Proceedings of the 2006 IEEE International Conference on Mechatronics and Automation June 25 - 28, 2006, Luoyang, China

Biologically Inspired Approaches for Wireless Sensor Networks Hongliang Ren, Max Q.-H. Meng Department of Electronic Engineering The Chinese University of Hong Kong Hong Kong, China { hlren, max}@ee.cuhk.edu.hk Abstract - Biological systems present remarkable adaptation, reliability and robustness in the dynamic and even hostile environments. Also, the networks that govern the flow of resources and information are ubiquitous in nature. Most of them are performed by individuals in a distributed and selforganized way. These mechanisms provide useful paradigms for designing the wireless sensor networks, where the individual nodes operate without central control. This paper investigates the existing sensor network design paradigms that inspired from the biological system. The investigation ranges from cell level mechanisms to collective activities performed by social insects, in an effort to apply these inspirations to the sensor networks. This paper involves the problems in the design of wireless sensor networks, including routing, congestion control, medium access control, and clustering problem.

mechanisms of indirect interactions via environment such as pheromone-laying and pheromone-following. From a biological perspective, take V-Formations of geese flocking as another example, the benefits of flocking are geese flying in Vs can extend their energy range by over 70% because each bird rides on the vortex cast off by the wing-tip of the one in front. Similarly, benefits of shoaling are to resist the ocean flow turbulence so as to save energy. Biological study [1] has shown their activities are coordinated by selforganization, explicit peer-to-peer local communication, or implicit communication through the environment. This paper investigates the algorithms inspired from robust biological system, and draws algorithmic profiles when these algorithms are applied to wireless sensor networks. The rest of this paper is organized as follows. Section II provides a brief introduction to the Ant Colony Optimization (ACO) inspired from the collective shortest path searching behaviours in ant colonies and investigates various flavors of ant-based routing algorithms for sensor networks. Section III reviews the Genetic Algorithm (GA) that simulates the process of genetic selection and natural elimination in biological evolution and discusses its application in sensor networks. Section IV outlines Particle Swarm Optimization (PSO) motivated by foraging behaviour of bird flocking and its application in clustering problems. Section V describes the reaction-diffusion mechanisms in the formation of morphogenesis and its inspiration for wireless medium access methods. Section VI introduces other biologically inspired approaches: a clustering algorithm that motivated by the bacterial signalling process called quorum sensing in VI (A); the artificial immune system and cell biology analogy for networks security problems in VI (B); and the grouping strategies for the mobile sensor nodes inspired from swarm behaviours in VI (C). Section VII concludes the paper.

Index Terms - WSN, Network Protocol, Swarm Intelligence, Biologically Inspired

I. INTRODUCTION Many desirable characteristics, such as scalability and robustness are exhibited in biological system despite of individual simplicity, especially in social insects. Emergence, kind of collective phenomena, is defined as the global complex structure emerges from a set of simple individual rules accomplished according to limited local information. The global complex structures are typically flexible to adapt to a new environment and robust to resist individual failure. For individuals, their behavioural repertoire is limited and their cognitive systems are not powerful to acquire global knowledge. Collective behaviours of the nature species (e.g., insects) provide a natural model for distributed problem solving, without externa guidance or central control. The key is to investigate how the collective complexity comes out of individual simplicity. Study has shown that Self-Organization and Stigmergy are two key ideas in Swarm Systems [1]. The basic ingredients of Self-Organization include Positive Feedback, Negative Feedback, Fluctuation Amplification, and Multiple Interactions. Take ant colony as an example to illustrate these ingredients: the action of depositing pheromone is a positive feedback mechanism to recruit more ants, so as to more pheromones are deposited on the shorter path; however, the evaporation of pheromone is a negative feedback to reduce the pheromone level; in this way the shortest paths to the food source are found. Moreover, Stigmergy is defined as the

1-4244-0466-5/06/$20.00 ©2006 IEEE

II. ANT COLONY OPTIMIZATION META-HEURISTIC ALGORITHM A. Ant Colony Optimization Meta-Heuristic Algorithm Ant colony, a typical social insect, exhibit complex collective behaviours such as foraging and cemetery construction, although the individual ant behaviours are very simple. This has motivated a Meta heuristic optimization algorithm called Ant Colony Optimization [1] by considering the foraging problem that the ants start out searching for food from their nests. The ants deposit a chemical substance called

