmission requests from SSR. When accepting requests from the SSR, the PR checks policy conformance. It replies with yes, no, or no with constraints to be satisfied. The PR has a policy base with all active policies, which can be modified and updated as required. 978-1-4673-0915-8/12/$31.00 ©2012 IEEE 2012 3rd International Conference on System Science, Engineering Design and Manufacturing Informatization 159

Transcript of A Novel Method to Represent Spectrum Management Policies

Page 1: A Novel Method to Represent Spectrum Management Policies

A Novel Method to Represent Spectrum Management Policies

Li Da-li Nanjing Telecommunication Technology Institute

Nanjing, China [email protected]

Cao Long Institute of Communications Engineering

PLA University of Science and Technology, Nanjing, China

[email protected]

Abstract—Policy Reasoner (PR) plays a key role in the implementation of the policy-based approach. This paper introduces a novel method to express spectrum management policies, which can facilitate the design of the reasoning algorithms. We compare it with the traditional way and give the simulation results. Finally its advantages and applicable condition are summarized.

Keywords-spectrum management; policy-based; policy reasoner; decision tree; binary decision diagrams

I. INTRODUCTION The static spectrum allocation policy, where

governmental agencies assign wireless spectrum to licensed holders on a long-term basis for large geographical regions, leads to two significant problems: spectrum scarcity and deployment delays. Assigning frequencies statically has left little of the spectrum available for growth of new technologies. Moreover, studies have shown that most of the assigned spectrum goes unused for a majority of the time, which is a waste of resource.

As the solution to these problems, a policy-based[1]

approach has been proposed by the DARPA’s neXt Generation (XG) Communications Program to manage spectrum resources more effectively and efficiently. In such a policy-based mode, regulators develop and distribute spectrum policies dynamically, while cognitive radios adjust their own transmission activities in order to conform to the policies. To implement the policy-based spectrum management, radios must understand and reason policies from regulators, which can be realized by the Policy Reasoner (PR).

This paper is outlined as follows: After introducing the XG radio architecture and the reasoning process in Section II, we propose a novel way to represent spectrum policies using decision tree in Section III. Section IV gives the reasoning process of policies represented by decision tree and compares it with the traditional binary decision diagram method (BDD) way. The paper is concluded with a discussion on advantages and drawbacks of the two policy representation methods in Section V.

II. XG RADIO ARCHITECTURE AND POLICY REASONER

A. XG Architecture DARPA’s XG Program presents a radio architecture

which ensures that the transmission activities conform to spectrum policies. As Fig.1 indicates, an XG radio[2] contains four main components.

Figure 1. XG architecture

1) Sensors. XG radios need sensors to discover

available spectrum and transmission opportunities 2) Radio frequency (RF). The RF component transmits

and receives. 3) System Strategy Reasoner (SSR). The SSR controls

the radio’s transmissions. It builds transmission requests based on sensor data received from the sensors and its current strategies. Replies to its requests from the PR may affect strategy.

4) Policy Reasoner (PR). The PR accepts transmission requests from the SSR and checks policy conformance. The PR has all active policies loaded.

In general, the PR oversees every transmission activity by checking the transmission requests from SSR. When accepting requests from the SSR, the PR checks policy conformance. It replies with yes, no, or no with constraints to be satisfied. The PR has a policy base with all active policies, which can be modified and updated as required.

978-1-4673-0915-8/12/$31.00 ©2012 IEEE

2012 3rd International Conference on System Science, Engineering Design and Manufacturing Informatization

159

Page 2: A Novel Method to Represent Spectrum Management Policies

B. Binary Decision Diagram-based Reasoner for Spectrum Access Policies

Reference [3] gives a prototype for a device independent policy reasoner: the Binary decision diagram-based REasoner for Spectrum Access Policies (BRESAP), which enables dynamic spectrum access. Fig. 2 illustrates the architecture of BRESAP and its reasoning process.

Figure 2. BRESAP architecture

Each BRESAP component and its functionality are detailed in Reference [4].

III. REPRESENTING POLICIES BY DECISION TREE Any spectrum policy can be converted into a Boolean

function by the Policy Parser, and the Boolean function can be represented by a data structure inside the device. The decision tree is a data structure that can be used to represent a Boolean function[5]. In this section, we use the data structure of decision tree to represent policies, and compared it with the binary decision diagram (BDD) approach.

In what follows, we provide an example that explains how a policy is represented by BDD and decision tree.

Policy example. Consider the spectrum management policy: Allow transmission in [235MHz, 267MHz] or [315MHz, 322MHz] frequency bands, if the transmission power level is less than 25 dBm.

In this example, there are three policy constraints. We now map each policy constraint to an atomic Boolean expression, represented by variable xi:

x1 f [235MHz, 267MHz] x2 f [315MHz, 322MHz] x3 Pt 25dBm

Illogical reasoning results can be avoided by adding auxiliary rules into Boolean function. Consequently the example can be transformed into the function:

1 2

1 2 3

0 , if , else

x xp

x x x

��� � � ��

(1)

The Policy Converter component in BRESAP can transform the Boolean function into a BDD which can be

understood and reasoned by the device. The BDD created by the Policy Converter is as shown in Fig. 3.

Figure 3. BDD representation for example policy

The Boolean function can also be converted into a

decision tree as shown in Fig. 4.

