1 / 56

An Energy and Delay Efficient Scheduling Algorithm for Data-Centric Wireless Sensor Networks

國立臺灣大學資訊管理學研究所碩士論文審查. An Energy and Delay Efficient Scheduling Algorithm for Data-Centric Wireless Sensor Networks. 具資料集縮能力無線感測網路之 低能耗與延遲排程演算法. 指導教授:林永松 博士 研 究 生:王弘翕 中華民國 95年 7月 27日. Outline. Introduction Problem Description Getting Primal Feasible Solution Experiment Result Conclusion.

winkfield
Télécharger la présentation

An Energy and Delay Efficient Scheduling Algorithm for Data-Centric Wireless Sensor Networks

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. 國立臺灣大學資訊管理學研究所碩士論文審查 An Energy and Delay Efficient Scheduling Algorithm for Data-Centric Wireless Sensor Networks 具資料集縮能力無線感測網路之低能耗與延遲排程演算法 指導教授:林永松 博士 研 究 生:王弘翕中華民國 95年 7月 27日

  2. Outline • Introduction • Problem Description • Getting Primal Feasible Solution • Experiment Result • Conclusion

  3. Background • Wireless Sensor Networks (WSNs) • WSNs consist of a number of small nodes with sensing, computation, and wireless communication abilities. • Each sensor in the WSNs is capable of probing and collecting environmental information such as temperature, ocean current, and atmospheric pressure. • The deployment of these sensors would be typically in a random fashion. • Recharging the batteries of a moribund sensor would not be feasible. • Several principal issues of WSNs • Data Aggregation • Routing • Minimize End-to-end Delay

  4. Background(cont’d) • Data-centric Routing Algorithms * • Shortest Path Tree (SPT) • Center Nearest Source (CNS) * B. K., D. E. & S. W., “Modeling Data-Centric Routing in Wireless Sensor Networks”

  5. Active Active Active Sleep Sleep S-MAC Active Active Active Sleep Sleep T-MAC TA TA TA Background(cont’d) • Data-centric Routing Algorithms • S-MAC * • T-MAC ** • D-MAC *** * S-Mac : W. Y., J. H., & D. E., “An energy-efficient MAC protocol for wireless sensor networks.” ** T-Mac : T. V. D., & K. G. L “An Adaptive Energy-Efficient MAC Protocol for Wireless Sensor Networks *** G. L., B. K., & C. R., “An Adaptive Energy-Efficient and Low-Latency MAC for Data Gathering in Sensor Network”

  6. Sink S S S Motivation • Drawback of the existing protocols • SPT, CNS : idle listening • S-MAC, T-MAC : sleep latency • D-MAC : data aggregation • The tradeoff between energy consumption and latency

  7. Problem Description Routing Problem Sensor Node Source Node Sink Node

  8. Problem Description(cont’d) Scheduling Problem

  9. Problem Description (cont’d) • Assumption • There is a centralized node to determine the activities of all sensors • A network with strict synchronization • Propagation delay can be ignored • Link delay can be considered as a constant • Given • The set of all sensor nodes • The set of all data sources • The sink node • The set of all candidate paths for each data source to reach sink node • Longest hops along shortest path from sink node to reach the farthest data source • An arbitrary large number M • End-to-end delay requirement T • Objective: • To minimize the energy consumption of the entire wireless sensor network

  10. Problem Description(cont’d) • Subject to: • Routing constraint • Tree constraint • Number of Neighbors constraint • Collision Free constraint • To determine: • Routing path for each data source • Transmission radius for each sensor node • Whether a link should be on the data aggregation tree • The data aggregation tree • The wake up time of each sensor node on data aggregation tree • The aggregation complete time of each sensor node on data aggregation tree • The transmission finish time of each sensor node on data aggregation tree

  11. Notation – Decision Variables

  12. Notation – Decision Variables(cont’d)

  13. Sleeping Aggregation Sending Idling Receiving Routing Constraints Tree Constraints Problem Formulation Objective Function Zip = min Subject to:

  14. Scheduling Constraints Radii Constraints Problem Formulation

  15. Collision Free Constraints Number of Neighbors Constraints Problem Formulation Go

  16. Lagrangean Relaxation • In (IP), by introducing Lagrangean multiplier vector u1, … , u16 we dualize Constraints (2), (5), (10), (11), (12), (16), (17), (18), (22) , (23), (24), (25), (26), (27), (28), and (29) to obtain the following Lagrangean relaxation problem(LR). We can decompose (LR) into 11 independent subproblems

