BeesAnts: a new nature-inspired routing algorithm Eslam Al ...

Int. J. Communication Networks and Distributed Systems, Vol. 10, No. 1, 2013

BeesAnts: a new nature-inspired routing algorithm Eslam Al Maghayreh*, Sallam Abu Al-Haija and Faisal Alkhateeb Computer Science Department, Faculty of Information Technology and Computer Sciences, Yarmouk University, Irbid 21163, Jordan E-mail: [email protected] E-mail: [email protected] E-mail: [email protected] *Corresponding author

Shadi Aljawarneh Science and IT Faculty, Isra University, Amman 11622, Jordan E-mail: [email protected]

Emad Al-Shawakfa Computer Information Systems Department, Faculty of Information Technology and Computer Sciences, Yarmouk University, Irbid 21163, Jordan E-mail: [email protected] Abstract: In this paper, we have introduced a new multi-agent nature-inspired routing algorithm. The algorithm is referred to as the BeesAnts algorithm. It is a combination of the ant colony-based routing algorithm (ARA) and the BeeHive-based routing algorithm. The proposed routing algorithm works effectively on networks consisting of two parts; one is a fixed network and the other is a mobile ad hoc network (MANET). It applies the ARA routing algorithm on the mobile part and the BeeHive routing algorithm on the fixed part. The experimental results and the statistical analysis have demonstrated that the BeesAnts routing algorithm outperforms its ancestor, the ARA routing algorithm, in terms of the propagation delay, the queue delay, and the number of hops. Keywords: mobile ad hoc networks; MANETs; routing algorithms; swarm intelligence; ant colony-based routing; beehive-based routing.

Copyright © 2013 Inderscience Enterprises Ltd.

83

84

E. Al Maghayreh et al. Reference to this paper should be made as follows: Al Maghayreh, E., Al-Haija, S.A., Alkhateeb, F., Aljawarneh, S. and Al-Shawakfa, E. (2013) ‘BeesAnts: a new nature-inspired routing algorithm’, Int. J. Communication Networks and Distributed Systems, Vol. 10, No. 1, pp.83–97. Biographical notes: Eslam Al Maghayreh is an Assistant Professor at the Department of Computer Science, Yarmouk University (Jordan) since June 2008. He received his PhD in Computer Science from Concordia University (Canada) in 2008, his MSc in Computer Science from Yarmouk University in 2003, and his BSc in Computer Science from Yarmouk University in 2001. His research interests include distributed systems, multi-agent systems, and runtime verification of distributed programmes. Sallam Abu Al-Haija is currently pursuing her PhD at Hamburg-Harburg Technical University, Hamburg, Germany in the field of swarm intelligence. She obtained both her MSc and Bachelor degree in Computer Science from Yarmouk University in 2010, 2006 respectively. Her research interests includes swarm intelligence, natural language processing and computer networks. Faisal Alkhateeb is currently an Assistant Professor at the Department of Computer Science, Yarmouk University. He holds a BSc from Yarmouk University in 1999, an MSc from Yarmouk University in 2003, an MSc from Grenoble 1 in 2004, and a PhD from Grenoble 1 in 2008. He is interested in knowledge-based systems, knowledge representation and reasoning, intelligent systems, e-learning, constraint satisfaction and optimisation problems. He has published more than 16 articles in journals, books, and conference and workshop contributions. Shadi Aljawarneh holds a BSc in Computer Science from Yarmouk University, Jordan, an MSc in Information Technology from Western Sydney University and a PhD in Software Engineering from Northumbria University, England. He is currently an Assistant Professor at the Faculty of IT, Isra University, Jordan since 2008. His research is centred in web and network security, e-learning, bioinformatics, and ICT fields. He has presented at and been on the organising committees for a number of international conferences and is a board member of the International Community for ACM, ACS, and others. Emad Al-Shawakfa is an Assistant Professor at the Computer Information Systems Department, Faculty of IT, Yarmouk University since September 2000. He holds a PhD in Computer Science from Illinois Institute of Technology (IIT) – Chicago, USA in the year 2000, an MSc in Computer Engineering from Middle East Technical University – Ankara, Turkey in 1989, and a BSc in Computer Science from Yarmouk University – Jordan in 1986. His research interests are in computer networks and natural language processing.