762

pheromone as they move between the food source and nest. At first, the ants randomly select the paths to search for food. After a transition period, ants are gradually attracted by the optimal path with stronger pheromone scent. If the path is truly optimal, they will deposit more pheromone on this path in return. At the same time, pheromone evaporates with time elapsed, so that the pheromone will finally be exhausted if no more deposition is available. In this way, the longer paths will have much fewer pheromones than the shorter ones. This trail-laying trail-following behaviour is modelled in [1] mainly by considering pheromone update rule, probabilistic transition rule and heuristic desirability. Probabilistic transition rule determine the probability of choosing an edge as an ant’s path by considering the existing pheromone level and its heuristic desirability, as shown in Equation (1). Pheromone updating rule specifies how the concentration of pheromone on the edges are modified by both depositing and evaporating procedure. This process is described in Equation (2), where the first term of the right hand side is caused by pheromone decay and the second term is the increment of pheromone depositing. In this shortest path problem, besides pheromone level, desirability is another criterion defined as the inverse of the edge length, by which the ants choose a shorter path.

pijk (t )

Mij (t ) K ij k

Ji

Į, ȕ

B. Review of ACO based Routing Algorithm Recently, some ACO based routing algorithms dubbed “ant-based [2;3]” or “ant-like [4]” are emerging for circuitswitch voice network [3], wired networks [5] and Mobile Adhoc and Sensor Networks [6]. Although these different biologically inspired routing algorithms are designed in various ways to meet respective objectives, they share the common mechanisms of ACO, such as pheromone updating rules and transition rules. Typically, the pheromone level represents the goodness of the route, which is commented by the ants. Different design consideration or metrics can be reflected in the pheromone updating and transition rules. The table III lists recent ACO based routing algorithms with different design objectives and focuses. Some of them are reviewed in this paper to show how the ACO is applied to the routing problem, and how the pheromone updating and probabilistic transition rules are formulated. For completeness, we will first start from the routing problem in wired network and then draw focus on sensor networks. 1) Ant-Based Control Ant-Based Control (ABC) [2] is a routing algorithm designed for circuit-switched telephone networks. Each node maintains a routing table where each row is a destination and each column a neighbour. Each entry is the next hop probabilities for each destination. Ant packets traverse the network, updating routing tables at every node depending on the age of the arriving packet. The effect of an ant’s arrival at a node is decreased with the age of the ant. The reciprocal of the ant age can be deemed pheromone although there is no explicit pheromone formulation. So if the ant arrives at a node, the normalized probabilistic transition rule for the new cell entry is formulated as: p ( pold  'p) /(1  'p) ,

TABLE I NOTIONS IN THIS PAPER Transition rule, the probability for ant k to go from i to j while building its tth tour Pheromone trail, on the edge connects i to j Heuristic desirability, for choosing j when in i, depend on problems, e.g. for TSP, it is the inverse of distance between i and j Set of nodes that ant still has to visit when it is on node i

'M ijk (t )

Tunable parameters that control the relative weight of trail intensity and desirability. Decay coefficient of pheromone The quantity of pheromone laid on edge (i, j) by ant k

m

The number of ants

ȡ

The model can be applied to many combinatorial problems when the problem can be appropriately represented and the above mentioned three factors are properly defined. Take the Travelling Salesman Problem (TSP) as an example [1] in Table II to show how this model are applied. Researchers have shown that ACO is able to achieve ideal results in other combinatorial optimization problems such as quadratic assignments problem, graph colouring, jobshop scheduling [1]. More importantly, ACO shows good performance in solving the stochastic time-varying problems such as routing in networks because of its flexibility and decentralization nature. ACO presented many desirable features in solving dynamic and distributed routing problem because of their similarities between ants foraging and routing. In the following Part B, we will review several variants of ACO that has been used to solve network problems.

Probabilistic transition rule:

pijk (t )

[M ij (t )]D ˜ [K ij ] E

¦

lJ ik

Pheromone updating rule:

(1)

[M il (t )]D ˜ [K il ] E

Mij (t ) m (1  U ) ˜ Mij (t )  'Mij (t ) m

(2)

'M ij (t ) m ¦k 1 'M (t ) Heuristic desirability:

Problem representation:

Heuristic desirability: Probabilistic transition rule: Pheromone updating rule: Additional constraint:

Ș

k ij

(3)

TABLE II SOLVING TSP USING ACO TSP problem is represented as a graph G= (V, E) where the set of vertices V is the citied and the set of edges E the paths. Ants find the solutions by visiting the nodes. Ș = (1/Distance) Same as Equation (1), the probability of determining the path is a function of pheromone and desirability. Same as Equation (2), pheromone evaporates on all edges and new pheromone is deposited by the ants on visited edges. Do not revisit the same node.

where 'p A / age  B , A and B are design parameters. This simplified ACO outlines its mechanism on solving routing problem and was shown good performance in the implementation of British Telecom [2].

