MANET routing protocols based on swarm intelligence

Transcription

1 MSC SEMINAR - PERVASIVE AND ARTIFICIAL INTELLIGENCE RESEARCH GROUP, DEPARTMENT OF INFORMATICS, UNIVERSITY OF FRIBOURG. JUNE MANET routing protocols based on swarm intelligence Iliya Enchev Pervasive and Artificial Intelligence Research Group Department of Informatics University of Fribourg iliya.enchev@unifr.ch Abstract A mobile ad-hoc network (MANET) is formed by a group of mobile nodes which communicate over wireless connection. The nature of this kind of networks makes them suitable for deployment in places where network infrastructure is hard to build and maintain. They are flexible, adaptive and require few resources but in the same time they are challenging in finding paths that support optimal communication between their component nodes. In order to fulfil this task, suitable routing protocols are required that are robust, reliable, efficient and at the same time as simple as possible. There exist a number of swarm intelligence based protocols that try to meet these criteria. They are based on the behaviour of animals that form swarms. Two popular group of such protocols are bee- and antinspired protocols, which take their principles from ant and bee colonies. In this work the ant-inspired protocols - ARA, SARA, ANSI, AntHocNet, HOPNET and the bee-inspired protocol - BeeAdHoc are regarded and their main features, advantages and disadvantages are compared between each other as well as to some well established non swarm intelligence protocols, like AODV. Some results are presented at the end of the paper. is becoming less occupied, probably due to some deterioration. Fig. 1 shows a scenario in which the best route between two choices is chosen by the ants after a while. Bee-inspired algorithms are mainly based on the foraging principle of honey bees. Two main types of bees are utilized for doing routing in MANET - scouts and foragers. Scouts discover new nodes from their source node to their destination node. When a scout reaches its destination, it starts a backward journey to its source. When returned to the source, a scout recruits the foragers for its route by using the metaphor of dance [1]. I. INTRODUCTION Swarm intelligence is defined as the collective behaviour of decentralized, self-organized groups. They are made up of simple agents that interact with the environment (so called stigmergy) and between each other. The agents follow simple rules and possess themselves limited capabilities. They don t follow centralized orders for each individual and interact locally and randomly but together, from a global point of view, their behaviour emerges as intelligent. Examples for such behaviour are searching for food by ants, or searching for nectar by bees. The nature of swarms largely resembles mobile ad-hoc networks (MANETs) and that is why ideas from swarm animals like ants and bees are used for creating suitable routing protocols for MANETs. The basic idea behind ant-based routing algorithm is taken from the food searching strategy of real ants. They start searching food from their nest and walk towards the food, sampling different routes. When an ant reaches an intersection it has to make a decision which way to take next. Also while walking (to the food source and back), ants leave pheromone, a chemical substance, which marks the route they took. Other ants can smell the pheromone. They can distinguish its concentration as well, which gives a hint to them for the usage of the route and influences their choice. With time the concentration of pheromone decreases due to diffusion. This property is important for knowing which route Fig. 1. All ants take the shortest path after an initial searching time. [2] A. Ant-inspired routing protocols The first protocol taken into account in this work is Ant- Colony-Based Routing Algorithm (ARA)[2]. It is a highly adaptive, scalable efficient on-demand (reactive) MANET routing protocol. The next algorithm considered in this work is the Simple Ant Routing Algorithm (SARA) proposed by F. Correia and T. Vazao [3] which is a kind of evolution of the one above - ARA. It offers low overhead facilitated through all three routing phases - discovery, maintenance and recovery (which will be explained later in this work). This is done by means of three complementary mechanisms - Controlled Neighbour Broadcast (CNB), maintaining active sessions paths and deep search procedure that restricts the number of nodes searched.