BeesAnts: a new nature-inspired routing algorithm

85

1 Introduction A mobile ad hoc network (MANET) is a group of mobile nodes connected via wireless links where these nodes use a decentralised way in organising the network and passing messages among them (Frodigh et al., 2004). Due to the limited transmission range of wireless network interfaces, the data to be exchanged with other nodes across the network has to pass through several intermediate nodes. A vital challenge, which is considered as the heart of building a MANET, is finding the best path from a source node to a destination one. This gives rise to the routing problem. Therefore, developing a MANET goes hand in hand with developing a good routing protocol to find the optimal path in such a dynamic network. A good routing protocol increases the performance of the network as a whole (Layuan et al., 2007; Kurkowski et al., 2005, 2006; G¨unes et al., 2002; Abolhasan et al., 2004). Realising the importance of a good routing algorithm in a MANET droves researchers into the search for the best way to find such an algorithm. Hence, the turn to the nature has attracted researchers to societies similar to the MANET environment in terms of the decentralised behaviour and the dynamic environmental changes, like ants’ and bees’ colonies. Consequently, a new branch focusing on applying biological features into a network environment has been adopted by many researchers (Wedde and Farooq, 2006; Ducatelle et al., 2005, 2008; Blum and Merkle, 2008; Abraham et al., 2006; Lee et al., 2008). The ant colony-based routing algorithm (ARA) is inspired by ants’ behaviour in the real world where it is observed that foraging ants converge to the shortest path between their nest and the food source, depending on a volatile chemical substance called pheromone (Woo et al., 2008; Zanjani and Haghighat, 2009; G¨unes et al., 2002; Blum and Merkle, 2008; Olariu and Zomaya, 2005; Caro and Dorigo, 1998b; Attia et al., 2009; Maekawa et al., 2006). Initially, ants start their way from the nest looking for a food source in arbitrary directions, and while walking, they mark the route taken by depositing a pheromone. Arriving to a branch, ants should decide which direction to choose, initially, this is done randomly, where ants divide themselves into groups each of which tries a different branch. After a short time, ants that took the shortest path will return more quickly causing an increase in the pheromone’s concentration on the route taken by the moving ants back and forth, thus, announcing the best route to the food source. On the other hand, the longer paths will take ants longer time to return and thus, the pheromone amount will decay by time informing other ants that it is not the best way (G¨unes et al., 2002). Now we have to translate these concepts into a network language. One of the major issues handled by this algorithm is the route discovery, where ants are broadcasted from the sender to the neighbours. Each node receives an ant will create a routing table record then passes it to its neighbours and so on until the ant reaches the destination node where it is destroyed and a new ant is created and sent backward to the source node. This process might result in multiple paths, which is considered as an advantage to this algorithm. The ARA algorithm is characterised by local work, where routing tables do not have to be transmitted to all nodes in the network but only to the neighbouring nodes (Olariu and Zomaya, 2005; Osagie et al., 2008). Similarly, the BeeHive routing algorithm is inspired by the behaviour of bees. Bees can find the optimal path to a food source, and they communicate directly via dancing.

86

E. Al Maghayreh et al.

There are two types of dancing; round dances, which indicate that a food source is close to the hive, and the waggle dances, which indicates farther food sources. The duration and orientation of a dance inform other bees about the exact place of the food source and its goodness (Olariu and Zomaya, 2005; Wedde et al., 2004). Each node in the network is considered as a hive that consists of bee agents, which are launched to other network nodes in order to explore the network and to collect information about routes. Bee agents provide the nodes they visit with information about the paths they have explored so far. In addition, the dance floor represents the routing table, where the information about the quality of the paths that have been traversed is stored (Olariu and Zomaya, 2005; Wedde et al., 2004). On contrary to the ARA algorithm, the beehive routing algorithm uses short distance bee agents to collect and disseminate routing information to neighbourhood and long distance bee agents to disseminate routing information to all nodes (Wedde et al., 2004). This paper presents a new multi-agent nature-inspired routing algorithm where bee agents and ant agents work together to find the best path between a given source and destination nodes. In this work, we have revised the algorithm presented in Maghayreh et al. (2010). We have conducted more through experimental and statistical analysis to demonstrate the effectiveness of the proposed routing algorithm.

