Probabilistic Broadcasting Based on Coverage Area and Neighbor Confirmation in Mobile Ad Hoc Networks Jae-soo Kim
Qi Zhang and Dharma P. Agrawal
Department of Computer Engineering Sangju National University Sangju ,South Korea
[email protected]
OBR Center for Distributed and Mobile Computing, ECECS University of Cincinnati Cincinnati, OH {qzhang, dpa}@ececs.uc.edu available. For instance, AODV (Ad Hoc On-demand Distance Vector Routing) protocol adopts broadcasting mechanism as a route request in MANET [1, 2].
Abstract—Broadcasting is a fundamental and effective data dissemination mechanism for route discovery, address resolution and many other network services in ad hoc networks. While data broadcasting has many advantages, it also causes some problems such as the broadcast storm problem, which is characterized by redundant retransmission, collision, and contention. Although many approaches have been proposed to solve them, none of them guarantees the lowest bound. In this paper, we propose a probabilistic broadcasting based on coverage area and neighbor confirmation in mobile ad hoc networks. We use the coverage area of a node to adjust the rebroadcast probability. If a mobile node is located in the area closer to sender, which means it has small additional coverage and rebroadcast from this node can reach less additional nodes, so its rebroadcast probability will be set lower. On the other hand, if a mobile node is located in the area far from sender, which means that the additional coverage from this node is large, its rebroadcast probability will be set higher. The coverage area can be estimated from the distance between sender and receiver and the distance can be estimated by signal strength or global positional system. Our approach combines the advantages of probabilistic and area based approach. Simulation results show that our approach can improve the average performance of broadcasting in various network scenarios. Our approach is simple and can be easily implemented in MANET. MANET;
Many approaches [3, 4, 5, 6, 7, 8, 9, 10, 11, and 12] are proposed for broadcasting in MANET. But none of them have been considered as an optimal method for the broadcasting. The simplest one to achieve broadcasting is through flooding. Even though flooding is very simple and reliable approach, it produces a high overhead in the network. Therefore, it leads to excessive contention, collision and redundant rebroadcasts. More sophisticated solutions such as probability-based, counter-based, distance-based, location-based, and neighborknowledge-based approaches have been proposed to overcome the drawbacks of flooding. Probability-based approach is another simple one. It depends upon pre-defined fixed probability to determine whether it rebroadcast the packets or not. One problem of the probabilistic approach is how to set the rebroadcast probability. It is demonstrated in [4 and 5] that the optimal rebroadcast probability is around 0.65. Intuitively, this value does not seem globally optimal for different node density and relative location from the sender. For example, the mobile hosts close to sender will have more neighbors whose coverage areas significantly overlap. So, the rebroadcast probability of a node in MANET should be set dynamically according to its circumstances [6, 7].
As the results of advances in wireless communication technology, portable computers with wireless interfaces can communicate among themselves. It is argued that future wireless network will be converged to be more easily reconfigurable situations such as Mobile Ad hoc network (MANET). MANET is a special type of wireless mobile network in which mobile hosts can communicate without any aid of established infrastructure and can be deployed for many applications such as battlefield, disaster relief and rescue, etc. Broadcasting is to transmit a message from a source to all the other nodes in the network. It is widely used to resolve many network layer problems. In a MANET, in particular, due to host mobility, broadcastings can be applied to many areas, such as paging a particular host, sending an alarm signal, and finding a route to a particular host, etc. Several ad hoc network protocols assume that the broadcasting service is basically
In this paper, we propose a dynamic probabilistic broadcasting approach with coverage area and neighbor confirmation. We use the coverage area concept to adjust the rebroadcast probability of a node. If a mobile node is located in the area close to sender, which means it has small additional coverage area and its neighbors may receive the same broadcasting message from others, thus its rebroadcast probability will be set lower. On the contrary, if a mobile node is located in the area far from sender, which means its additional coverage area is large. Then its rebroadcast probability will be set higher. The coverage area can be estimated from the distance between sender and receiver node, and the distance can be estimated by signal strength or global positional system (GPS). We present simple coverage ratio related by distance in Section III. We also compare our approach with simple flooding approach under various network conditions through simulation. Simulation results show that our approach can improve the average performance of broadcasting in various network scenarios.
Keywords-Ad Hoc probabilistic broadcasting
I.