2 An unicast protocol for hybrid ad hoc networks - ANSI (Ad hoc Networking with Swarm Intelligence) [4] considers the existence of higher-capability mobile or stationary devices in a network. It is responsive to fluctuating topology and utilizes the common swarm intelligence routing strategies. Another proposal for ant-based protocol using hybrid routing is the AntHocNet [5]. It has reactive and proactive components. Final ant-based solution considered here is the HOPNET [6], which is again a hybrid protocol. It is based on ant colony optimization (ACO) and zone routing framework for broadcasting. HOPNET is highly scalable for large networks. B. Bee-inspired routing protocols The bee-inspired protocol BeeAdHoc [1] focuses on energy efficiency. It is a reactive source routing algorithm which strives to achieve performance similar to established MANET protocols - DSR, AODV and DSDV but on the cost of significantly less energy. A. Types of Networks II. COMPARISON CRITERIA There are three distinctive types of network that are taken into account in this work: proactive, reactive, hybrid. In proactive protocols nodes in the network maintain routing information to all other nodes in the network by exchanging periodically routing information. Nodes with reactive protocols delay the route acquisition until there is a demand for a route. Hybrid protocols use a combination of both proactive and reactive strategies to gather routes to the destinations in a network. There can be different level of proactive and reactive routing involved, e.g. a node can collect proactively information for favourite destination nodes or nodes from the own area and use reactive routing for other nodes. ARA and SARA are on-demand routing protocols or in other words reactive. BeeAdHoc is reactive is as well. ANSI uses hybrid routing with both proactive and reactive components. The pure mobile nodes in ANSI use only reactive routing and choose routes deterministically, while more capable (immobile) nodes, part of network infrastructure use a combination of both proactive and reactive routing and perform stochastic (random) routing when multiple paths are available. AntHocNet is reactive in the route discovery phase and in case of route failure. For the route maintenance phase it acts proactively. The HOPNET algorithm [6] comprises two strategies - the local proactive route discovery within a node s neighbourhood and reactive communication between the neighbourhoods. So it can be derived it is a hybrid protocol. There are two types of routing tables maintained by HOPNET - Intrazone Routing Table (IntraRT) and Interzone Routing Table (InterRT). The IntraRT is proactively maintained so that a node has quick access to the paths within its zone. This is done by periodically sending out forward ants (FANTS) to sample the path within the own zone and determine topology changes (nodes moving away or entering the zone, link failures, etc.), then when the FANT reaches its destination, a corresponding backwards ant is sent back along the discovered path. The other table - InterRT stores the paths to nodes beyond its zone. The table is updated on demand as route outside one s own zone is required. The peripheral nodes of the zone are used to find routes between zones. B. Topology The ARA algorithm supports dynamic topology with multihop paths between nodes. Its optimization method is based on individual ants and local information from them. No routing information in tables has to be transmitted to neighbours or the whole network. The SARA topology is again dynamic. When it is formed only the shortest routes (with the minimum hop count) from all the broadcasted routes are stored by the network nodes. The topologies supported by ANSI and AntHoc can be constantly changing as well. In the following sections the mechanism supporting the changing topology will be explained into more detail (e.g. updates on routing information). The HOPNET network is divided into zones which are nodes local neighbourhoods. The size of the zone is determined by the number of hops and not locally by the radius length from a node. Therefore, a routing zone consists of the nodes and all other nodes within the specified radius length. A node may be within multiple overlapping zones and zones could vary in size. The nodes can be categorized as interior and boundary (or peripheral) nodes. The distance between the boundary nodes and the central node is the specified radius. All other nodes less than the radius are interior nodes. In Fig. 2, if the radius of the zone is 2 then for node S, nodes A, D, F, G are boundary nodes, and nodes B, E, C are interior nodes. All other nodes are exterior nodes (outside the zone). In order to construct a zone, a node, and determining border nodes, a node needs to know its local neighbours. This is achieved by a detection process based on replies to hello messages transmitted by each node. Fig. 2. HOPNET zones [6] BeeAdHoc is inspired the foraging principle of a honey bee colony [1]. Its Bee Agent Model consists of four types of agents: packers, scouts, foragers, and swarms. The packers function like the food-storer bees. They always reside inside the node, receive and store the data packets that come from the transport layer. Their main job is to find a forager for their data packet and they die once they hand over to the foragers. 2