2 Related works G¨ unes et al. (2002) have developed an on-demand ad-hoc routing algorithm, which is based on swarm intelligence. Their algorithm is referred to as the ARA. It is assumed that the network is represented as a connected graph G = (V, E). The ant colony optimisation meta-heuristic can be exploited to find the shortest path between a source node vs and a destination node vd on the graph G. The number of nodes on the path represents its length. Each edge ei,j ∈ E connects two nodes in the graph (vi and vj ) and has a variable φi,j to represent the concentration of an artificial pheromone. The concentration of the pheromone is an indication of the usage of this edge. An ant located in node vi uses the pheromone value φi,j of edge ei,j to compute the probability of node vj as the next hop (Pi,j ). Assuming that Ci is the set of nodes reachable from vi in one hop, then Pi,j = ∑

φi,j j∈Ci

φi,j

When an ant traverses an edge between node vi and node vj the pheromone value φi,j is incremented by a constant value ∆φ. As it is the case with the real life ant pheromone, the concentration of the artificial pheromone decreases with time. At every constant interval t, the value of φi,j is decreased by α. If φi,j becomes less than zero, it is set to zero, indicating no pheromone. The ARA algorithm is composed of three phases, the main phase is the route discovery phase. In order to discover new routes, a node uses two artificial ant agents; a forward ANT (FANT) and a backward ANT (BANT). A FANT is a small packet carrying a unique sequence number; to enable the receiving node to recognise the duplicated packets. The main work of the FANT is to discover the route from the source node to the destination using pheromone tracks. The BANT establishes the way back to the source node.

BeesAnts: a new nature-inspired routing algorithm

87

At the beginning, the source node broadcasts FANTs to its neighbours, then for each neighbour, when it receives a FANT, it creates a corresponding new record in its routing table consisting of the destination (which is interpreted as the source node from which the FANT was sent), the next hop (which represents the previous node where the FANT came from), and lastly the pheromone value that is calculated depending on the number of hops a FANT passed through so far (G¨unes et al., 2002). These operations are repeated until the FANT reaches the destination. When the destination node receives a FANT, recognised by its unique sequence number, it kills this FANT and creates a BANT with the same unique sequence number and sends it back to the node from which it has received the FANT. After establishing the route along the way back to the source node by the BANT agent, data packets can be sent using this generated path (G¨unes et al., 2002; Cauvery and Viswanatha, 2008). G¨ unes et al. (2002) have tested their proposed algorithm using ns-2 simulator on a network of 50 mobile nodes and an area of 1,500 m x 300 m. Seven different pause times have been used to express mobility; 0, 30, 60, 120, 300, 600, and 900 seconds. They have compared the result to existing routing protocols for the delivery rate of packets. It was concluded that the algorithm has a good performance, especially in highly dynamic situations (Cauvery and Viswanatha, 2008). BeeHive is one of the primary routing algorithms that was presented by Wedde et al. (2004), where they benefit from the real bee behaviour with its colleagues. A bee can talk to its colleagues through dancing; where a dance duration and orientation tell other bees about a food source location and its goodness (Blum and Merkle, 2008; Abraham et al., 2006; Wedde and Farooq, 2006). In the BeeHive algorithm, each node represents a hive, and contains two types of agents; short distance bee agents and long distance bee agents. Moreover, the network is divided into parts; each node has its own foraging zone that contains the neighbouring nodes, also a foraging region that contains a representative node responsible for sending the long distance bee agents to the external part of the network or to other regions. The short distance bee agents traverse the nodes in the foraging zone (Wedde et al., 2004). Each node maintains three types of routing tables; foraging region membership (FRM), intra foraging zone (IFZ), and inter foraging region (IFR). The IFZ is a vector correlating the set of destinations with the set of neighbouring nodes, and each single entry is a combination between queue delay and propagation delay (qjk , pjk ) expressing the cost of reaching the destination k using the neighbour j. On the other hand, the IFR maintains the queue and the propagation delay for reaching the representative node of each foraging region through the neighbours of a node. Lastly, the FRM routing table contains the mapping of known destinations to a foraging region. The process begins when each node broadcasts short bee agents to its neighbourhood. Each neighbouring node updates its routing table based on the information carried by the bee agent if it is not duplicated, if duplicated; the bee agent will be killed. When receiving the bee agent, the node updates the queue delay and updates either the IFZ or the IFR depending on the place of the bee agent (the zone of the node or the region). A node can recognise whether the bee is in its zone or region depending on the number of hops that the bee agent has passed to reach the node (Wedde et al., 2004).

