1. On Reducing Broadcast Redundancy in Wireless Ad Hoc Network Author: Wei Lou, Student Member, IEEE, and Jie Wu, Senior Member, IEEE From IEEE transactions on mobile computing April-June 2002 Presented By 資管研一 R92725034 Lin Ming Yuan

2. Outline Introduction Preliminaries Enhanced dominant pruning algorithm Termination criteria Performance evaluation Conclusions

3. Introduction Some characteristics of ad hoc network  Without central infrastructure  Temporary and changing topology In this paper, the author focused on the topic of broadcast problem and try to find the minimum number of forward nodes.

4. Introduction (cont.) Traditionally it used the concept of the flood tree to broadcast packets in ad hoc networks. The efficiency of the algorithm depends on the number of total forwarding nodes. The importance and application of broadcast service  Route query process in several routing protocol  Send an error message to erase invalid routes  For reliable multicast

5. Introduction (cont.) The problem of finding minimum forwarding nodes can be reduce to a dominant set problem which is NP. Some previous algorithm  Blinding flooding (broadcast storming problem)  Dominating pruning (DP) algorithm

6. Introduction (cont.) The DP algorithm utilizes 2-hops neighborhood information to reduce redundancy transmissions and prolong the life of the network. The DP algorithm also can be considered as an approximation to the minimum flood tree problem. In this paper, the author proposed two extensive algorithm  TDP (Total Dominant Pruning algorithm)  PDP (Partial Dominant Pruning algorithm)

7. Outline Introduction Preliminaries  The approximation of MCDS (AMCDS) algorithm  The dominant pruning (DP) algorithm Enhanced dominant pruning algorithm Termination criteria Performance evaluation Conclusions

8. Preliminaries Lim and Kim prove that building a minimum flooding tree is the same as finding a minimum connected dominating set (MCDS) in a network, which is an NP-complete problem. Our approach is based on constructing a connected dominating set “on-the-fly” and it is suitable for dynamic networks with mobile hosts

9. Preliminaries (redundancy problem) U is the sender. The transmissions between v and w are redundant.

10. Outline Introduction Preliminaries  The approximation of MCDS (AMCDS) algorithm  The dominant pruning (DP) algorithm Enhanced dominant pruning algorithm Termination criteria Performance evaluation Conclusions

11. Notation G =(V,E), where V represents a set of wireless mobile hosts (nodes) and E represents a set of edges. Such a graph forms an unit disk graph. N(u) represents the neighbor set of u (including u) and N(N(u)) represents the neighbor set of N(u) (i.e., the set of nodes that are within two hops from u). Clearly, and if, then.

12. Assumption 2-hop neighborhood information can be obtained by periodic “Hello” packets, each of which contains the sender’s identification and the list of its neighbor. In this paper, the author assumed that v (sender) and u (receiver) are neighbors.

13. Outline Introduction Preliminaries  The approximation of MCDS (AMCDS) algorithm  The dominant pruning (DP) algorithm Enhanced dominant pruning algorithm Termination criteria Performance evaluation Conclusions

14. The approximation of MCDS (AMCDS) algorithm Step 1: At the start of the algorithm, all nodes are colored white and, then, the node with the maximum node degree is selected (put in set C) and colored black, and all of its neighbors are colored gray. Step 2: A recursive selection process runs until no white node exists: Choose a gray node that has the maximum number of white neighbors. Color the selected node black and its white neighbors gray.

15. AMCDS algorithm (cont.) The drawback of this algorithm is that it needs to know the global network topology and, therefore, it is not suitable for ad hoc wireless networks. The result of the AMCDS algorithm can be used as the lower bound to compare with algorithm.

16. Outline Introduction Preliminaries  The approximation of MCDS (AMCDS) algorithm  The dominant pruning (DP) algorithm Enhanced dominant pruning algorithm Termination criteria Performance evaluation Conclusions

17. The dominant pruning (DP) algorithm (selection process) 1. Let (empty list), (empty set), and, where for. 2. Find set S i whose size is maximum in K. (In case of a tie, the one with the smallest identification I is selected.) 3.,, and for all. If, exit; otherwise, go to step 2.

18. The dominant pruning (DP) algorithm (cont.) F(u,v) is the forward bode list between sender v and receiver u. B(u,v)=N(v)-N(u) to covers nodes in U(u,v)=N(N(v))-N(v)-N(u). Z is a subset of U(u,v) and S i is the neighbor set of vi. K is the set of S i. Specifically, the greedy set cover algorithm is used for the selection of forward node.

19. The dominant pruning (DP) algorithm (cont.) 1. Node v uses N(N(u)), N(u), and N(v) to obtain U(u, v) = N(N(v)) - N(u) - N(v) and B(u, v) = N(v) - N(u). 2. Node v then calls the selection process to determine F(u, v).

20. DP algorithm (graph.)

21. Outline Introduction Preliminaries Enhanced dominant pruning algorithm  The total dominant pruning (TDP) algorithm  The partial pruning dominant pruning (PDP) algorithm Termination criteria Performance evaluation Conclusions

22. The total dominant pruning (TDP) algorithm If node v can receive a packet piggybacked with N(N(u)) from node u, the 2-hop neighbor set that needs to be covered by v’s forward node list F is reduced to U = N(N(v)) – N(N(u)). The total dominant pruning (TDP) algorithm uses the above method to reduce the size of U and, hence, to reduce the size of F.