3 The scouts discover new routes from their launching node to their destination node. A scout is broadcasted to all neighbour nodes. Once a scout reaches the destination, it finds back its route to the source. Foragers are the main workers in the BeeAdHoc algorithm. They receive the data packets from packers and then transport them to their destination. Each forager has a special type: delay or lifetime. The delay foragers collect delay information from the network, while the lifetime foragers collect remaining battery capacity of the nodes they visit. The first type tries to route packets along a path with minimum delay, while the second type selects such a path so that the life time of the network is increased. When a forager reaches its destination it remains there until it can be piggybacked to the source. This brings the optimization of less control packets, thus saves energy in the network. The swarms have a role when nodes do not need explicit acknowledgements for the received data packets. When the difference between incoming foragers from a certain node i and the outgoing foragers to the same node i reaches a certain threshold value at a node j, then the node j launches a swarm of foragers to the node i. When the swarm arrives at the node i then the foragers are extracted from the payload part and they are stored like they would have arrived at the node in a normal fashion. C. Phases In this section the three routing phases - discovery, maintenance and recovery of the regarded protocols will be presented with their more specific features. First is the,route Discovery phase, followed by the Route Maintenance phase, then the Route Recovery phase. 1) Route Discovery Phase: In this phase new routes are searched and established. In the case of ARA [2], the creation of new routes requires the use of forward ant (FANT) and backward ant (BANT). A FANT agent establishes the pheromone track to the source node, while a BANT establishes the pheromone track to the destination node. The FANT is a small packet baring unique sequence number. A forward ant is broadcasted by the source node s and is relayed by the neighbour nodes(fig. 3). A node that receives a FANT for the first time creates a record in its routing table, consisting of destination address, next hop, pheromone value. The pheromone value is computed based on the number of hops the FANT needed to reach the node. Then the node relays the FANT to its neighbours. Duplicate FANTs are identified through the unique sequence number and source address and are destroyed by the nodes. When the FANT reaches the destination node d, its information is extracted and the FANT itself is destroyed. Subsequently d creates a BANT and sends it the source s (Fig. 4). The BANT has the same task as the FANT, i.e. establishing a track to s. When the source node receives the BANT from the destination node, the path is established and data packets can be sent. By AntHocNet [5] a source node s checks its routing information if it has up-to-date route with the destination node and if not it sends a reactive forward ant similar to the ARA Fig. 3. Route discovery phase. A forward ant (F) is send from the sender (S) toward the destination node (D). The forward ant is relayed by other nodes, which initialize their routing table and the pheromone values. [2] Fig. 4. Route discovery phase. The backward ant (B) has the same task as the forward ant. It is send by the destination node toward the source node. [2] above, to look for paths to d. These ants gather information concerning the quality of the path they followed and then when they reach the destination they become backward ants which trace back their path and update routing tables. In the routing tables there is information on the goodness of the paths to each destination through each next hop. This is the pheromone. With its help multiple paths between s and d can be indicated, and packets can be routed as datagrams. The choice of the path happens stochastically - in each node the packets select the next hop with a probability proportional to the pheromone values. AntHocNet provides better ways to select which FANTS to be propagated and which to be discarded than ARA. It considers not only the number of hops but as well the time to reach a hop. By discarding some worse FANTS overhead is limited. By the route discovery phase SARA brings innovative approach compared to for example AntHocNet. Usually, as was seen above, the source node starts the route discovery process by sending a forward ANT, which is replicated by all network nodes until it reaches the destination node or neighbourhood. Upon receiving the first FANT, the destination sends a BANT through the shortest known path(s). If the packet arrives at the source then path is established and data flow may start. This approach requires two-way routing as with AntHocNet and creates significant amount of control information in case of long paths. The FANT is also replicated, creating flooding, which deteriorates performance. To cope with this SARA introduces a more efficient mechanism to disseminate FANTs: Controlled Neighbour Broadcast (CNB). Here each node broadcasts the FANT to all of its neighbours, who process it, but only one of them broadcasts the FANT again to its own neighbourhood. The selection of the node responsible for re-broadcasting the FANT is done in a probabilistic manner. HOPNET makes a separation in the route discovery phase 3