Network;
broadcasting;
INTRODUCTION
IEEE Communications Society Globecom 2004 Workshops
96
0-7803-8798-8/04/$20.00 ©2004 IEEE
The rest of this paper is organized as follows: In Section II, we introduce the background and related work of broadcasting in MANET. In Section III, we describe our approach, dynamic probabilistic broadcasting based on coverage area and neighbor confirmation, highlighting the differences from other similar approaches. In Section IV, we simulate our approach by using GloMoSim v2.03, present the simulation results, and compare its performance with flooding. In section V, we conclude the paper and present directions for future work. II.
whether there is any node in the calculated coverage area. So, some nodes may not receive broadcasting packets. Neighbor knowledge scheme maintains neighbor node information to decide whether it or the neighboring nodes have to rebroadcast or not. To use neighbor knowledge method, each node has to explicitly exchange neighborhood information among mobile hosts using periodic Hello packets. The length of the period affects the performance of this scheme: If it is set too short then it could cause collision or contention while setting it too long would degrade its ability to cope with mobility.
RELATED WORK
Haas et al. [6] proposed Gossip-Based ad hoc routing. Four different gossip versions are presented for ad hoc routing. The most basic scheme is GOSSIP1(p, k), and it rebroadcast with probability 1 until it reaches K hops. When it exceeds K hops, it will act the same as the static probability scheme, rebroadcast with a predefined probability p. GOSSIP2 is presented to improve the adaptability for changing density. GOSSIP2(p1, k, p2, n) determines its probability using the number of neighbor nodes. If a node has a smaller number of neighbors than threshold n, it decides to rebroadcast with probability p2 which is greater than p1. Otherwise, it rebroadcasts with probability p1. To prevent the “die out” of a message with too low probability, GOSSIP3 is presented. GOSSIP3(pψkψm) is just like GOSSIP1(pψk), except for the following modification. A node that originally did not broadcast a received message and did not get the message from at least mψother nodes within some timeout period broadcasts the message immediately after the timeout period. The zone concept is applied to GOSSIP4. Zone consists of all the nodes that are at most ρψhops away from u, for some appropriately chosen zone radius ρ . GOSSIP4(pψkψk’) is just like GOSSIP1(pψk), except that each node has a zone of radius k’.
One of the earliest broadcast mechanisms is flooding, where every node in the network retransmits a message to its neighbors upon receiving it for the first time. Although flooding is extremely simple and easy to implement, it can be very costly and can lead to serious problem, named as broadcast storm problem, which is characterized by redundant packet retransmissions, network bandwidth contention and collision. Ni et al. [4] studied the flooding protocol analytically and experimentally and showed that a rebroadcast can provide only 61% additional coverage at most and only 41% additional coverage in average over that already covered by the previous transmission. So, rebroadcasts are very costly and should be used with caution. Ni et al. [4] also classified broadcasting schemes into five classes to reduce redundancy, contention, and collision: probabilistic, counter-based, distance-based, location-based and cluster-based. In probabilistic scheme, a mobile host rebroadcasts packets according to a certain probability. In counter-based scheme, a node determines whether it rebroadcast a packet or not by counting how many identical packets it receives during a random delay. It is assumed in counter-based scheme that the expected additional coverage is so small that rebroadcast would be in vain when the number of recipient broadcasting packets exceeds a threshold value. In distance-based scheme, they used the relative distance between a mobile node and previous sender to make the decision whether it rebroadcast a packet or not. In location-based scheme, additional coverage concept is used to decide whether to rebroadcast a packet or not. Additional coverage is acquired by the locations of broadcasting hosts using the geographical information of a MANET. In cluster-based scheme, MANET is divided into clusters, which is a set of mobile hosts. There are one cluster head and several gateways in a cluster. Cluster head is representative of a cluster and its rebroadcast can cover all hosts in that cluster. Only gateways can communicate with other clusters and have responsibilities to propagate the broadcast message.
Although the goal of all the above protocols is to minimize the number of retransmissions, none of them guarantee the best suited bound of retransmissions. In general, Neighbor knowledge methods perform better than Area based methods, while Area based methods perform better than Probability based methods. This is due to the complexity and increased overhead of the complex schemes. Our approach combines the advantages of probabilistic and area-based approach, and it has higher throughput, better reachability, and lower latency as compared with general probabilistic or area-based approach. Moreover, it is simple enough for easy implementation. We describe the details of our approach in Section III. III. DYNAMIC PROBABILISTIC BROADCASTING WITH COVERAGE AREA AND NEIGHBOR CONFIRMATION In this section we describe our protocol in detail. Probability based methods use certain predefined rebroadcast probability p (0 < p ≤ 1) to decide whether to rebroadcast or not. The biggest problem with these schemes is determining the suitable probability. We propose Probabilistic Broadcasting with Coverage Area and Neighbor Confirmation.