88

E. Al Maghayreh et al.

After these calculations, the node forwards this bee agent again to its neighbours. Many bees explore the neighbourhood whereas few bees travel to the external part. The routing table represents a dance floor where bee agents communicate (Wedde et al., 2004). At the end, the node must calculate the goodness of the neighbours. Wedde et al. (2004) have tested the performance of the suggested algorithm, and compared it with the AntNet algorithm (Caro and Dorigo, 1998a), it was deduced that the BeeHive routing algorithm has a better or similar performance to AntNet (Wedde et al., 2004). On the other hand, Rahmatizadeh et al. (2009) have developed a algorithm that exploits ants and bees in finding the best path in a network. The algorithm is referred to as the Ant-Bee routing (ABR) algorithm. It is based on the AntNet routing algorithm. In the ABR algorithm, a node generates regularly an ant called forward-ant. Each forward-ant explores the network to find the best path and then updates the routing tables. When an ant reaches a node, if it is not the destinations, the ant simply continues its way to the destination, but if it was the destination node, the ant is killed and a bee is created to be sent back in the same path to the source node. All the data collected and carried by the ant is transferred to the newly born bee. This collaboration between artificial ants and bees aims at increasing the speed of delivering data, and as a result, the throughput and the performance of the network will be enhanced. The ABR algorithm has been tested in static and dynamic situations. It has been shown that the ABR algorithm performs well in static situations in which the routers are working properly. In dynamic situations, where routers may fail, the ABR algorithm delivers packets with a higher speed but keeps the same level of throughput as of that of the AntNet algorithm. Consequently, the ABR algorithm is expected to be more suitable for real networks where routers may fail temporarily (Rahmatizadeh et al., 2009). Compared to our algorithm, the ABR algorithm works for fixed networks only. It is an extension of the AntNet routing algorithm which is also suitable for fixed networks. Our algorithm can be used in a network consisting of mobile and fixed parts. The following section presents the details of the proposed algorithm.

3 The BeesAnts routing algorithm In this paper, we are presenting a new routing algorithm based on swarm intelligence. We will call it the BeesAnts algorithm. It is an on demand routing algorithm, which means that whenever a node wants to send a message to another node, it should run this algorithm to discover the best path to the destination node. There are two ancestors for this algorithm, the ARA and the BeeHive routing algorithm. The former one is a well known routing algorithm over MANETs, whereas the latter one is usually applied on fixed wireless networks. The merge between the two algorithms is therefore expected to be applied on a hybrid network. At the beginning, we divide the network into two parts; the first part contains mobile ad hoc nodes (mobile part), whereas the second part contains the fixed nodes only (fixed part). When a node wants to send a message to another node, the source node runs the BeesAnts routing algorithm to discover the best path to the destination node. Initially, the BeeHive routing algorithm is used to find a path between the source and the destination. This path is then used to figure out the node that resides at the

BeesAnts: a new nature-inspired routing algorithm

89

border between the two parts of the network (the mobile and the fixed parts). We will refer to this node as the border node. Next, the ARA algorithm is applied to find (if possible) a better route between the source and the border node, where both nodes are in the mobile part of the network. Finally, the path found by the ARA routing algorithm is merged with the rest of the path found by the BeeHive routing algorithm (from the border node to the destination node). The new path represents the best path between the source and the destination that can be found by the BeesAnts algorithm. Algorithm 1 shows the pseudocode for the proposed algorithm. Algorithm 1

The BeesAnts routing algorithm

Let (S) be the source node and (D) be the destination node. Step 1: Find a path from S to D using the BeeHive routing algorithm. Step 2: Figure out the node on the border (B) between the mobile part and the fixed part of the network. Step 3: Find the optimal path between S and B using the ARA routing algorithm. Step 4: Merge the path established in Step 4 with the portion of the path established in Step 1 that connect the border node and the destination node. Step 5: Use the path established in Step 4 to send data packets from S to D.