  17. Getting Primal Feasible Solution S6 S2 Sink S1 S4 S5 S3

  18. S1 S1 S3 S2 S4 N5 N6 S2 Sink Assign a Slot Assign a Slot S2 S4 S1 S3 S1 S4 S2 Push Push Topology Sort S4 S4 S4 S4 S1 POP POP POP POP POP POP S3 S3 S4 S2 S2 S2 Sort byOut-degree S2 S2 S4 S1 Swap S3 Getting Primal Feasible Solution N5 N6

  19. Sink S Rerouting Heuristic

  20. Experiment Environment

  21. Experiment Scenarios • Random Network with Different Number of Sensor nodes • Depolyment Manner : Ransom Source, Congregated Source. • Grid Network with Different Number of Sensor nodes • Random Network with Different Number of Source nodes

  22. Random Network with Different Number of Sensor Nodes (Random Source)

  23. Random Network with Different Number of Sensor Nodes (congregated source)

  24. End-to-end Delay of DifferentSensor Nodes

  25. Grid Network with Different Number of Sensor Nodes

  26. Energy Cost of Different Source Number No. of Sensor Nodes = 100

  27. End-to-end Delay of Different Source Number

  28. Energy Cost : 1 + 1 + 1 = 3 Energy Cost : 2 + 2 = 4 B C Delay : 1 + 1 + 1 = 3 Delay : 1 + 1 = 2 A Experiment Discussion • Number of Source Nodes • Sensor Deployment manner • The Sequence of Paths Selection

  29. Conclusion • Contribution • We propose a mathematical formulation to model this problem as an Integer Programming Problem. • By Lagrangean Relaxation and our two-phases heuristic, we can near-optimally solve this problem. • By the reroute heuristic, we can verifies whether the algorithm we proposed achieves energy efficiency, data aggregation and ensures the latency within a reasonable range.

  30. These nodes will use up their energy soon. Conclusion • Future Work • More experiments for the different sequenceof paths selection • Take load-balance or the other factors into consideration

  31. The End Q & A - Thanks for your listening

  32. The End Appendix

  33. Notation – Given Parameters

  34. Notation – Given Parameters(cont’d)

  35. wE wE Relation between wu, nv, and mv Scheduling Problem Back

  36. Lagrangean Relaxation • In (IP), by introducing Lagrangean multiplier vector u1, … , u16 we dualize Constraints (2), (5), (10), (11), (12), (16), (17), (18), (22) , (23), (24), (25), (26), (27), (28), and (29) to obtain the following Lagrangean relaxation problem(LR). ZLR (u1, … , u16 ) = min

  37. Relaxation(cont’d) We can decompose (LR) into 11 independent subproblems.

  38. Relaxation(cont’d) Subject to:

  39. Relaxation(cont’d)

  40. Relaxation(cont’d) We can decompose (LR) into 11 independent subproblems.

  41. independent shortest path problems (SUB 1) can be further decomposed into For each shortest path problem, it can be easily solve by Dijkstra’s algorithm. Computational Complexity Subproblem 1 (related to decision variable ) min Subject to:

  42. For all outgoing links of node u, find the smallest coefficient. For all data source node s, check whether there is a outgoing link from s to the other node. For the sink node q, check whether there is at least one incoming link. For each link (u,v) compute the coefficient Computational Complexity Subproblem 2 (related to decision variable ) min Subject to:

  43. Subproblem 3 (related to decision variable ) min Subject to:

  44. calculate Computational Complexity Transformation After transforming, we can decompose (SUB 3) into | V | independent subproblems. For each node u, s.t. min

  45. calculate Computational Complexity Subproblem 4 (related to decision variable ) min Subject to: After transforming, we can decompose (SUB 4) into | V | independent subproblems. For each node u, s.t. min

  46. calculate Computational Complexity Subproblem 5 (related to decision variable ) min subject to: After transforming, we can decompose (SUB 4) into | V | independent subproblems. For each node u, s.t. min

  47. calculate Computational Complexity Subproblem 6 (related to decision variable ) min subject to: We can be further decomposedthis subproblem into |V| independent subproblems. For each node u s.t. min

  48. calculate Computational Complexity Subproblem 7 (related to decision variable ) min subject to: We can decompose this subproblem into | V | independent subproblems. For each link ( u,v ), s.t. min

  49. For all incomg links of node u, find the smallest coefficient. For each link (u,v) compute the coefficient Computational Complexity Subproblem 8 (related to decision variable ) min subject to: We can decompose this subproblem into | V x V | independent subproblems. For each link ( u,v ), s.t. min

  50. calculate Computational Complexity Subproblem 9 (related to decision variable ) min subject to: We can decompose this subproblem into | V x V | independent subproblems. For each link ( u,v ), s.t. min

More Related