Williams et al. [5] classified the broadcasting techniques into four groups and compared their performances: simple flooding, probability-based, area-based and neighbor knowledge scheme. In flooding scheme, every node in the network retransmits the message to its neighbors after receiving it. Probability-based scheme is a very simple way of reducing rebroadcasts. Each node rebroadcasts with a predefined probability p, where p =1 activates blind flooding. In areabased scheme, a node determines whether it rebroadcast a packet or not by calculating its additional coverage area. Although area-based scheme works quite well, it doesn’t know IEEE Communications Society Globecom 2004 Workshops
A. Shadowing Effect As described in the previous section, the goal of our protocol is to achieve high reachability of broadcasting and reduction of rebroadcast. In a general probabilistic approach,
97
0-7803-8798-8/04/$20.00 ©2004 IEEE
the rebroadcast probabilities of all mobile nodes are fixed as same value. However, the same rebroadcast probabilities can’t reflect the situation of MANETs which topologies are frequently changed. For example, a MANET may be sparse or dense, and relative location of the rebroadcast node is far from or closer to the sender. If these circumstances are not taken into consideration, the rebroadcast probabilities value might be set too small or too large, and after all, the reachability will be poor or many rebroadcast packets will be generated. So, the rebroadcast probabilities need to be adjusted by the circumstances of the node.
acquire good output. Figure 2 shows how to get the rebroadcast probability that is proportional to the distance. Let R be the communication range of node s, and A be the covering area of node s. A can be obtained by the equation: A = πR 2. A1, A2, and A3 is sub area of A with radii r1, r2, and r3, respectively. We define the coverage ratio μψas follows: µ=
We allow each node to choose different probability according to its distance from the sender. The distance from a node to the sender can be calculated from the signal strength or GPS (Global Positioning System). In short of our protocol, each node examines how far it is from the sender and determines its retransmission probability. It is better for a node that is farther away from the sender to have high retransmission probability. This means that a node geographically further from the sender may potentially act as a relay node on behalf of a node closer to sender. Note that a node closer to sender might be shadowed from rebroadcast nodes in the extreme case.
A1 , A1 + A2 + A3 A2 , A1 + A2 + A3 A3 , A1 + A2 + A3
For Area 1 For Area 2 For Area 3
When a node receives the broadcast message, its relative distance can be obtained by comparing the signal strength with maximum signal strength or by using GPS. Then the distance to sender can be estimated and the node can determine its coverage ratio μ. We can get the rebroadcast probability (p) as follows.
p = αµ where, a is a sensitivity parameter to control the rebroadcast probability. This approach gives a good approximation in the general distribution of mobile node.
Figure 1 shows the shadowing effect of our approach. In Figure 1, the neighbors (node n1, n2, n3 and n4) that are geographically far away from the sender node (s) will act as relays with high retransmission probability. On the other hand, the neighbors (node n5 and n6) that are geographically close to the sender node (s) will be shadowed from relays with low retransmission probability. This shadowing effect will reduce the number of rebroadcast packets.
A r r1
s
r2
r3
A1 A2 A3
n1 n5 s n2
Figure 2. Rebroadcast probability is proportional to the distance from sender.
n4
C. Dynamic Probabilistic Rebroadcast with Coverage Area and Neighbor Confirmation An important problem is how to minimize the number of rebroadcast packets while retransmission latency and packets reachability are maintained appropriately. Even though a large number of rebroadcasts guarantees high reachability, it causes high network bandwidth wastage and many packets collisions. On the other hand, the small number of rebroadcasts results in low reachability, because it cause rebroadcast chain broken so that some hosts may not receive the broadcast packets.
n6 n3
Figure 1. shadowing effect of inner nodes
B. Dynamic Probabilistic Rebroadcast with Coverage Area Our scheme is based on the shadowing effect. When a node receives a broadcasting packet, it refers to its distance from sender to determine its rebroadcast probability. As explained above, if a node is an outer node, the node assumes that it has larger coverage area, and if a node is inner node, the node assumes that it has smaller coverage area. As the coverage area is proportional to the distances to sender, the rebroadcast probability of a node should be considered according to its coverage area. Our scheme uses its coverage area to determine its rebroadcast probability.