4 within a neighbourhood of nodes and between neighbourhoods (zones) [6]. Within a zone the discovery is proactive and between zones it is reactive. The route discovery phase within a zone is accomplished with the help of the intrazone routing table - IntraRT, which was mentioned earlier, and internal forward ants, which are different than the external forward ants. The internal forward ants are sent periodically to all neighbours to maintain the IntraRT. In the IntraRT there is information on the neighbour nodes in a zone - pheromone, visited times, hops, SeqNo. Pheromone is for the pheromone concentration of a link, visited times gives the number of times a link has been visited by the ants, hops is the number of hops between a nodes and another node in its zone. The format of the ants contains Source, Destination, SequenceNo, Type, Hops and Path. The sequence number which is present also in the IntraRT table is for distinguishing packets and avoid duplications. The Source field stores the source address. The Destination stores the destination address, which is left blank for internal forward ants and stores the destination node s address only for external ants. The Type field indicates one of five possible ant types: internal forward ant, external forward ant, backward ant, notification ant, error ant. The Hops field indicates the number of hops a forward ant can move. This field is assigned only for internal forward ants and is left blank for external. Path field represents the sequence of nodes between source and destination [6]. When a source node s has to send data to a destination node d, s first checks the columns of its IntraRT table if the destination lies in its zone. If this is the case then route discovery is done. If d is within the zone of s there will always be a route between the nodes, which is proactively maintained with periodically send ants. In the case when the source node s fails to find its destination within its zone, the InterRT table is used, which is responsible for the between zones route discovery phase. The InterRT contains paths to destinations with expiration time and sequence number (to avoid duplications). When a route outside a source s zone is searched it is checked in the table if this route has not been already discovered recently (before the expiration time), if this is the case, the source node sends its packet along the path that is pointed in the InterRT table. Otherwise external forward ants have to be sent. It is forwarded to peripheral nodes (which are known in the IntraRT table), and then possibly to other zones peripheral nodes until the destination s node is reached. Then the external forward ant transforms into a backward ant and functions similarly to the backward ants of the other ant-based protocols seen here. When the backward ant reaches its source, the path found by it is added to the InterRT table and communication between the source and destination may take place. ANSI has similar to HOPNET local proactive management and non-local reactive management. Its phases are not clearly described by its authors so its phases description will be skipped. In the case of BeeAdHoc, route discovery is done by the scout bees [1]. They are similar to FANTs and BANTs. As was explained above, a scout is broadcasted to all neighbour nodes with an expanding time to live timer (TTL), which controls the number of time the scout could be re-broadcasted. When it reaches the destination, it starts to trace back its route to the source (much like a BANT). One difference with ant-based algorithms though is that when a scout is at the source node, it recruits the foragers for its route by using the metaphor of dance, just like scout bees in nature. 2) Route Maintenance: By the route maintenance in most cases the ant-inspired protocols use the metaphor of pheromone. The most used links experience the highest pheromone levels, while the unused ones have the lowest levels. This happens with two complementary mechanisms of increasing and decreasing the pheromone intensity. The first happens with every packet that crosses a link, reaches its destination and then produces a backward link. The decreasing happens with time when no traffic happens through nodes. By this phase ARA utilizes fully the pheromone tracks established by FANTs and BANTs and updated by transported data packets. The protocol doesn t create special extra packets for route maintenance, the only improvement to the basic principle it adds is the duplicate packets elimination based on sequence numbering and source address. The routing maintenance phase of HOPNET [6], as described by its authors, corresponds rather to the route recovery (repair) phase as defined below in this work. For the route maintenance (as considered in this current section) HOPNET makes use of pheromone concentration like the other protocols presented. The pheromone values are updated as packets traverse from source to current node, which influences ants coming subsequently from the source. Because of this, the mechanism is described with the term source update algorithm [6]. AntHocNet - When the paths are established and a data session is running, s starts to send proactive forward ants to d. These ants follow the pheromone values as well and monitor the used paths. They also have a small probability of being broadcasted, so they can find new paths. The improvement that SARA brings in the route maintenance phase is in the case of asymmetric traffic (like UDP). Such traffic does not generate backward packets, which prevents maintaining correct pheromone levels. While AntHocNet solves this problem by generating additional control traffic, SARA uses special FANT messages calls super FANTs. The super FANT is generated at an end node (source or destination) where asymmetric traffic is detected. It has a pheromone reinforce equivalent capability of n standard FANTs. This super FANT is generated with lower frequency, than the arriving packet rate, which reduces overhead. Route maintenance by BeeAdHoc is executed partly by forager bees and partly by swarm bees, as far as their definitions are given. The foragers collect state information about the network depending on their type (delay foragers and lifetime foragers) and also bring optimizations (less control overhead). The swarms help in the case where no or scarce acknowledgement are sent by destination nodes. 3) Route Recovery: The route recovery is a process that is initiated when a broken link between two nodes is detected. The broken link may happen for several reasons, like a node 4