4 Experimental results In order to achieve the goal behind this research, we have implemented the BeesAnts routing algorithm using MATLAB. The number of nodes in the network used in our experiments is 50 (G¨ unes et al., 2002). The network covers an area of 1,000 m x 1,000 m unit squares. The node mobility of the mobile part of the network is expressed by the pause time; that is to measure the dynamic behaviour of the nodes in the mobile part of the network, the nodes move, and after few seconds they are paused and the routing algorithm is executed. For example, the nodes are allowed to move for 30 seconds, and then, the scene is stilled and the algorithm is applied for the current positions of the nodes. Tables 1 to 4 show the results of our experiments. The pause times chosen are 0, 30, 60, 120, 300, 600 and 900 seconds. We have assumed that the source is node number 1 and the destination is node number 50. We have executed our algorithm and the ARA algorithm 500 times each. For each run, we have measured the propagation delay, the queue delay and the number of hops. Table 1 shows the measured results for the propagation delay. The table is divided into three main parts; the first part shows the pause time, the second part shows the results obtained using our algorithm and the last part shows the results obtained using the ARA algorithm. Each of the last two parts is divided into six columns; each of the first five columns (T1 to T5) shows the average propagation delay measured after executing the corresponding algorithm 100 times, and hence the AVG column shows the average propagation delay after executing the algorithm 500 times. Tables 2 and 3 can be interpreted in the same way. Table 4 summarises all the results shown in Tables 1 to 3. Each result in this table was obtained after running the corresponding routing algorithm 500 times, and then taking the average.

0 30 60 120 300 600 900

Pause time

12,474 20,590 104,459 227,767 354,987 12,438 12,438

T1 24,272 85,211 77,515 165,671 500,758 232,997 871,764

T3 23,121 105,071 98,463 36,675 549,038 452,646 566,359

T4

T5 15,212 22,416 22,475 286,802 163,570 195,228 226,794

BeesAnts routing algorithm 17,879 142,105 47,513 21,752 35,302 664,738 765,257

T2 18,592 75,079 70,085 147,733 320,731 311,610 488,522

AVG 18,361 21,860 107,388 255,698 354,111 18,361 18,361

T1 21,971 140,518 54,069 28,805 183,235 679,821 773,346

T2 30,209 90,833 79,091 177,154 527,346 235,915 856,628

T3

26,325 115,628 109,786 56,433 533,753 454,147 573,512

T4

ARA algorithm 16,669 27,832 231,190 287,955 164,990 216,686 266,743

T5

22,707 79,334 116,305 161,209 352,687 320,986 497,718

AVG

90 E. Al Maghayreh et al.

Table 1 A summary of our experimental results (propagation delay)

0 30 60 120 300 600 900

Pause time

0.0303 0.0062 0.0086 0.0097 0.0051 0.0307 0.0291

T1 0.0573 0.0080 0.0075 0.0057 0.0037 0.0102 0.0043

T3 0.0541 0.0060 0.0045 0.0076 0.0037 0.0157 0.0041

T4

T5 0.0350 0.0082 0.0071 0.0032 0.0074 0.0033 0.0109

BeesAnts routing algorithm 0.0534 0.0026 0.0113 0.0434 0.0094 0.0043 0.0080

T2 0.0460 0.0062 0.0078 0.0139 0.0058 0.0129 0.0113

AVG 0.0376 0.0076 0.0086 0.0135 0.0061 0.0424 0.0380

T1 0.0698 0.0035 0.0114 0.0449 0.0128 0.0024 0.0048

T2 0.0672 0.0114 0.0088 0.0068 0.0045 0.0140 0.0035

T3

0.0653 0.0087 0.0055 0.0084 0.0028 0.0217 0.0093

T4

ARA algorithm 0.0468 00115 0.0024 0.0037 0.0093 0.0048 0.0154

T5

0.0573 0.0085 0.0074 0.0155 0.0071 0.0171 0.0142

AVG

BeesAnts: a new nature-inspired routing algorithm

Table 2 A summary of our experimental results (queue delay)

