Contemporary Engineering Sciences, Vol. 10, 2017, no. 14, 673 - 689

HIKARI Ltd, www.m-hikari.com

https://doi.org/10.12988/ces.2017.7766

Evolutive Algorithm for Spectral Handoff

Prediction in Cognitive Wireless Networks

Diego Giral

Universidad Distrital Francisco José de Caldas, Bogotá, Colombia

Cesar Hernández

Universidad Distrital Francisco José de Caldas, Bogotá, Colombia

Hans Marquez

Universidad Distrital Francisco José de Caldas, Bogotá, Colombia

Copyright © 2017 Diego Giral, Cesar Hernández and Hans Marquez. This article is distributed

under the Creative Commons Attribution License, which permits unrestricted use, distribution, and

reproduction in any medium, provided the original work is properly cited.

Abstract

Context: Meta-heuristic algorithms have received considerable attention in the

solution of spectrum assignment problems since they have proven to be efficient

with lower computational load and complexity and fast execution time in complex

applications. Genetic algorithms are meta-heuristic optimization models inspired

in the process of genetics and evolution that can offer local or global optimal

solutions; their implementation in cognitive radio has led to using old techniques,

generate new models and formulate hybrid structures.

Objective: This work proposes a spectral handoff prediction using evolutive

algorithms and the multivariable approach of fuzzy feedback analytic hierarchic

process (FFAHP).

Method: To develop the proposed model, an availability matrix is generated

through a genetic algorithm which is used as training strategy; the evaluation of

the algorithm is performed for Wi-Fi technology with a 500-channel spectral

occupation of real data and a transmission time of 10 minutes.

Results: Results are summarized in five evaluation metrics: average number of

accumulated failed handoffs, average number of successful handoffs, average

bandwidth, cumulative average delay and cumulative average throughput.

674 Diego Giral, Cesar Hernández and Hans Marquez

Conclusion: According to the metrics obtained through simulations, the proposed

strategy is efficient and improves performance in terms of spectral mobility in

cognitive radio networks.

Keywords: cognitive radio networks, genetic algorithms, meta-heuristics, spectral

assignment, spectral handoff

1. Introduction

Cognitive radio (CR) is a wireless communication technology capable of

dynamically assigning the radio-electric spectrum. The concept was defined by

Mitola in 1999 [1] and various entities have generalized the definition. According

to the International Telecommunications Union (ITU), the CR is a radio or system

that detects and is aware of its surroundings and can be adjusted dynamically and

autonomously according to its radio operation parameters; according to the

National Telecommunications and Information Administration (NTIA), CR is a

radio or system that detects the electromagnetic operation environment and may

adjust dynamically and autonomously its radio operation parameters to modify the

system’s functionality which leads to maximizing performance, reducing

interference and facilitating interoperability [2].

In a wireless network, each user uses a segment or channel from the radio-electric

spectrum to transmit and receive information. In CR, secondary users (SU), which

are known for having a low priority under the licensed spectrum, can use a

channel dynamically and opportunely when it is not being used by primary users

(PU) and does not affect them [3]. The process in which the SU changes from one

frequency channel to another is known as spectral handoff [4]–[6].

During a spectral handoff it is inevitable that communication is interrupted

temporarily since it necessary to perform a spectral availability search process; the

spectral assignment techniques allow the SU to change to a new spectral band

with minimum degradation and reducing latency [7]–[10].

Meta-heuristic algorithms have been widely used to solve spectral assignment

issues, and have proven to be efficient in assignment problems due to their low

computational load and complexity and fast execution time in complex

applications [11], [12]. The most analyzed meta-heuristic methods in literature are

related to evolutive and swarm-based techniques [11], [13].

This work proposes a spectral handoff prediction based on an evolutive technique

with the purpose of generating an availability matrix through a genetic algorithm

and using it a training strategy. The algorithm is evaluated for Wi-Fi technology

with a 500-channel spectral occupation of real data and transmission time of 10

minutes. The smart selection of the frequency channels is achieved with a fuzzy

Evolutive algorithm for spectral handoff prediction 675

feedback analytic hierarchy process algorithm (FFAHP) [14]. The validation and

evaluation of the proposed model is carried out with the Spectrum Mobility

Analytical Tool software [15].