763

TABLE III ACO BASED ROUTING SCHEMES Authors Algorithm Name Objective Schoonderwoerd [2] ABC (Ant-Based Telephone Control) Network Di Caro [5] AntNet Wired Network Heissenbuttle [3] MABR (Mobile Ants Topology Based Routing) Abstracting Gunes [6] ARA (Ant Colony Reduce Based Routing) Overhead Liu Zhenyu [7] EARA-CG Congestion (Emergent Ad hoc control routing Routing Algorithm ) Liu Zhenyu [8] EARA-QoS QoS Provision (Emergent Ad hoc Routing Routing Algorithm ) ZhengXiangquan ADRA (Ant-based Congestion and [9] Distributed Routing Convergence Algorithm) M. T. Islam [10] All-Pair ACO Trip time Routing F. Ducatelle [11] AntHocNet Multipath Routing Fair energy O. Hussein [12] ARAMA (Ant routing Algorithm for usage MANET) Srisathapornphat ABEC+AODV (AntEnergy [13] Based Energy Conservation Conservation) Shivanajay[14] Ant-AODV Reduce delay of starting phase D. Camara [4] GPSAL(GPL/AntGPS localization like routing) aided M. Roth [15] Termite (Ad-Hoc Fast Routing with convergence Stigmergy) Pirzada [16] Trusted routing Security

used to maintain the path. In ARA, the pheromone updating rule resembles the rule formulated in Equation (2) in the ACO algorithm. The transition probabilities for a node to choose next hop is derived from Equation (1) without considering the heuristic desirability and set Į to be 1: [M ij (t )]D (4) p ijk (t ) ¦lJ k [M il (t )]D

Year 1996 1998 2003 2002

i

4) EARA-CG and EARA-QoS EARA-CG (Emergent Ad hoc Routing Algorithm) [7] is an ACO based routing scheme for WSN. It adopts the crosslayer design approach by considering the local network congestion reside in the MAC layer. Two congestion metrics are involved. First, average MAC layer utilization was defined as the ratio of average busy time of the medium to the duration of medium time, and was used in the route discovery period to determine the probability of broadcasting route request packets according to the MAC layer congestion status. The other one was transmission queue length in the nodes and it also served as the heuristic value Ș in Equation (1). In EARA-CG, each node maintains two tables, probabilistic routing table and pheromone table. A probabilistic routing table has the entry of transition probability as shown in Equation (1) and a pheromone table tracks the amount of pheromone on each neighbour link as shown in Equation (2). The operation of the algorithm consists of three phases, route discovery, route reinforcement and route maintenance, which is similar to ARA. EARA-QoS [8], as its name implies, is an extension of EARA with QoS provision. This ACO based routing scheme integrates QoS consideration by adding the link delay Di,j to Equation (1): [M ij (t )]D ˜ [K ij ] E [ Dij ] r p ijk (t ) (5) ¦lJ [M il (t )]D ˜ [K il ]E [ Dij ]r

2005

2005

2004

2003 2005 2003

2003

2002 2000 2003

2004

2) AntNet AntNet [5] solves routing problems in wired networks, in which each node maintains probabilistic entries in the routing table like ABC. In actual network operation, the next hop with the highest probability is always chosen. Each node periodically sends an exploration agent called forward ant to randomly selected destinations. The ants record their arrival time and the node identity in a stack when every node is visited. An ant reaching its destination is converted to a backward ant and it returns to the source node following the path in reverse. Each intermediate node updates its routing tables according to the information extracted from the backward ant. For example, the probability of the route with shorter round trip time will be increased.

k i

where r is the tunable parameters that control the weight of link delay. 5) ADRA The ADRA (Ant-based Distributed Route for Ad hoc network) [9] algorithm improves the convergence rate of antbased routing by introducing another two ants: the shortcut reinforce ant dubbed enforce-ant, and the congestion repression ants dubbed anti-ant. When an intermediate node’s load exceed its predefined congestion threshold, it will send a congestion-control anti-ant to its upstream neighbour nodes to modify their probability route tables. Upon receiving enforceant or anti-ant, the node updates the routing table respectively to reinforce the good route and avoid the congested route.