91

92

E. Al Maghayreh et al.

Table 3 A summary of our experimental results (number of hops) BeesAnts routing algorithm

Pause time 0 30 60 120 300 600 900

ARA algorithm

T1

T2

T3

T4

T5

AVG

T1

T2

T3

T4

T5

AVG

18 31 46 41 44 18 18

17 42 44 19 35 38 45

29 45 43 38 43 45 43

35 40 26 36 46 44 41

14 39 36 35 41 35 35

23 39 39 34 42 36 36

25 32 48 44 44 25 25

21 42 47 24 43 38 45

36 49 44 39 43 47 43

38 41 32 38 46 45 41

17 44 45 35 42 37 41

27 42 43 36 44 38 39

Table 4 A summary of our experimental results BeesAnts routing algorithm Pause time Propagation delay Queue delay No. of (distance unit) (seconds) hops 0 30 60 120 300 600 900

18,592.2427 75,079.0993 70,085.3756 147,733.7301 320,731.5828 311,609.9553 488,522.6441

0.046104 0.006194 0.007782 0.013928 0.005848 0.012862 0.011290

23 39 44 19 35 38 36

ARA algorithm Propagation delay Queue delay No. of (distance unit) (seconds) hops 22,707.6140 79,334.5286 116,305.2535 161,209.5339 352,687.5094 320,986.7027 497,718.4537

0.075333 0.008580 0.007350 0.015475 0.007105 0.017053 0.014211

27 42 43 36 44 38 39

4.1 Analysis and charts Compared with the ARA algorithm, the results shown in Table 4 as well as Figures 1 to 3 demonstrate that our BeesAnts routing algorithm has proven superiority over the ARA algorithm. In almost all experiments, the BeesAnts algorithm finds a better route with less propagation delay, less queue delay, and less number of hops than the ARA algorithm. Figure 1

The propagation delay of the BeesAnts algorithm and the ARA algorithm depicted at different pause times

BeesAnts: a new nature-inspired routing algorithm Figure 2

The queue delay of the BeesAnts algorithm and the ARA algorithm depicted at different pause times

Figure 3

The number of hops needed by the BeesAnts algorithm and the ARA algorithm depicted at different pause times

93

Figure 4 demonstrates the difference between the proposed BeesAnts routing algorithm and the ARA algorithm. It is assumed that the source node is the third node and the destination node is node number 47. The network contains 50 nodes and it is divided into a fixed part and a mobile part separated by the curved line in the figure. The figure depicts the two paths from the source to the destination generated by the BeesAnts algorithm (the solid path) and the ARA algorithm (the dotted path). It is obvious that the paths are almost identical in the mobile part, because only the ARA algorithm is applied in this part. However, the difference is mainly produced in the second part where the BeeHive algorithm is used instead of the ARA algorithm. The resulting path (solid path) is better than the path generated by pure ARA algorithm (dotted path) in terms of the propagation delay (19,684.7507 compared to 20,759.0151), the queue delay (0.046022 compared to 0.05319) and the number of hops (17 compared to 20).

94

E. Al Maghayreh et al.

Figure 4

An example demonstrating the difference between the the proposed BeesAnts routing algorithm and the ARA algorithm

Other experiments were conducted changing the source and the destination nodes. The obtained results were similar to the results obtained in the figures. In order to determine if the results are statistically significant, statistical tests were performed. The following section presents the details of the statistical analysis performed.