5 being turned off, failure in radio coverage or congestion that causes a higher number of collisions, etc. In MANET, these kinds of situations may occur frequently and thus, the route recovery procedure must be quickly executed with low overhead. ARA detects route failure if there is a missing acknowledgement [2]. If a node gets a route error message for a link, it deactivates the link by setting the pheromone value to 0. Then an alternative link is searched in the routing table. If the nodes finds such, it sends the packet vie this path. Otherwise the node informs its neighbours to check if they can relay the packets. Either they can or the backtracking continues to the source node. If the packet can not reach the destination, a new route discovery phase has to be initiated. To recover the route, AntHocNet and SARA will try to find alternative routes in the neighbourhood of the broken link. However, AntHocNet will attempt to reach the destination node with a standard FANT, but with a limited number of allowed broadcast transmissions, so that the ant is confined in the area of the destination. SARA also uses broadcast transmission with RFANT messages, but will attempt to find an alternative path that can reach the other side of the broken link instead of reach the destination node. If several nodes go down simultaneously it might not be possible to find an alternative route, using this local repair procedure. If the local repair procedure fails to succeed, an error message is sent to the source node and the route discovery procedure is initiated. As noted in the prior section, HOPNET s route recovery, as considered here, is described in the section Route maintenance [6] by its authors. If there is an invalid route within a zone it is periodically repaired proactively, because of the IntraRT table which is proactively maintained. If the route is between zones, the upstream node of the broken link will try a local repair procedure, aiming to find an alternative path to the destination, while buffering the incoming packets. If the node is successful in finding alternative path, it sends a notification ant to the source that lets it know of the route change. All nodes on the path of the notification change their inter routing tables as well to remove the invalid nodes. If this can not happen, an error ant is sent to the source, which initiates a new route discovery procedure. With its mechanisms HOPNET provides evolution of the methods adopted in ARA and AntHocNet presented before. By BeeAdHoc the recovery is possibly done by the foragers, but there is not enough information provided by its authors [1]. III. EXPERIMENTAL RESULTS AND COMPARISON TO WELL KNOWN ALGORITHMS LIKE AODV Most of the algorithms considered in this work are compared by their authors to classical routing solutions like AODV, DSR, etc. In this section some results are shown. A. Metrics for comparison In this work the most common metrics are used to measure the performance of the presented MANET routing protocols - packet delivery ratio, number of packets and end to end delay. 1) Packet Delivery: Packet delivery ratio (PDR) is the ratio of successfully delivered data packets to the total data packets sent from the source to the destination. 2) Overhead: Overhead is calculated by the ratio between the amount of information needed to carry control traffic over the total amount of traffic that has been sent. 3) End to end delay: This is the difference between the time once a packet is received by the application layer of a destination node and the time when a packet is originated at the application layer of a source node. B. ARA The ARA protocol is compared to the Ad-hoc On-demand Distance Vector (AODV), Destination-Sequenced Distance Vector (DSDV) and Dynamic Source Routing (DSR) [2].On Fig. 5 the fraction of successfully delivered packets is shown as function of pause time between which nodes move. It can be seen that in the case of higher dynamics only DSR outperforms ARA, but ARA s performance is still satisfying and in general its performance is better than AODV and DSDV. Fig. 5. Successfully delivered packets as a function of pause time [2]. C. SARA The Fig. 6 presents the overhead results of SARA compared to the non-swarm-intelligence protocol AODV, and two other protocols that are also analysed in this work - ARA and AntHocNet. The graphic shows that SARA presents the best performance, concerning overhead, which is almost independent of the nodes velocity. According to the authors of SARA [3], this fact is caused by the use of the Constant Bit Rate (CBR) and by the optimizations that were realized in the route maintenance. The use of a flooding mechanism and the continuous generation of control traffic for the active routes are responsible for the worst performance of AntHocNet, AODV and ARA. The use of a source-routing mechanism, where all the entire path information is carried by the FANT messages is also responsible for the worst performance exhibited by AntHocNet. These relative results are shared by UDP and FTP/TCP traffic Fig. 7. Nevertheless, smaller overhead values are experienced when FTP traffic is used, because the more significant amount of data traffic is generated. 5