3) Ant Colony Based Routing Algorithm This algorithm (ARA) [6] introduces a ACO based routing scheme for Mobile Ad hoc Networks (MANET). Its route discovery mechanisms are similar to the dynamic on demand routing scheme Dynamic Source Routing (DSR). Route discovery is performed by flooding forward ants to the destination while establishing reverse links to the source. Routes maintenance in ARA does not need any special packets. Once the pheromone tracks have been established for the source and destination nodes, subsequent data packets are

6) ARAMA ARAMA (Ant Routing Algorithm for Mobile Ad-hoc Networks) [12] uses the same two ants as AntNet. Forward ants are used to collect information about the network and backward ants are used to update the routing information in the nodes. It emphasizes the fair energy usage across the network by including the local battery information of the node into the ant packets.

764

7) Termite Routing Termite routing [15] operate in the same way as ARA, but using a different pheromone decay formulation is used. Like real pheromone, the artificial pheromone concentration decreases with time to prevent a fast convergence of pheromone on the edges. In the termite routing, it use exponential decay instead of Equation (2): M ij (t ) e  t ˜ M ij (t )  'M ij (t ) , where t is the time elapsed.

information on the network state is delivered for updating the local routing tables. Swarms packets were used to solve unreliable transport protocol problems, such as UDP in [18]. III. GENETIC ALGORITHM (GA) Genetic Algorithm (GA) is a computational mechanism that simulates the process of genetic selection and natural elimination in biological evolution [19]. Through the genetic operations (e.g., reproduction, crossover and mutation), it is able to find out a better solution in the form of new generations that have inherited advantages or survivalenhancing traits from previous generations. Unlike traditional search algorithms in Artificial Intelligence (AI), a generic algorithm is able to automatically acquire and accumulate implicit knowledge about search space. GA has been utilized as an optimization technique to optimize sensor node distribution or layout [20], to optimize clustering [21] and to compute mobile agent routes [19] recently. Next, we will outline the algorithms in optimal node distribution. With some mobile nodes and most of random scattered fixed nodes, node layout optimization algorithm aims to enhance sensor coverage and lifetime of the network by topology reconfiguration. In [22], initial populations in GA are set up by using coding mapping: Popi={gi1,gi2,…gij,… giN} where gij is the gene represent of node coordinate of the No.j gene in No.i individual and every chromosome of individual is composed of N genes. Then Circle Crossover between parents brings new offspring individuals: (6) g Chd ( g Pop  g Pop ) / 2  \ g ,

8) Trusted Routing Trusted Routing [16] draws more attention on secure routing by making use of ACO, and a pragmatic mechanism was presented for establishing trust in Ad hoc network to secure the routing process against malicious behaviours. The trustworthy routes are found based on sharing and maintaining trust levels represented by pheromone levels which are spread by ants. Once a malicious node starts destructive activity, its corresponding pheromone levels that are being maintained by other nodes start depleting. In trusted routing, the trust levels, described by pheromones, are computed based upon the number and accuracy of packets that have traversed a particular route. From the survey of existing ACO based routing schemes, it can be concluded that three factors should be well defined to meet particular objectives. Firstly, probabilistic transition rules is the expression to emphasize particular parameters while random selecting the routes. Secondly, the updating operation of pheromone level, including the definition of pheromone level and its increase or decrease operations, should be clearly specified. Thirdly, the operation and implementation of ant packets must be examined, e.g., how the ant packets are constructed and what information should be carried with it. Normally, a large ant packet will increase the communication overhead, so the tradeoffs should be examined in actual implementation.

ik

where

\ g ~ N (0, g Pop

ik

( i 1 ) k

( i 1 ) k

obeys

 g Popik )

Gaussian

distribution. Afterward, new contents are brought to the population by Mutation operation: g Mutation g k  \ k .