Figure 4. BDD representation for example policy

Each path in the BDD and the decision tree from the root

node to the terminal node represents an assignment of values to the variables along the path. Any given transmission request corresponds to a variable assignment, which in turn corresponds to a path. As the example is a permissive policy, the 1-terminal node means Yes (Y), in the condition of which the transmission request is approved; while the 0-terminal node means Not Applicable (NA), in the condition of which the transmission request is not approved, and the SSR should adjust its transmission activity in accordance with the reply received from the PR.

After converting the example policy into BDD, there are only 6 nodes; while the decision tree needs 15 nodes, which occupy much more storage. Fig. 5 shows the number of nodes required for storing the policy as the number of variables increases.

Figure 5. Number of nodes required to represent example policy by

decision tree and BDD

160

Page 3: A Novel Method to Represent Spectrum Management Policies

Although policies represented by decision tree requires more nodes than those represented by BDD, the decision tree method is more convenient for policy reasoning when the transmission request is not approved, which will be discussed in next section.

IV. COMPARATION OF TWO METHODS IN POLICY REASONING

In this section, we show the difference caused by using different data structures to represent policy, when executing policy reasoning algorithm.

In what follows, the example in Section III is used to demonstrate the process for policy reasoning.

A. Request-valid Condition

In request-valid condition, the path generated by transmission request arrives at the 1-terminal node and the transmission will be approved.

We assume the transmission request is r1 = {(f, 318MHz� f �322MHz), (Pt, 20dBm)}. So the values to the variables can be assigned as in Table I.

TABLE I. VARIABLE ASSIGNMENT IN EXAMPLE POLICY FOR REQUEST 1

Variable Variable Assignment

Constraints Request Valuex1 235MHz � f �

267MHz 318MHz � f � 322MHz

0

x2 315MHz � f � 322MHz

318MHz � f � 322MHz

1

x3 Pt 25dBm Pt = 20dBm 1

Fig. 6 illustrates the path generated by request r1 in the

BDD and decision tree of the example policy. 0

1

x1

1

x2

x3

0

1

x2

Figure 6. Reasoning path generated by r1

In request-valid condition, the reasoning algorithms of BDD and decision tree both judge the value of the Boolean function and have the similar complexity.

B. Request-invalid Condition

In request-invalid condition, the path generated by transmission request arrive at the 0-terminal node, therefore the transmission will not be approved.

We assume the transmission request is r2 = {(f, 318MHz� f �322MHz), (Pt, 30dBm)}. So the values to the variables can be assigned as in Table II.

TABLE II. VARIABLE ASSIGNMENT IN EXAMPLE POLICY FOR REQUEST 2

Variable Variable Assignment

Constraints Request Valuex1 235MHz � f �

267MHz 318MHz � f �322MHz

0

x2 315MHz � f � 322MHz

318MHz � f �322MHz

1

x3 Pt 25dBm Pt = 30dBm 0

Fig. 7 illustrates the path generated by request r2 in the

BDD and decision tree of the example policy.

Figure 7. Reasoning path generated by r2

As the path leads to 0-terminal node, the request will not

be approved. Consequently, the PR will execute the FindPath algorithm to generate a path which achieves the 1-terminal node, and return the SSR a reply that consists of the parameters with value in conformance with policies.

To find a path achieving the 1-terminal in a BDD, the algorithm will traverse it from top to bottom and its complexity is O(n) where n is the number of variables. In a decision tree, however, once a 1-terminal node is determined, there will be a unique path lead to the start node. Therefore the design of algorithm will be simplified compared with the BDD condition. Fig. 8 shows the time required for executing the FindPath algorithm. The simulation was conducted on Intel Core 2 CPU 2.0 GHz machine with 2GB RAM.

Figure 8. Time required for executing FindPath algorithm

V. CONCLUSION In this paper we propose a novel method to represent

spectrum policies, and compare it with the traditional BDD method. From our analysis and the simulation results it can be concluded that policies represented by the decision tree can be reasoned more efficiently when the transmission request is not approved, compared with the BDD representation. However, the decision tree method has obvious drawbacks. This can be attributed to the fact that the nodes increased exponentially as the complexity of policy increased, which may occupy overmuch storage space.

161

Page 4: A Novel Method to Represent Spectrum Management Policies

It can be summarized that the decision tree method can be utilized where high reasoning speed is required and the number of policy variables is small.

ACKNOWLEDGMENT This work is supported by Natural Science Foundation of

China under Grant No. 61102092 and pre-research project of PLA Univ. of Sci.&Tech. under Grant No. 20110601

REFERENCES [1] F. Perich, “Policy-based network management for neXt Generation

spectrum access control,” in Proc. of 2007 IEEE DySPAN, Apr. 2007, pp. 496-506.

[2] D. Wilkins, G. Denker, “Policy-based cognitive radios”s, IEEE Wireless Communications Magazine, vol. 14, 2007, pp. 41-46.

[3] B. Bahrak, A. Deshpande, M. Whitaker, and J. Park, “BRESAP: A Policy Reasoner for Processing Spectrum Access Policies Represented by Binary Decision Diagrams”, in Proc. of 2010 IEEE DySPAN, Apr. 2010, pp 1-12.

[4] A. Deshpande, Policy Reasoning for Spectrum Agile Radios. Blacksburg, USA: Virginia Tech, 2010.

[5] GU Tian-long, XU Zhou-bo, Ordered binary decision diagram and its application. Beijing: Science Press, 2009, 19-25.

162