The probabilistic approached presented above may cancel a non-redundant packets retransmission. Figure 3 shows the distribution of mobile node in MANET, where a) is good distribution and b) is bad distribution. In the case of b), early die-out of rebroadcast may happen. When a non-redundant retransmission is dropped in first few steps, the number of unreachable nodes increases fast hop by hop.
We introduce simple equation that defines the relationship between the coverage area and rebroadcast probability to IEEE Communications Society Globecom 2004 Workshops
98
0-7803-8798-8/04/$20.00 ©2004 IEEE
N-1 possible rebroadcasts, where N is the total number of nodes. In general probabilistic approach, each node decides to rebroadcast or not according to a fixed probability P. Since their decisions are independent, the total number of rebroadcasts is P*(N-1) on average. In our approach, the rebroadcast probability is dynamically set according to the distance to sender. While the probability of a node far from the sender is set to high, the probability of a node close to the sender is set to low. Figures 4 and 5 show the number of relay nodes to transmit broadcast message without mobility and host mobility of 20m/sec. As shown in Figures 4 and 5, our approach can substantially reduce the number of rebroadcasts in other approach. This indicates that our approach is the most efficient among the four algorithms.
n2 n1 s
s
a) Good Distribution
b) Bad Distribution
Figure 3. The distribution of mobile nodes in MANET
To prevent from early die-out of rebroadcast, neighbor confirmation scheme is applied. The neighbor confirmation scheme is to get high reachability to the last nodes by second retransmission of a packet. It is executed by the nodes which do not rebroadcast the packet according to dynamic probability with coverage area. After a given amount of time t, a node checks if all the neighbors have received the broadcast packet. If not, it rebroadcast the packet. Let’s assume that n1 in Figure 3 has the probability that doesn’t not rebroadcast the packet. And also assume that n2 is one of the n1’s neighbors, and n2 doesn’t not receive broadcast packet. n1 waits a given amount of time t, and checks if all the neighbors have received the broadcast packet. If n2 didn’t receive the packet, n1 must rebroadcast it. IV.
1200 Flooding alpha=1.0 1000
alpha=2.0
Relay
800
600
400
200
0 10
15
20
25
30
35
40
45
50
Number of Connections
Figure 4. Number of retransmission (No mobility)
PERFORMANCE EVALUATION
In this section, we evaluate the performance of our algorithm by comparing with a simple flooding algorithm. We use network simulator Glomosim [13] to evaluate our scheme. All the simulations were performed in a simulation area of 2000m x 2000m, and the number of connections is varied from 10 to 50. Other simulation parameters are shown in Table 1.
600 Flooding alpha=1.0
500
alpha=2.0
Relay
400
300
200
TABLE I.
SIMULATION PARAMETERS
Simulation Parameter Simulator Network Range Transmission Range Number of Connections Bandwidth Traffic Type Packet Rate Packet Size Simulation Time Number of Trials
100
Value
0 10
20
25
30
35
40
45
50
Number of Connections
GloMoSim (v2.03) 2000m X 2000m 250m 10, 20, 30, 40 and 50 2Mbps CBR (Constant Bit Rate) 10 packets per second 512 bytes 90s 10
Figure 5. Number of retransmission (Mobility = 20m/sec)
B. Reachability Flooding approach guarantees that all nodes can receive the broadcast packets in the expense of redundant rebroadcasts if no packet is dropped. But, in probabilistic approach, some packets will be dropped with their probability, which may cause some mobile node will not receive broadcast packets. We define reachability as the number of mobile hosts receiving the broadcast message i.e. number of mobile hosts that are reachable directly or indirectly. Figures 6 and 7 show the simulation results for reachability with no mobility and host mobility of 20m/sec. It shows that there is no significant difference with respect to the reachability of flooding approaches. Flooding has best reachability. This indicates that our scheme is a good solution for maintaining high reachability. High reachability is the result of neighbor confirmation. We can also infer that our approach can achieve high performance without sacrificing reachability.