The article is structured as follows; the second section includes related work on

the subject at hand, the third section describes the general features of genetic

algorithms, the fourth section presents the adopted methodology, the fifth section

details the results and, finally, the sixth section presents a set of conclusions.

2. Related Work

In literature, there is a high volume of studies that analyze and gather strategies

for the assignment of the spectrum based on evolutive and swarm techniques.

In [16], various artificial intelligence techniques are presented that have been

implemented in CR such as: artificial neural networks (ANN), meta-heuristic

hidden Markov models (HMM), ontology-based systems (OBS) and case-based

system (CBS). The document analyzes factors such as response capacity,

complexity, safety, robustness and stability. Table 1 describes the features of

different meta-heuristic techniques.

Table 1. Features of some meta-heuristic techniques

Algorithm Advantages Disadvantages

Classification

techniques

Provides a globally optimal

solution for convex

optimization problems. Its

convergence properties are

well analyzed.

It could produce undesirable

solutions for ill-behaved

functions (branch-and-bound,

clustering) and techniques that

access global information and

computationally intensive.

Genetic

algorithms

Widely studied in wireless

applications.

The convergence has not been

fully researched and efficiency

depends on the selection of

parameters.

Simulated

annealing

The method has a good

classification which is easily

scalable. It can predict based

on experience.

The convergence rate can be

slow and it only converges to

global optimal value a time

approaches infinity for a finite

search space.

676 Diego Giral, Cesar Hernández and Hans Marquez

Table 2. (Continued): Features of some meta-heuristic techniques

Ant colony

It easily adapts to changes in

real-time.

It is inferior to simulated

annealing in local searches.

Taboo search

Asymptotically converges to

an optimal solution with a

probability of 1. Can be easily

implemented.

Efficiency relies on an

appropriate selection of

parameters.

Source: Adapted from [16].

2.1. Meta-heuristic algorithms based on swarms

Swarm techniques are intelligence artificial methods based on the study of

collective behavior present of Nature systems which are applied to problems in

diverse areas and can offer global or local optimal solutions. These strategies have

been well received in CR and some of documents are now described where swarm

techniques have been applied in CR.

In [13], a strategy is proposed to minimize the required frequency spectrum for a

given quality level of reception through the network. The author proposes a meta-

heuristic approximate non-deterministic tree search (ANTS) method that includes

a model derived from ant colonies.

In [11] an algorithm also inspired on ant colonies is defined to solve spectrum

assignment issues obtained from coloration problems of NP-hard graphs. In [4], a

solution to the cognitive radio spectrum problem is developed taken from the

meta-heuristic Chemical Reaction Optimization (CRO) approach.

In [17], an optimal solution for Quality of Service (QoS) parameters in cognitive

radio is presented. The paper compares the evolutive optimization techniques

Annealing (SA) and Genetic Algorithm (GA). Additionally, the led research

proposes a swarm intelligence technique called Spider Monkey Optimization

(SMO) that is inspired on the spider monkeys’ foraging behavior. The method is

used to optimize the QoS parameters in terms of minimum energy consumption,

minimum bit error rate (BER), maximum performance, minimum interference and

maximum spectral efficiency.

2.2. Evolutive meta-heuristic algorithms

Genetic algorithms are optimization models inspired in genetic and evolution

processes, and can offer global or local optimal solutions like swarm-based

techniques do. Their implementation in CR has facilitated the use of previously

defined techniques, generate new models and formulate hybrid structures.

Evolutive algorithm for spectral handoff prediction 677

In [18], it is analyzed how evolutive calculation methods such as the genetic

algorithm (GA) are capable of designing new types of systems. A binary-coded

genetic algorithm is used to optimize the geometry of an antenna to achieve the

largest possible bandwidths with a minimum number of commuters. This has the

purpose of increasing the bandwidth performance from a single CR antenna.

In [19], a genetic algorithm is used as a strategy for optimal assignment of the

radio-electric spectrum, guaranteeing a satisfactory quality of service. The

document highlights the ability of handling multiple targets from genetic

algorithms as a heuristic strategy.

In [20], an optimal power assignment scheme is suggested using a genetic

algorithm for a MIMO (multiple input multiple output) system in cognitive radio.

The goal is to maximize the performance of the secondary user under interference

restrictions in a system model with multiple pairs of SUs coexisting with multiple

pairs of PUs.