4.2 Statistical tests We have applied the T statistical test to examine the goodness of the BeesAnts routing algorithm compared to the ARA routing algorithm in terms of the propagation delay, the queue delay, and the number of hops. In the first test, we have considered the propagation delay. The first sample represents the propagation delay results of the BeesAnts algorithm calculated for each of the pause times 0, 30, 60, 120, 300, 600, and 900 (the second column of Table 4). The second sample represents the propagation delay results of the ARA algorithm calculated for each of the pause times 0, 30, 60, 120, 300, 600, and 900 (the fifth column of Table 4). Let µ1

mean value for the first sample (BeesAnts results)

µ2

mean value for the second sample (ARA results)

BeesAnts: a new nature-inspired routing algorithm

95

We have the following hypotheses: Ho = µ1 < µ2 H1 = µ1 > µ2 Obviously, we want to prove hypothesis Ho in order to prove that the propagation delay of the BeesAnts algorithm is less than that of the ARA algorithm. Using MS-Excel ready function T T EST to calculate the T value, we found Tpropagation = 0.015575274. Using the T table for a sample size = 7 (because in the sample used in the test, we have executed the algorithm using seven different pause times, so each sample contains seven numbers; one corresponding to each pause time), we found that the expected T value (T = 1.943) for α = 0.05. Eventually, comparing the calculated T value with the expected value indicates that the actual T value is less than the expected T value which confirms the null hypothesis Ho , and ensures the superiority of the BeesAnts routing algorithm over the ARA algorithm with a confidence interval of 95%. The same steps can be followed to apply the T test on the samples representing the number of hops and the queue delay. We found that Tqueue = 0.092947857 and Thops = 0.038252422. Comparing the calculated T values with the expected values has indicated that the actual T values are less than the expected T values; which confirms the null hypothesis Ho in the last two tests, and ensures the superiority of the BeesAnts routing algorithm over the ARA algorithm (in terms of the queue delay and the number of hops) with a confidence interval of 95%.

5 Conclusions and future work In this paper, we have proposed and implemented a new multi-agent nature-inspired routing algorithm referred to as the BeesAnts algorithm. The BeesAnts algorithm combines the ARA and the BeeHive routing algorithm. The experimental results have demonstrated that building a routing algorithm inspired by the behaviour of both ants and bees has been proved to be a promising choice. The statistical analysis proved that BeesAnts indeed surpasses its ancestor, ARA algorithm, in terms of the propagation delay, the queue delay, and the number of hops. Therefore, adopting BeesAnts is a good choice especially for networks involving two parts; MANET part, and fixed wireless part. The results of this paper can be expanded and improved by launching ants in parallel with the discovery of the border node. Consequently, bees are moving to the farthest nodes and ants are enhancing the path between the source node and the border node simultaneously.

References Abolhasan, M., Wysocki, T. and Dutkiewicz, E. (2004) ‘A review of routing protocols for mobile ad hoc networks’, Ad Hoc Networks, Vol. 2, No. 1, pp.1–22. Abraham, A., Grosan, C. and Ramos, V. (2006) Stigmergic Optimization (Studies in Computational Intelligence), Springer-Verlag, New York, Inc., Secaucus, NJ, USA.