C. BeeHive

k

BeeHive [17] is a routing algorithm that mimics the communicative methods and foraging process of honey bees. During an actual honey source searching course, upon their return from a foraging trip, bees share the distance, direction, and quality of a flower site with other foragers by making waggle dances. By dancing zealously for a good foraging site they recruit more foragers and in this way a good flower site is exploited and reinforced. In BeeHive algorithm, each node in BeeHive has a hive, which consists of three parts: packing floor, entrance and dance floor. BeeHive uses four types of agents: packers, scouts, foragers, and swarms. Packers, like the task of a real food-storer bee, always stay in the node, receive and store the data packets from the upper layer. Scouts are used in route discovery phase to discover new routes to their destination node. Once a scout returns to its source node from destination, it recruits the foragers by using the metaphor of dance as real scout bees do. Foragers are the main workers for receiving the data packets from packers and then transporting them to their destination. On their way the

Finally, selection generates next generation population from all parents and offspring based on certain metrics. After finite generations of iteration, the algorithm selects optimal layout solution. In another paper [20], Multiple Objective GA (MOGA) was used for layout optimization, where sensor coverage and lifetime are the main concerns. Paper [21] apply GA to improve the performance of an existing weighted clustering algorithm (WCA). During the mapping process, each chromosome contains information about cluster head and its members as obtained from the original WCA. It optimizes WCA such that the number of cluster-heads is minimized while load in the network is evenly balanced among clusters. After a series of genetic operations, such as encoding, crossover, mutation, the one with the lowest fitness value is replaced as the best chromosome in the population. GA is also applied to compute an approximate mobile agent route solution for data fusion [19] by employing two-level encoding.

765

phenomena of morphogenesis can be explained by the reaction together and diffusion of a system of chemical substances through a tissue. This reaction-diffusion process involves two substances, the activator and inhibitor that both diffuse and react inside the system boundaries. Such conceptions are mimicked in the wireless access process to generate transmission patterns [26], where an existing transmission inhibits the simultaneous transmission in its interference range and activates those out of its exclusion range. Firstly [26], this reaction-diffusion process can be described by [27] second order partial differential equations in terms of the concentration of activator a (z, t) and inhibitor h (z, t) at location z and time t: wa (8) f ( a, h)  D ’ 2 a

Because of mutation operation in GA, typically random alteration of a bit in the genotype, there is no guarantee that the results of a reproduction will be much fitter or survivable. That is the reason why GA is not suitable for life-dependent processes work. IV. PARTICLE SWARM OPTIMIZATION (PSO) AND CLUSTERING IN SENSOR NETWORKS

Particle Swarm Optimization (PSO) [23] is an optimization tool motivated by foraging behaviour of bird flocking. PSO in fact runs a population-based search procedure, where particles fly around in the problem space and adjust their speed and direction according to past best fitness experience. The past experience not only includes its own past best position (pbest) experienced, but also the experience of a neighbouring particle, which may be the global best position (gbest). PSO and Genetic Algorithms (GA) have some qualities in common in terms of population evolution. However, the particles in PSO fly through the problem space by following the current optimum particles, rather than evolution operations like crossover and mutation in GA. The particles are manipulated by changing their velocity toward its pbest and gbest locations at each time step: Vnew = wi * Vold + c1* rand() * (Pbest-Pold)+c2 * rand() *(Gbest –Pold)

wh wt

g ( a, h )  D h ’ 2 h

where f and g are nonlinear function of a and h, Da and Dh are the respective diffusion rate. Furthermore, it has been shown that reaction-diffusion process can be denoted by the Cellular Neural Network, which consists of activation and inhibition influence from around cells. Finally the Medium Access Probability (MAP, p) updating algorithm was constructed to realize the global transmission patterns by using only local information: (9) p i (t  1) f ( p i (t ), p h (t ), p a (t )) ,

(7)

Pnew = Pold+ Vnew

Vnew /Vold Pnew /Pold Pbest Gbest wi c1 /c2

a

wt

TABLE IV Notions in PSO New/old velocity calculated for each particle New/old position calculated for each particle Particle’s own past best position The best position a particle attained in the whole swarm Inertial weight constant Weights values

where f is the user defined function. Hence the reactiondiffusion mechanism is embedded in the MAP updating process. It thus create an efficient scheduling of transmissions with high traffic loads, while maximizes the number of successful transmissions per unit area, i.e., density of spatial reuse.