3. Genetic Algorithms

3.1. Definition

Genetic algorithms are optimization models based on genetic and evolutive

processes. A simple model is composed by an initial population of individuals and

a set of operations that interact over that population to obtain new generations of

individuals. As seen in the optimization model described in equation (1), simple

genetic algorithms are applied to the solution of optimization problems with

continuous parameters [21]–[24].

1 2min ( , ,... )

, , , 1,...,

k k kn

ki i i i i i

f x x x

x l u l u n (1)

The population is constituted by a set of individuals represented through an

equivalent number in binary. This binary representation is called Chromosome

and each bit of the chromosome is called gen. In general, a genetic algorithm is

characterized by five definitions or genetic equivalents described in Table 3. They

are graphically represented for a specific population in Figure 1.

Table 3. Genetic equivalents

Genetic parameter Description

Allele Each different state that gens can present in the same position.

Gen The value of an allele in an array.

Chromosome It is an array-formed collection of genes.

Position It is the place occupied by gen in the chromosome.

Index It is the place of the individual within a population.

Source: [22]

678 Diego Giral, Cesar Hernández and Hans Marquez

Figure 1. Genetic equivalents for a specific population

Source: Adapted from [22]

3.2. Generations

Generations or iterations from a systematic point of view allow the initial

population to evolve into a new population with a better genetic code. This

process is performed through three operators on the current population [21].

Selection operator

Cross operator

Mutation operator

3.2.1. Selection

It is the operation with the highest impact and it is in charge of transmitting the

genetic code from the most fitted individuals to future generations. A new

population called intermediate is generated from the current population and it

must have the same initial size. However, genetic structure must be diverse so it

does not only take the best codes but it also transmits a number of codes with

lower performance. Algorithmically, there are diverse selection methods [22]–

[25]:

Direct selection: The population is ordered from the most suitable

individuals to the ones with lower performances and the cross is

performed from the inferior index up to the superior index, allowing the

ill-adapted individuals to interact with individuals that are better adapted.

Tourney selection: n individuals are randomly selected from the

population and the ones with the highest fitness are chosen among them.

The process is repeated until a new population is formed.

Proportional or roulette selection: Each individual has a proportional

cross probability. A random array is generated with ascending

Evolutive algorithm for spectral handoff prediction 679

probabilities while keeping the same length as the population. Then, the

selection probability of the individual is compared with the ascending

array’s probability. If the value of the random array falls in the

individual’s interval, then the individual is chosen.

3.2.2. Cross or recombination

The crossover or crossing operator selects two chromosomes called parents and

assigns them a crossing point. The crossing is then performed between the

selected individuals with the purpose of having new combinations that are called

children. A sexed operator is known as being allowed to exchange genetic

material to produce descendants. The central idea is to match parents with

different genetic codes.

Figure 2 shows the matching carried out by the crossing operator for two

randomly selected individuals [26], [27].

Figure 2. Crossing operator

3.2.3. Mutation

The mutation operator for simple genetic algorithms with low mutation

percentages is not relevant. The mutation is performed by randomly modifying the

genes of a population by a certain percentage. Small mutation rates guarantee that

new populations do not vary in comparison to the genetic code of previous

populations [27], [28].

4. Methodology

In this section, the description of the proposed model for spectral handoff

evaluation is presented. Genetic algorithms such as evolutive techniques are used

to present the general description of the model, the input data, the operators

implemented for every generation (number of iterations), the channel selection

methodology and the evaluation of metrics.

680 Diego Giral, Cesar Hernández and Hans Marquez

4.1. Proposed model

Figure 3 presents the block diagram’s proposed model, the Occupation and

Processing blocks are associated to the spectral occupation database for the Wi-Fi

network and to the respective processing that allows the definition of the

availability matrix for each channel; the proposed model has five sub-blocks and

is assessed with the Spectrum Mobility Analytical Tool software.

The model corresponds to the blocks enclosed by the discontinued line square,

formed by five algorithms, the Selection algorithm and Channel selection blocks

enable the selection of channel (number of columns) with two techniques: through

the multivariable FFAHP or randomly. The parameterization variable corresponds

to the number of channels (between 10 and 460). The blocks Initial population,

Population operators and Final population construct the training matrix with

genetic algorithms. First, a random initial population is established and its binary

representation is adjusted according to the number of channels. The operations of

direct selection, crossing and mutation (small mutation rate) are performed to

create the final population equivalent to the training data.