96

E. Al Maghayreh et al.

Attia, R., Rizk, R. and Mariee, M. (2009) ‘A hybrid multi-path ant QoS routing algorithm for MANETs’, in IFIP International Conference on Wireless and Optical Communications Networks, WOCN ‘09, pp.1–5. Blum, C. and Merkle, D. (Eds.) (2008) Swarm Intelligence: Introduction and Applications, Natural Computing Series, Springer, Berlin Heidelberg. Caro, G.D. and Dorigo, M. (1998a) ‘Antnet: distributed stigmergetic control for communications networks’, Journal of Artificial Intelligence Research, Vol. 9, No. 1, pp.317–365. Caro, G.D. and Dorigo, M. (1998b) ‘Two ant colony algorithms for best-effort routing in datagram networks’, in Proceedings of the Tenth IASTED International Conference on Parallel and Distributed Computing and Systems (PDCS98), pp.541–546, IASTED/ACTA Press. Cauvery, N.K. and Viswanatha, K.V. (2008) ‘Enhanced ant colony based algorithm for routing in mobile ad hoc network’, in Proceedings of the International Conference on Artificial Intelligence and Neural Networks, pp.30–35, Bangkok, Thailand. Ducatelle, F., Di Caro, G. and Gambardella, L. (2008) ‘An evaluation of two swarm intelligence MANET routing algorithms in an urban environment’, in Swarm Intelligence Symposium, pp.1–8. Ducatelle, F., Di Caro, G. and Gambardella, L.M. (2005) ‘using ant agents to combine reactive and proactive strategies for routing in mobile ad-hoc networks’, International Journal of Computational Intelligence and Applications, Special Issue on Nature-Inspired Approaches to Networks and Telecommunications, Vol. 5, No. 2, pp.169–184. Frodigh, M., Johansson, P. and Larsson, P. (2004) ‘Wireless ad hoc networking: the art of networking without a network’, Ericsson Review, No. 4, pp.248–263. G¨ unes, M., Sorges, U. and Bouazizi, I. (2002) ‘ARA – he ant-colony based routing algorithm for MANETs’, in ICPP Workshops, pp.79–85. Kurkowski, S., Camp, T. and Colagrosso, M. (2005) ‘MANET simulation studies: the incredibles’, SIGMOBILE Mob. Comput. Commun. Rev., Vol. 9, No. 4, pp.50–61. Kurkowski, S., Camp, T. and Navidi, W. (2006) ‘Two standards for rigorous MANET routing protocol evaluation’, in IEEE International Conference on Mobile Adhoc and Sensor Systems (MASS), pp.256–266. Layuan, L., Chunlin, L. and Peiyan, Y. (2007) ‘Performance evaluation and simulations of routing protocols in ad hoc networks’, Comput. Commun., Vol. 30, No. 8, pp.1890–1898. Lee, H., Choi, G., Kim, H. and Lee, K. (2008) ‘A nature-inspired QoS routing algorithm for next generation networks’, in Fourth International Conference on Autonomic and Autonomous Systems, ICAS ‘08, pp.226–232. Maekawa, T., Tada, H., Wakamiya, N., Imase, M. and Murata, M. (2006) ‘An ant-based routing protocol using unidirectional links for heterogeneous mobile ad-hoc networks’, in International Conference on Wireless and Mobile Communications, ICWMC ‘06, p.43. Maghayreh, E.A., Al-Haija, S.A., Alkhateeb, F. and Aljawarneh, S. (2010) ‘Bees-ants based routing algorithm’, in International Conference on Intelligent Systems, Modelling and Simulation, pp.344–349, IEEE Computer Society. Olariu, S. and Zomaya, A.Y. (2005) Handbook of Bioinspired Algorithms and Applications, Chapman & Hall/CRC Computer & Information Science. Osagie, E., Thulasiraman, P. and Thulasiram, R.K. (2008) PACONET: Improved Ant Colony Optimization Routing Algorithm for Mobile Ad Hoc Networks, pp.204–211, Los Alamitos, CA, USA, IEEE Computer Society. Rahmatizadeh, S., Shah-Hosseini, H. and Torkaman, H. (2009) ‘The ant-bee routing algorithm: a new agent nature-inspired routing algorithm’, Journal of Applied Sciences, Vol. 9, No. 5, pp.983–987.

BeesAnts: a new nature-inspired routing algorithm

97

Wedde, H., Farooq, M. and Zhang, Y. (2004) ‘BeeHive: an efficient fault-tolerant routing algorithm inspired by honey bee behavior’, in ANTS Workshop, pp.83–94. Wedde, H.F. and Farooq, M. (2006) ‘A comprehensive review of nature inspired routing algorithms for fixed telecommunication networks’, Journal of Systems Architecture, Vol. 52, Nos. 8–9, pp.461–484. Woo, M., Dung, N.H. and Roh, W.J. (2008) ‘An efficient ant-based routing algorithm for MANETs’, in 10th International Conference on Advanced Communication Technology, ICACT ‘08, pp.933–937. Zanjani, V. and Haghighat, A. (2009) ‘Adaptive routing in ad hoc wireless networks using ant colony optimization’, in the International Conference on Computer Technology and Development, ICCTD ‘09, pp.40–45.