23. The total dominant pruning (TDP) algorithm (cont.) 1. Node v uses N(N(u)), N(u), and N(v) to obtain U(u, v) = N(N(v)) – N(N(u)) and B(u, v) = N(v) - N(u). 2. Node v then calls the selection process to determine F.

24. The total dominant pruning (TDP) algorithm (theorem) Theorem 1. If a node w 2 N(N(v)) is also in N(N(u)), then w can be excluded from U. Proof: consider all possible conditions of w  w is 1-hop neighbor of the node u, then it has received broadcast packet during the transmission of u and v.  w is 2-hop neighbor of the node u, then it will receive broadcast packets from the 1-hop neighbor of u like v.

25. The total dominant pruning (TDP) algorithm (theorem) Theorem 2. Let U = N(N(v)) - N(N(u)) and B = N(v) - N(u); then, U = N(B). Proof: by the concept of the complement set  with x= N(v) and Y = N(U)

26. The total dominant pruning (TDP) algorithm (graph)

27. Outline Introduction Preliminaries Enhanced dominant pruning algorithm  The total dominant pruning (TDP) algorithm  The partial pruning dominant pruning (PDP) algorithm Termination criteria Performance evaluation Conclusions

28. The partial pruning dominant pruning (PDP) Besides excluding N(u) and N(v) from N(N(v)), as addressed in the DP algorithm, more nodes can be excluded from N(N(v)). These nodes are the neighbors of each node in. Such a node set is donated as. Therefore, the 2-hop neighbor set U in the PDP algorithm is.

29. The partial pruning dominant pruning (PDP) (cont.) 1. Node v uses N(N(u)), N(u), and N(v) to obtain and U = N(N(u)) - N(u) - N(v) – P, and B = N(v) – N(u). 2. Node v then calls the selection process to determine F.

30. The partial pruning dominant pruning (PDP) (theorem) Theorem 3. Let ; U = N(N(v)) - N(u) - N(v) – P and B = N(u) – N(v), then. Proof: by the concept of the set subtraction and with X = N(u) and Y = N(v). So, N(B) can cover U.

31. The partial pruning dominant pruning (PDP) (graph)

32. Example

33. Result

34. Result (cont.)

35. As the lower bound by using the AMCDS algorithm, the minimum connected dominating set is {2, 6, 7, 11}, so the number of forward nodes is 4. The number of the original DP is 8, TDP’s is 5 and PDP’s is 6. (The more near global information, the better performance.)

36. Outline Introduction Preliminaries Enhanced dominant pruning algorithm Termination criteria Performance evaluation Conclusions

37. Termination criteria The first one assigns a marked/unmarked status to each node. A node v is called marked if v has received a packet; otherwise, v is called unmarked. We assume that, v knows the current marked/unmarked status of the nodes in N(v) at the time v decides its forward node list. When all nodes in N(v) are marked, v will stop rebroadcasting and discard the packet.

38. Termination criteria (cont.) The second approach assigns a relayed/ unrelayed status to each node. A node v is called relayed when v has sent a packet; otherwise, v is called unrelayed. Forward node v will stop rebroadcasting a packet only when v has sent that packet.

39. Outline Introduction Preliminaries Enhanced dominant pruning algorithm Termination criteria Performance evaluation Conclusions

40. Performance evaluation Static environment 400 randomly generated graph and parameters  r : the fixed transmitter range  d : the fixed average node degree (density) No contention in MAC layer

41. Performance evaluation (No. of forward node) Transmission range=25/40 and use marked/unmarked approach 15% improvement

42. Performance evaluation (No. of forward node) Transmission range=55/70 and use marked/unmarked approach 20% improvement

43. Performance evaluation (No. of forward node) Transmission range=25/40 and use relayed/unrelayed approach

44. Performance evaluation (No. of forward node) Average degree=6/10 and use marked/unmarked approach

45. Performance evaluation (No. of forward node) Average degree=6/10 and use relayed/unrelayed approach

46. Performance evaluation (No. of received packets) Transmission range=25/40 and use marked/unmarked approach

47. Performance evaluation (No. of received packets) Transmission range=25/40 and use relayed/unrelayed approach

48. Performance evaluation (No. of received packets) Average degree=6/10 and use marked/unmarked approach

49. Performance evaluation (No. of received packets) Average degree=6/10 and use relayed/unrelayed approach

50. Performance evaluation (broadcast delivery rate) Transmission range=25/40 and No. of nodes=100 X axis represents the speed of the nodes.

51. Explanation of the result The larger transmission range the more covered neighborhood node information and can reduce more redundancy forward nodes. The higher degree the more redundancy transmission. Broadcast rate decreases as the speed of each node increases.

52. Explanation of the result (cont.) Performance : AMCDS>TDP>PDP>DP The marked/unmarked approach contains more neighbor information than relayed/ unrelayed approach and is better.

53. Outline Introduction Preliminaries Enhanced dominant pruning algorithm Termination criteria Performance evaluation Conclusions

54. Original DP algorithm and improved TDP and PDP algorithm. Trade-off between broadcast redundancy (v.s the life of the ad hoc network) and broadcast delivery rate. Extend the proposed schema from 2-hops to k- hops.