Figure 3. Proposed model

4.2. Input data

The Spectrum Mobility Analytical Tool software takes the information

availability, bandwidth and channel ranking for the FFAHP handoff model. The

database corresponds to the power information for high traffic in a Wi-Fi network

in the city Bogotá, Colombia. The time steps of the validation availability matrix

are adjusted according to the software’s database.

Figure 4 presents the input and output data, the input information belongs to the

software parameterization which requires the definition of the following variables:

threshold, noise floor, bandwidth and multichannel [15].

Evolutive algorithm for spectral handoff prediction 681

Figure 4. Input and output data

4.3. Genetic algorithm

A genetic algorithm is designed for the training matrix. The model establishes an

initial random population which is verified by a transition function (transition

matrix). The goal is to assure that the population has coherent values. The

selection, crossing and mutation create the final population equivalent to the

training data. The number of generations (iterations) is adjusted under

performance parameters such as simulation which is a variable that is

parameterized with a trial and error criterion. Figure 5 presents the flow diagram

for the genetic algorithm.

4.4. Selection of studied channels

Figure 6 shows the block diagram for the selection of the number of channels

(between 10 and 460). The selection involves two techniques: the FFAHP handoff

model and a second method based on random selections that will allow the

comparison of results. The number of channels variable adjusts the channels for

the bandwidth, the availability validation matrix and the input condition for the

initial population of the genetic algorithm.

4.5. Evaluation of spectral handoff

The evaluation of spectral handoffs is performed through the training and

validation matrices. The Spectrum Mobility Analytical Tool software is used in

Custom mode. The results correspond to the figures of bandwidth, handoff, failed

handoffs, delay and throughput. Although it is not part of the evaluation, the

following information of the genetic algorithm is collected: number of

generations, initial population, mutation percentage, etc. Figure 7 shows the

evaluation of diagram.

682 Diego Giral, Cesar Hernández and Hans Marquez

Figure 5. Flow diagram for the genetic algorithm

Figure 6. Selection of studied channels, validation matrix and training matrix

Evolutive algorithm for spectral handoff prediction 683

Figure 7. Evaluation of the algorithm

5. Results

The evaluation of the spectral handoff algorithm using genetic algorithm for a

high traffic Wi-Fi network is carried out with five metrics: number of cumulative

failed handoffs, number of cumulative total handoffs, bandwidth average, the

cumulative average delay and cumulative average throughput. The results

obtained for each metric with a transmission time of 10 minutes are shown in

Figures 8-12.

Figure 8 presents the number of cumulative failed handoffs, the difference

between the two strategies is 131 failed handoffs, where the FFAHP algorithm has

the lowest number of failed handoffs (114).

Figure 8. Number of cumulative failed handoffs for a Wi-Fi network with high

traffic

684 Diego Giral, Cesar Hernández and Hans Marquez

Figure 9 presents the number of cumulative of handoffs. The results present high

values in the 10 minutes of transmission for the random distribution algorithm

with a final average value of 1044 and a total of 685 for the FFAHP.

Figure 9. Number of cumulative total handoffs for a Wi-Fi network with high

traffic

Figure 10. Average bandwidth for a Wi-Fi network with high traffic

Figure 10 presents the average bandwidth and there is no significant difference

between the two selection strategies. However, the average bandwidth for FFAHP

Evolutive algorithm for spectral handoff prediction 685

is higher than for the random distribution during the 10 minutes of transmission.

The highest value in both cases is obtained at 5 minutes and the average

difference was 79.33 kHz.

Figure 11. Average cumulative delay for a Wi-Fi network with high traffic

Figure 12. Average cumulative throughput for a Wi-Fi network with high traffic

Figure 11 presents the average delay for each algorithm during the 10 minutes of

transmission. The delays correspond to the difference between the total handoffs

686 Diego Giral, Cesar Hernández and Hans Marquez

and failed handoffs. In general, the two strategies present an acceptable behavior

with a lineal growth under 200 seconds; the lowest and the highest differences

between the techniques is obtained in the extreme values of the data, 12.07 s for

1000 kB and 60.3 s for 10000 kB.

Figure 12 describes the average throughput with a standard 16QAM modulation;

the cumulative average throughput is superior for the FFAHP distribution with

4575 kbps. For the random distribution its value is 3594 kbps.