PSO is typically exploited to solve the stochastic nonlinear optimization problems in a faster, cheaper way with few parameters to adjust. More importantly, it shows potential in solving the clustering or other optimization problems in WSN because of its inexpensive requirements: a few lines of code, primitive mathematical operators and less occupation of memory. Divided Range Particle Swarm Optimization (DRPSO) [24] was developed to solve the clustering problem in sensor networks. It uses the objective (fitness) function similar to the Weighted Clustering Algorithm (WCA), where the clusterheads are elected based on the weight of each node. PSO is also employed to solve another cluster-head identification problem in [25], where its objective is to minimize the energy expenditure while maximizing the total data gathered.

VI. OTHER BIOLOGICALLY INSPIRED APPROACHES Next we will briefly investigate other related biologically inspired approaches in wireless sensor networks. A. Quorum Sensing and Clustering Clustering in sensor networks is an efficient process for data aggregation by dividing the whole networks into clusters. In [28;29], an application layer clustering algorithm was inspired by the bacterial signalling process called quorum sensing. Quorum sensing is an intercellular signalling process, through which bacteria monitor cell population density via the exchange of autoinducers. By releasing the signalling molecules, bacteria are able to measure the number of the molecules within a population and co-ordinate their behaviour accordingly. For example, the squid benefits a lot from the bacteria's quorum sensing and their bioluminescence abilities: Visible light is emitted and used by the squid for antipredation strategies or attracting a mate. During the course, the autoinducers accumulate as the number of cells in the light organ of the squid multiplies and reach a very high concentration. The same conception motivated the clustering

V. BIOLOGICAL MORPHOGENESIS AND TRANSMISSION PATTERNS In [26], collision free transmission patterns for wireless medium access are constructed based on the reaction-diffusion mechanism, which is proposed by Turing in 1952 [27] to account for the formation of patterns in biological morphogenesis. It is indicated by Turing [27] that the main

766

control and under-layer communication issues are firstly assumed in these algorithms.

algorithm in sensor networks where the sensor node need to know if there are enough of them to form a cluster for the purpose of monitoring a particular area of the network. In the implementation of [28], two types of algorithm control packets, Node Synchronization (NS) packet and Group Synchronization (GS) packet, act like autoinducers during the course of intra-cluster and inter-cluster communication. The sensor nodes decide to join a cluster according to the spatial gradient of the signal they are monitoring. It uses two parameters to represent variation of the signalling: the number of broadcasts before a gradient updating and the interval between two continuous broadcasts.

VII. CONCLUSIONS AND REMARKS The goal of this position paper is to investigate and present a comprehensive review on the biologically inspired algorithms concerning some important issues in wireless sensor networks. Some biologically inspired methodologies are investigated including Ant Colony Optimization (ACO), Quorum Sensing (QS), Genetic Algorithm (GA), ReactionDiffusion (RD), and Particle Swarm Optimization (PSO). The implementations of these methods in various network problems are investigated, including routing, flow control, congestion control, and security problems. The remarkable adaptation, reliability and robustness in the dynamic and even hostile environments of the biological systems present useful paradigms for designing the wireless sensor networks, where the individual nodes operate without central control. The investigation ranges from cell level mechanisms to collective activities performed by social insects, in an effort to apply these inspirations to the sensor networks.

B. Biologically Inspired Security Mechanism for WSN It is also a hot spot to study on the utilization of biological mechanisms for security engineering, such as using artificial immune system for virus detection[30], intrusion detection [31], and cell biology as a key paradigm for networks [32]. An artificial immune system for computer networks was presented in [30]. Artificial antibodies are generated to detect previously encountered attack agents and self replication mechanism is involved in the process. The similar idea was used in [31] to detect intrusion by mimicking many properties of real immune system such as diversity, error tolerance, and self monitoring, and a detection system is built up to distinguish ‘self’ from ‘non-self’. In the cell and molecular biology metaphor [32], structures of organisms and networks show high similarities: organisms consists of organs, tissues, and cells, whereas a network consists of clusters and nodes. Literature [32] extracts the information exchange in cellular environments and applies it to networks from security perspective. It modeled the behaviour patterns of routers, monitor systems, and firewalls based on the analysis cellular metabolism.