We used three kinds of measures: rebroadcast number, reachability, and collision, to evaluate the performance of our approach. Rebroadcast number is the number of rebroadcast packets. Reachability is the sum of mobile node that receives the broadcast message directly or indirectly. Collision is the number of collided packet that mobile nodes send. A. Rebroadcast Number In flooding, a mobile node rebroadcasts all routing request packets that are received for the first time. Therefore, there are
IEEE Communications Society Globecom 2004 Workshops
15
99
0-7803-8798-8/04/$20.00 ©2004 IEEE
these intervals cause the overall delay of broadcast packets in our scheme. We plan to pursue to reduce the overall delay of broadcast packets in our further research.
120 Flooding alpha=1.0 alpha=2.0
100
80
2500
alpha=1.0 alpha=2.0
2000 40
Collisions
Reachability
Flooding 60
20
1500
1000
0 10
15
20
25
30
35
40
45
50
Number of Connections 500
Figure 6. Reachability (No Mobility) 0 10
15
20
25
30
35
40
45
50
Number of Connections 45
Figure 8. Collision (No Mobility)
40 35 30
1400 Flooding alpha=1.0
1200 20
alpha=2.0 1000
15 Flooding
10
alpha=1.0 5
800
Collisions
Reachability
25
alpha=2.0
600
0 10
15
20
25
30
35
40
45
50
400
Number of Connections
200
Figure 7. Reachability (Mobility = 20m/sec)
0 10
15
20
25
30
35
40
45
50
Number of Connections
D. Delay We also measured the delay of broadcast packets to reach all of the nodes in network. For the simulation of the delay, the start times are recorded when source node send broadcast packets as well as the end times are recorded when the broadcast packet reaches the last node. The broadcast delay is defined as the time difference between these two values. Figure 10 and 11 shows the simulation result of delay with no mobility and host mobility of 20m/sec. In Figures 10 and 11, our scheme shows more delay time than flooding. Our scheme use neighbor confirmation to get high reachability. Neighbor confirmation is conducted by the node which doesn’t rebroadcast after some random intervals. It is analyzed that IEEE Communications Society Globecom 2004 Workshops
Figure 9. Collision (Mobility = 20m/sec)
12
10
8
Delay (s)
C. Collision As mentioned above, collision is measured as the number of collision packet that mobile nodes send. Consider the scenario where several neighbor nodes hear a broadcast from node x. After hearing broadcast message, they (i.e. neighbor nodes of x) all try to start rebroadcasting at around the same time. Flooding scheme causes enormous collisions. Most broadcast methods obviously aim to reduce the collision. Figures 8 and 9 show the simulation result of our scheme for collision with no mobility and host mobility of 20m/sec. We can see that our scheme can reduce the collision packets more than 50 % average compared with flooding. Reducing the collision regardless of the reachability drop would be meaningless, since it would be omitting the original concept of broadcasting. The results show that our scheme has good performance without sacrificing high reachability.
6
4
Flooding 2 alpha=1.0 alpha=2.0 0 10
15
20
25
30
35
40
45
50
Number of Connections
Figure 10. Delay (No Mobility)
12
10
Delay (s)
8
6
4
Flooding
2
alpha=1.0 alpha=2.0 0 10
15
20
25
30
35
40
45
50
Number of Connections
Figure 11. Delay (Mobility = 20m/sec)
100
0-7803-8798-8/04/$20.00 ©2004 IEEE
In this paper we introduced a dynamic probabilistic broadcasting approach with coverage area and neighbors confirmation for MANETs. Our scheme combines probabilistic approach with the area-based approach. A mobile host can dynamically adjust the value of the rebroadcast probability according to its additional coverage in its neighborhood. The additional coverage is estimated by the distance from the sender. Our scheme combines neighbor confirmation concept to prevent early die-out of rebroadcast. Our simulation results show this approach generates fewer rebroadcasts than flooding approach. It also incurs lower broadcast collision without sacrificing high reachability. We plan to build more elaborated dynamic probabilistic approach considering neighbor density in order to facilitate optimal adaptation strategy. We believe that further saving may be achieved in rebroadcast and collision by improving the proposed algorithm.
E. Throughput It is important to evaluate the overall throughput of a protocol. We define throughput as the amount of broadcast data (bits) transmitted during a second in the MANET. To evaluate the throughput of our scheme, we measured how many data (bits) are transmitted per second. Figures 12 and 13 show the simulation result of throughput with no mobility and host mobility of 20m/sec. From our results, we observe that the average data transmission rate (bit per second) of our scheme is more than twice as much speed as flooding. It is analyzed that these good throughput is the results of decreased rebroadcast node and reduced collision packets in our scheme. Even though the transmission delay of our scheme, which was described in Figures 12 and 13, is larger than flooding, they outperform the deficiency of delay in the aspect of overall throughput.