Table 4 presents the results of the two selection strategies using as training matrix

the population generated with the genetic algorithm.

Table 4. Evaluation metrics

Failed

handoffs

Total

handoffs

Average

bandwidth

Average

delay

Average

throughput

FFAHP 114 685 771.05 139.5 4575

Random 245 1044 691.72 199.8 3594

6. Conclusions

Meta-heuristic algorithms have been widely considered to solve spectrum

assignation problems and have proven to be efficient in assignments problems.

They are optimization models that can offer local or global optimal solutions.

The spectral handoff prediction model using genetic algorithms, described in this

article gave conclusive results for both channel selection methodologies.

However, in contrast to the random strategy, the FFAHP algorithm assures

repeatability in the results. According to the metrics obtained through simulations,

the proposed method improves the performance of spectral mobility in cognitive

radio networks.

Referencias

[1] J. Mitola and G. Q. Maguire Jr., Cognitive radio: making software radios

more personal, IEEE Pers. Commun., 6 (1999), no. 4, 13–18.

https://doi.org/10.1109/98.788210

[2] L. F. Pedraza, C. Hernández, K. Galeano, E. Rodríguez-Colina and I. P. Páez,

Ocupación espectral y modelo de radio cognitiva para Bogotá, Primera.

Bogotá: Editorial UD, 2016.

[3] C. Hernández, H. Márquez, and L. F. Pedraza, Modelo de asignación

multicanal con equidad para la movilidad espectral en redes de radio

cognitiva. Bogotá: Editorial UD, 2017.

Evolutive algorithm for spectral handoff prediction 687

[4] A. Y. S. Lam, V. O. K. Li, and J. J. Q. Yu, Power-Controlled Cognitive

Radio Spectrum Allocation with Chemical Reaction Optimization, IEEE

Transactions on Wireless Communications, 12 (2013), no. 7, 3180–3190.

https://doi.org/10.1109/twc.2013.061713.120255

[5] I. F. Akyildiz, W.-Y. Lee, M. C. Vuran, and S. Mohanty, A survey on

spectrum management in cognitive radio networks, IEEE Commun. Mag., 46

(2008), no. 4, 40–48. https://doi.org/10.1109/mcom.2008.4481339

[6] I. F. Akyildiz, L. Won-Yeol, M. C. Vuran and S. Mohanty, NeXt

generation/dynamic spectrum access/cognitive radio wireless networks: A

survey, Comput. Networks, 50 (2006), no. 13, 2127–2159.

https://doi.org/10.1016/j.comnet.2006.05.001

[7] D. A. López, E. R. Trujillo, and O. E. Gualdron, Elementos Fundamentales

que Componen la Radio Cognitiva y Asignación de Bandas Espectrales, Inf.

Tecnológica, 26 (2015), no. 1, 23–40.

https://doi.org/10.4067/s0718-07642015000100004

[8] I. Christian, S. Moh, I. Chung and J. Lee, Spectrum mobility in cognitive

radio networks, IEEE Commun. Mag., 50 (2012), no. 6, 114–121.

https://doi.org/10.1109/mcom.2012.6211495

[9] I. F. Akyildiz, W.-Y. Lee and K. R. Chowdhury, CRAHNs: Cognitive radio

ad hoc networks, Ad Hoc Networks, 7 (2009), no. 5, 810–836.

https://doi.org/10.1016/j.adhoc.2009.01.001

[10] C. Hernández, I. Páez and D. Giral, Modelo adaptativo multivariable de

handoff espectral para incrementar el desempeño en redes móviles de radio

cognitiva, Primera Ed. Bogotá: Editorial UD, 2017.

[11] H. Salehinejad, S. Talebi and F. Pouladi, A metaheuristic approach to

spectrum assignment for opportunistic spectrum access, 2010 17th

International Conference on Telecommunications, (2010), 234–238.

https://doi.org/10.1109/ictel.2010.5478746

[12] J. N. J. Moon, L. A. Hughes and D. H. Smith, Assignment of frequency lists

in frequency hopping networks, IEEE Transactions on Vehicular Technology,

54 (2005), no. 3, 1147–1159. https://doi.org/10.1109/tvt.2005.844659

[13] M. Dorigo and L. M. Gambardella, Ant colony system: a cooperative learning

approach to the traveling salesman problem, IEEE Transactions on

Evolutionary Computation, 1 (1997), no. 1, 53–66.

https://doi.org/10.1109/4235.585892

688 Diego Giral, Cesar Hernández and Hans Marquez

[14] C. Hernández, L. F. Pedraza and E. Rodriguez-Colina, Fuzzy Feedback

Algorithm for the Spectral Handoff in Cognitive Radio Networks, Rev. Fac.