6 Fig. 8. Packet delivery and delay. [4] ratio and average delay. AODV has only a little advantage when all packets are taken into account for the delay measurement, but if the 99-th percentile of the delay is taken it can be seen that the decrease in performance of AODV is mainly due to a small number of packets with a very high delay. Fig. 6. CBR Traffic [3] Fig. 9. Packet delivery and delay. [5] F. HOPNET On Fig. 10 the delivery ratio of HOPNET and AODV is shown. It can be seen that with fewer nodes HOPNET is worse than AODV. When there is a link failure HOPNET tries to pass the packets to other alternative proximate nodes and if there are less nodes in general this proves to be inefficient, but it is seen also that with increased number of nodes HOPNET performs better. Fig. 7. FTP Traffic [3] D. ANSI ANSI is compared to AODV similarly to the other protocols above. AODV is suited for hybrid ad hoc networks which makes it good for comparison with ANSI as well. The results presented by the authors of ANSI [4] show the performance of ANSI vs. AODV over a hybrid network using UDP flows. It can be concluded that ANSI consistently outperforms AODV in terms of packet delivery and delay (Fig. 8). The reason why ANSI performs better than AODV delivering more packets with better metrics such as delay and number of route errors is because ANSI manages the local network information better than AODV does, and performs congestion-aware routing. E. AntHocNet A comparison between AntHocNet and AODV shows (Fig. 9) that AntHocNet can outperform AODV in terms of delivery Fig. 10. Delivery ratio. [6] The control overhead of HOPNET Fig. 11 is generally higher than AODV, because AODV is purely reactive protocol and HOPNET is proactive within a zone. The proactive packets create the extra overhead. Fig. 12 shows a comparison between the end to end delay of HOPNET and AODV. It is clearly seen that HOPNET 6

7 Fig. 14, the energy expenditure is shown instead of overhead, because energy efficiency is one of the main motivations for BeeAdHoc and it can be clearly seen that BeeAdHoc has the lowest energy expenditure. Fig. 11. Routing overhead. [6] produces better results. This is due to the division in zones by HOPNET and the intrazone routing tables as well as the inerzone routing tables. With them the paths for transmission are readily accessible. Fig. 14. Energy expenditure. [1] Fig. 15, the delay is shown. The decreased delay with increased load is due to the fact that route discovery time is shared between more packets and thus it becomes less significant. BeeAdHoc shows lowest delays. Fig. 12. End to end delay. [6] Fig. 15. Packet delay. [1] G. BeeAdHoc Three comparison measurements are shown for BeeAdHoc depending on varying packet send rates. On Fig. 13 the packet delivery ratio is shown. Although all algorithms perform well with increased loads, BeeAdHoc shows the smallest decrease in success rates and keeps the second highest rates in general. IV. CONCLUSION A number of state of the art swarm-intelligence inspired MANET protocols are considered in this work and put to partial comparison. It is shown that by using ideas taken from the simple behaviour of ants and bees optimization and innovations in routing protocols can be done, that help outperform the standard MANET routing protocols like AODV, DSDV, DSR. Depending on application needs the presented protocols provide also customizing and tuning capabilities that can make them suitable for a wide range of MANET applications. REFERENCES Fig. 13. Packet delivery ratio. [1] [1] H. F. Wedde, M. Farooq, T. Pannenbaecker, B. Vogel, C. Mueller, J. Meth, and R. Jeruschkat, Beeadhoc: an energy efficient routing algorithm for mobile ad hoc networks inspired by bee behavior, in Proceedings of the 2005 conference on Genetic and evolutionary computation, ser. GECCO 05. New York, NY, USA: ACM, 2005, pp [Online]. Available: [2] M. Gunes, U. Sorges, and I. Bouazizi, Ara-the ant-colony based routing algorithm for manets, in Parallel Processing Workshops, Proceedings. International Conference on, 2002, pp

8 [3] F. Correia and T. Vazo, Simple ant routing algorithm strategies for a (multipurpose) manet model, Ad Hoc Networks, vol. 8, no. 8, pp , [Online]. Available: [4] S. Rajagopalan and C.-C. Shen, Ansi: A swarm intelligencebased unicast routing protocol for hybrid ad hoc networks, Journal of Systems Architecture, vol. 52, no. 8-9, pp , 2006, nature-inspired Applications and Systems. [Online]. Available: 4JS23DF-1/2/052fef335b630aa4885ebad05dbc1ef9 [5] G. D. Caro, F. Ducatelle, and L. M. Gambardella, Anthocnet: an ant-based hybrid routing algorithm for mobile ad hoc networks, in In Proceedings of Parallel Problem Solving from Nature (PPSN) VIII. Springer-Verlag, 2004, pp [6] J. Wang, E. Osagie, P. Thulasiraman, and R. K. Thulasiram, Hopnet: A hybrid ant colony optimization routing algorithm for mobile ad hoc network, Ad Hoc Networks, vol. 7, no. 4, pp , 2009, i. Bio-Inspired Computing and Communication in Wireless Ad Hoc and Sensor Networks; II. Underwater Networks. [Online]. Available: 8