ACKNOWLEDGMENT This project is supported by RGC Competitive Earmarked Research Grant #CUHK4213/04E of the Hong Kong government, awarded to Max Meng. REFERENCES [1]

[2]

[3]

[4]

C. Swarm behaviour for Swarm Based Sensors Swarm behaviours also inspired the grouping strategies for the mobile sensor nodes [33;34], which maintain safe separation between nodes while enforcing a level of cohesion to provide region coverage. The swarm model is developed based on three controlling behaviours that govern movements of particles within the swarm: separation that avoid collisions with neighbouring nodes, alignment that attempt to match velocity with nearby nodes, and cohesion that try to stay close to nearby nodes. This is not like the way in aforementioned PSO where the particles just follow the best global or local ones. Herein, this intuitive knowledge learned from flocks of birds provides a good paradigm to the distributed mobile sensors. In [33], two metrics, connectivity and coverage efficiency, are proposed to evaluate the swarm model and new routing schemes are devised to meet the dynamics challenge in such an environment. The paper [33] takes great effort on the swarm behaviours model for the purpose of a stable swarm formation. However, we should note that the autonomous

[5]

[6]

[7]

[8]

[9]

[10]

767

Eric Bonabeau, Marco Dorigo, and Guy Theraulaz, Swarm Intelligence:From Natural to Artificial Systems Oxford University Press, 1999. Ruud Schoonderwoerd, Owen Holland, Janet Bruten, and L.Rothkrantz, "Ant-Based Load Balancing In Telecommunications Networks," Adaptive Behavior, 1996. M.Heissenbuttle and T.Braun, "Ants-Based Routing in Large Scale Mobile Ad-Hoc Networks," Kommunikation in Verteilten Systemen (KiVS), 2003. D. Camara and A. A. F. Loureiro, "A GPS/ant-like routing algorithm for ad hoc networks," 2000 IEEE Wireless Communications and Networking Conference, 2000. WCNC., vol. 3, pp. 1232-1236, 2000. G.Di Caro and M.Dorigo, "AntNet: distributed stigmergetic control for communications networks," Journal of Artificial Intelligence Research, vol. 9, pp. 317-365, 1998. M. Gunes, U. Sorges, and I. Bouazizi, "ARA-the ant-colony based routing algorithm for MANETs," International Conference on Parallel Processing Workshops, 2002. Proceedings., pp. 79-85, 2002. Liu Zhenyu, M. Z. Kwiatkowska, and C. Constantinou, "A Biologically Inspired Congestion Control Routing Algorithm for MANETs," Third IEEE International Conference on Pervasive Computing and Communications Workshops, 2005. PerCom 2005 Workshops., pp. 226-231, 2005. Liu Zhenyu, M. Z. Kwiatkowska, and C. Constantinou, "A Biologically Inspired QoS Routing Algorithm for Mobile Ad Hoc Networks," 19th International Conference on Advanced Information Networking and Applications, 2005. AINA 2005., vol. 1, pp. 426-431, 2005. Zheng Xiangquan, G. Wei, and L. Renting, "An ant-based distributed routing algorithm for ad-hoc networks," 2004 International Conference on Communications, Circuits and Systems, 2004. ICCCAS 2004., vol. 1, pp. 412-417, 2004. M. T. Islam, P. Thulasiraman, and R. K. Thulasiram, "A parallel ant colony optimization algorithm for all-pair routing in MANETs,"