Ing. la Univ. Antioquia, 81 (2016), 47-62.

https://doi.org/10.17533/udea.redin.n81a05

[15] C. Hernández and D. Giral, Spectrum Mobility Analytical Tool for Cognitive

Wireless Networks, Int. J. Appl. Eng. Res., 10 (2015), no. 21, 42265–42274.

[16] A. He, K. K. Bae, T. R. Newman, J. Gaeddert, K. Kim, R. Menon, L.

Morales-Tirado, J. Neel, Y. Zhao, J. H. Reed and W. H. Tranter, A survey of

artificial intelligence for cognitive radios, IEEE Trans. Veh. Technol., 59

(2010), no. 4, 1578–1592. https://doi.org/10.1109/tvt.2010.2043968

[17] V. Rana and P. S. Mundra, Simulation of QoS parameters in cognitive radio

system using SMO algorithm, 2017 International Conference on Inventive

Communication and Computational Technologies (ICICCT), (2017), 292–

297. https://doi.org/10.1109/icicct.2017.7975206

[18] Z. Aizaz and P. Sinha, A survey of cognitive radio reconfigurable antenna

design and proposed design using genetic algorithm, 2016 IEEE Students’

Conference on Electrical, Electronics and Computer Science (SCEECS),

(2016), 1–6. https://doi.org/10.1109/sceecs.2016.7509263

[19] Y. El Morabit, F. Mrabti and E. H. Abarkan, Spectrum allocation using

genetic algorithm in cognitive radio networks, 2015 Third International

Workshop on RFID And Adaptive Wireless Sensor Networks (RAWSN),

(2015), 90–93. https://doi.org/10.1109/rawsn.2015.7173287

[20] A. M. Benaya, A. A. Rosas and M. Shokair, Maximization of minimal

throughput using genetic algorithm in MIMO underlay cognitive radio

networks, 2016 33rd National Radio Science Conference (NRSC), (2016),

141–148. https://doi.org/10.1109/nrsc.2016.7450847

[21] J. Khazaii, Genetic Algorithm Optimization, Chapter in Advanced Decision

Making for HVAC Engineers, Springer, 2016, 87–97.

https://doi.org/10.1007/978-3-319-33328-1_10

[22] N. Hernández Romero, J. Medina Martín and J. C. Seck Tuoh Mora,

Introducción a Matlab para Resolver Problemas de Ingeniería Aplicando

Algoritmos Genéticos, Primera Ed. México, 2012.

[23] R. L. Haupt and D. H. Werner, Anatomy of a Genetic Algorithm, Chapter in

Genetic Algorithms in Electromagnetics, Wiley-IEEE Press, 2007, 29–43.

https://doi.org/10.1002/9780470106280.ch2

Evolutive algorithm for spectral handoff prediction 689

[24] M. Eremia, C.-C. Liu and A.-A. Edris, Genetic Algorithms, Chapter in

Advanced Solutions in Power Systems:HVDC, FACTS, and Artificial

Intelligence, Wiley-IEEE Press, 2016, 845-902.

https://doi.org/10.1002/9781119175391.ch19

[25] P. E. Valencia, Optimización mediante algoritmos genéticos, Anales del

Instituto de Ingenieros de Chile, 109 (1997), no. 2, 83–92.

[26] M. R. Aldás and M. J. Flores, Modelo origen destino para estimar el flujo de

tráfico usando algoritmos genéticos, Maskana, 65 no. Supl., (2016).

[27] F. Herrera, M. Lozano and J. L. Verdegay, Algoritmos genéticos:

Fundamentos, extensiones y aplicaciones, Arbor: Ciencia, Pensamiento y

Cultura, 152 (1995), no. 597, 9-40.

[28] M. Wall, GALib, A C++ Libr. Genetic Algorithm Components, Massachusetts

Inst. Tech., 2000.

Received: August 9, 2017; Published: September 6, 2017