REFERENCES
45000 Flooding 40000
alpha=1.0
[1]
alpha=2.0 35000
Throughput
30000
[2]
25000
20000
[3]
15000
10000
5000
[4]
0 10
15
20
25
30
35
40
45
50
Number of Connections
[5]
Figure 12. Throughput (No Mobility)
80000
[6]
Flooding 70000
alpha=1.0 alpha=2.0
[7]
60000
Throughput
50000
40000
[8]
30000
20000
10000
[9]
0 10
15
20
25
30
35
40
45
50
Number of Connections
Figure 13. Throughput (Mobility = 20m/sec)
V.
[10]
CONCLUSION [11]
Broadcasting is an active research topic in MANETs. An important problem is how to minimize the number of rebroadcast packets while good retransmission latency and packets reachability are maintained. Even though the large number of rebroadcasts guarantees high reachability, it causes high network bandwidth wastage and so many packets collisions. On the other hand, the small number of rebroadcasts results in low reachability, because it cause rebroadcast chain broken so that some hosts may not receive the broadcast packets.
IEEE Communications Society Globecom 2004 Workshops
[12]
[13]
101
D. P. Agrawal and Q-A Zeng. “Introduction to Wireless and Mobile Systems,” Brooks/Cole Publishing, ISBN No. 0534-40851-6, 436 pages, 2003. S. Giordano and W. W. Lu, “Challenges in mobile ad hoc networking,” IEEE Communications Magazine, vol. 39, no. 6, pp. 129–181, June 2001. W. Lou and J. Wu, “On reducing broadcast redundancy in ad hoc wireless networks,” IEEE Transactions on Mobile Computing, vol. 1, no. 2, pp. 111–123, Apr.-June 2002. Y. C. Tseng, S. Y. Ni, Y. S. Chen, and J. P. Sheu, “The broadcast storm problem in a mobile ad hoc network,” Wireless Networks, vol. 8, no. 2/3, pp. 153–167, Mar.-May 2002. B. Williams and T. Camp, “Comparison of broadcasting techniques for mobile ad hoc networks,” Proceedings of the Third ACM International Symposium on Mobile Ad Hoc Networking & Computing (MobiHoc), 2002, pp. 194–205 Z. Haas, J. Halpern, and L. Li, “Gossip-based Ad Hoc Routing,” Proceedings of IEEE INFOCOM 2002. Y. Sasson, D. Cavin, and A. Schiper, “Probabilistic Broadcast for Flooding in Wireless Mobile Ad hoc Networks,” IEEE Wireless Communications and Networking Conference (WCNC), March 2003 M. Su, W. Feng, and T.H. Lai, "Location-aided Broadcast for Wireless Ad Hoc Network Systems," Special Issue of Kluwer MONET on Algorithmic Solutions for Wireless, Mobile, Ad Hoc and Sensor Networks, 2003. J. Cartigny and D. Simplot, “Border Node Retransmission Based Probabilistic Broadcast Protocols in Ad-Hoc Networks,” 36th Annual Hawaii International Conference on System Sciences (HICSS'03), January 2003 H. Lim and C. Kim, “Multicast tree construction and flooding in wireless ad hoc netowrks,’’ in Proceedings of the Third ACM International Workshop on Modeling, Analysis and Simulation of Wireless and Mobile Systems (MSWiM), Aug. 2000. W. Peng and X. Lu, “AHBP: An efficient broadcast protocol for mobile ad hoc netowrks,” Journal of Science and Technology, Beijing, China, 2002. A. Qayyum, L. Viennot, and L. Laouiti, “Multipoing relaying: An efficient technique or flooding in mobile wireless netowrks,” INRIARapport de recherche, Tech. Rep. 3898, 2000. X. Zeng, R. Bagrodia and M. Gerla, "GloMoSim: a Library for Parallel Simulation of Large-scale Wireless Networks", in Proceedings of the 12th Workshop on Parallel and Distributed Simulations. PADS '98, May 26-29, 1998 in Banff, Alberta, Canada.
0-7803-8798-8/04/$20.00 ©2004 IEEE