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

International Parallel and Distributed Processing Symposium, 2003. Proceedings., p. 8, 2003. F. Ducatelle, G. Di Caro, and L. M. Gambardella, "Ant agents for hybrid multipath routing in mobile ad hoc networks," Second Annual Conference on Wireless On-demand Network Systems and Services, 2005. WONS 2005., pp. 44-53, 2005. O. Hussein and T. Saadawi, "Ant routing algorithm for mobile ad-hoc networks (ARAMA)," Conference Proceedings of the 2003 IEEE International Performance, Computing, and Communications Conference, 2003., pp. 281-290, 2003. C. Srisathapornphat and C. C. Shen, "Ant-based energy conservation for ad hoc networks," Proceedings. The 12th International Conference on Computer Communications and Networks, 2003. ICCCN 2003., pp. 32-37, 2003. S. Marwaha, K. T. Chen, and D. Srinivasan, "Mobile agents based routing protocol for mobile ad hoc networks," IEEE Global Telecommunications Conference, 2002. GLOBECOM '02., vol. 1, pp. 163-167, 2002. M. Roth and S. Wicker, "Termite: ad-hoc networking with stigmergy," IEEE Global Telecommunications Conference, 2003. GLOBECOM '03., vol. 5, pp. 2937-2941, 2003. A. mir Pirzada, A. Datta, and C. McDonald, "Trusted routing in adhoc networks using pheromone trails," Congress on Evolutionary Computation, 2004. CEC2004., vol. 2, pp. 1938-1943, 2004. H. F. Wedde and M. Farooq, "A Performance Evaluation Framework for Nature Inspired Routing Algorithms," Lecture Notes in Computer Science, vol. 3449, pp. 136-146, Jan.2005. H. F. Wedde and M. Farooq, "A Performance Evaluation Framework for Nature Inspired Routing Algorithms," Lecture Notes in Computer Science, vol. 3449, pp. 136-146, Jan.2005. Q. Wu, N. S. V. Rao, J. Barhen, S. S. Iyenger, V. K. Vaishnavi, H. Qi, and K. Chakrabarty, "On computing mobile agent routes for data fusion in distributed sensor networks," IEEE Transactions on Knowledge and Data Engineering, vol. 16, no. 6, pp. 740-753, 2004. D. B. Jourdan and O. L. de Weck, "Layout optimization for a wireless sensor network using a multi-objective genetic algorithm," 2004 IEEE 59th Vehicular Technology Conference, 2004. VTC 2004-Spring, vol. 5, pp. 2466-2470, 2004. D. Turgut, S. K. Das, R. Elmasri, and B. Turgut, "Optimizing clustering algorithm in mobile ad hoc networks using genetic algorithmic approach," IEEE Global Telecommunications Conference, 2002. GLOBECOM '02., vol. 1, pp. 62-66, 2002.

[22]

[23]

[24]

[25]

[26]

[27] [28]

[29]

[30]

[31] [32]

[33]

[34]

768

Jianli Zhao, Yingyou Wen, Ruiqiang Shang, and Guangxing Wang, "Optimizing Sensor Node Distribution with Genetic Algorithm in Wireless Sensor Network," ISNN 2004, LNCS 3174, pp. 242–247, 2004. J. Kennedy and R. Eberhart, "Particle swarm optimization," IEEE International Conference on Neural Networks, 1995. Proceedings, vol. 4, pp. 1942-1948, 1995. J. Chunlin, Z. Yangyang, G. Shixing, Y. Ping, and L. Zhe, "Particle swarm optimization for mobile ad hoc networks clustering," 2004 IEEE International Conference on Networking, Sensing and Control, vol. 1, pp. 372-375, 2004. J. Tillett, R. Rao, and F. Sahin, "Cluster-head identification in ad hoc sensor networks using particle swarm optimization," 2002 IEEE International Conference on Personal Wireless Communications, pp. 201-205, 2002. Mathilde Durvy and Patrick Thiran, "Reaction-Diffusion Based Transmission Patterns for Ad Hoc Networks," IEEE Infocom 2005, Miami, 13-17 Mar, 2005. A.Turing, "The chemical basis of morphogenesis," Philosophical Transactions of the Royal Society B (London), 237,37-72, 1952. I. Wokoma, L. L. Shum, L. Sacks, and I. Marshall, "A biologicallyinspired clustering algorithm dependent on spatial data in sensor networks," Proceeedings of the Second European Workshop on Wireless Sensor Networks, 2005., pp. 386-390, 2005. I. Wokoma, L.Sacks, and I.Marshall, "Clustering in Sensor Networks using Quorum Sensing," London Communications Symposium, University College London, 8th-9th September, 2003. J.O.Kephart, "A Biologically Inspired Immune System for Computers," Proceedings of 4th International Workshop on Synthesis and Simulation of Living Systems, Cambridge, Massachusetts, USA, pp. 130-139, 1994. S.Hofmeyer and S.Forrest, "Architecture for an Artificial Immune System," Evolutionary Computation, vol. 8, pp. 443-473, 2000. Falko Dressler, "Bio-inspired mechanisms for efficient and adaptive network security mechanisms," Proceedings of Dagstuhl Seminar 04411 on Service Management and Self-Organization in IP-based Networks, Schloss Dagstuhl, Wadern, Germany, October, 2004. B.Anthony Kadrovach and Gary B.Lamont, "A particle swarm model for swarm-based networked sensor systems," Proceedings of the 2002 ACM symposium on Applied computing, pp. 918-924, 2002. B.Anthony Kadrovach and Gary B.Lamont, "Design and analysis of swarm-based sensor systems," Symposium on Circuits and Systems, Fairborn, OH, August, 2001.