1 / 29

Dresden, 12.Dec.2006

Dresden, 12.Dec.2006. Fakultät Informatik – Institut für Systemarchitektur – Professur Rechnernetze. Directed Diffusion for Wireless Sensor Networks Presented by Feng Long Sylvain Arnaud Supervisor: Dargie. Outline. Motivation Introduction Directed diffusion Interest propagation

dinos
Download Presentation

Dresden, 12.Dec.2006

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. Dresden, 12.Dec.2006 Fakultät Informatik – Institut für Systemarchitektur – Professur Rechnernetze Directed Diffusion for Wireless Sensor NetworksPresented by Feng LongSylvain ArnaudSupervisor: Dargie

  2. Directed Diffusion for Wireless Sensor Networks Outline Motivation Introduction Directed diffusion Interest propagation Gradients establishment Data propagation Reinforcement for path establishment and truncation Simulation Testbed and Improvements Conclusion

  3. Directed Diffusion for Wireless Sensor Networks Motivation Advances in processor, memory, and radio technology will enable small and cheap nodes capable of sensing, communication, and computation. Networks of such nodes can coordinate to perform distributed sensing of environmental phenomena. Now, we explore the directed-diffusion paradigm for such coordination.

  4. Directed Diffusion for Wireless Sensor Networks Introduction Directed diffusion offers robustness, energy-efficiency requirements. Interest and data propagation are determined by localized interactions (message exchanged between neighbors or nodes with some specified region) Conceptions An interest message is a query which specified what a user wants. Each interest contains a description of a sensing task that is supported by a sensor network for acquiring data. An event is a short description the sensed phenomenon. Gradient is direction state created in each node that receives an interest. The gradient direction is set toward the neighboring node form which the interest is received. An interest usually injected into the network at some node, called sink, in the network.

  5. Directed Diffusion for Wireless Sensor Networks Directed diffusion Naming Select a naming scheme is the first step in designing directed diffusion for the Network. Task descriptions are named by, for example, a list of attribute-value pairs that describe a task. Type = four-legged animal // detect animal location Interval = 20 ms // send back events every 20 ms (specify an event data rate) Duration = 10 seconds // .. For the next 10 seconds the task state is purged from the node Rect = [-100, 100, 200, 400] // from sensors within rectangle, based on GPS coordinates. The task description specified an interest for data matching the attributes.

  6. Directed Diffusion for Wireless Sensor Networks Interest propagation How interests are diffused through the sensor network? Sink periodically broadcasts an interest message to each of its neighbors. This initial interest contains a much larger interval attribute. The initial interest is called exploratory which tries to determine if there indeed are any sensor nodes that detect the specific type. The interest is soft state that will be periodically refreshed by the sink. The sink simply resends the same interest with a monotonically increasing timestamp attribute.

  7. Directed Diffusion for Wireless Sensor Networks Interest propagation Every node maintains an interest cache. Each item in the cache correspond to a distinct interest, in the example before, if their type attribute differ, or their rect attributes are disjoint. An entry in the interest cache has several fields: - A timestamp field It indicates the timestamp of the last received matching interest. - Several gradient fields It contains a data rate field requested by the specified neighbour.

  8. Interest propagation event source interests sink • When a node receives an interest, it checks to see if the interest exists in the cache. • No matching entry exists (type & rect), the node creates an interest entry. This entry has a single gradient toward the neighbor form which the interest was received, with the specified event data rate. • -If there exists an interest entry, but no gradient of the sender of the interest, then adds a gradient with the specified data rate. Fig.1.Simple schematic for directed diffusion Interest propagation Source [1] Directed Diffusion for Wireless Sensor Networks

  9. Directed Diffusion for Wireless Sensor Networks Interest propagation After receiving an interest, a node may decide to resent the interest to some subset of its neighbors. Sensor network queries may often be geographical, so use an energy efficient routing algorithm, Geographic and Energy Aware Routing (GEAR) algorithm to propagates a query to the appropriate geographical region. This can limit the topological scope for interest diffusion, there by resulting in energy savings. Or A node might use cached date to direct interests, rather than broadcasting to all neighbors. Note Not all received interests are resent. A node may suppress a received interest (timestamp) if it recently resent a matching interest.

  10. Gradients Establishment event source Gradients sink • Every pair of neighboring nodes establishes a gradient toward each other. Such two –way gradients can enable fast recovery form failed paths or reinforcement of empirically better paths. Fig.1.Simple schematic for directed diffusion (b) Initial gradients setup Source [1] Directed Diffusion for Wireless Sensor Networks

  11. Directed Diffusion for Wireless Sensor Networks Data propagation A sensor node that is within the specified rect processes interests, in Addition, the node tasks its local sensors to begin collection samples. A node that receives a data message form its neighbors attempts to find a matching interest entry in its interest cache. A matching interest entry is one whose Rect encompasses the sensor location The type of the entry matches the detected target type. If find one It computes the highest requested event data rate among all its outgoing gradients. The node tasks its sensor subsystem to generate event samples at this highest data rate.

  12. Directed Diffusion for Wireless Sensor Networks Data propagation A matching interest entry exists The node checks the data cacheassociated with the matching interest entry. Data cache keeps track of recently seen data items A received data message has a matching data cache entry The data message is dropped. If not The received message is added to the data cache and the data message is resent to the node’s neighbors. To resend a received data message, a node needs to examine the matching interest entry’s gradient list. If all gradients have a data rate > = the rate of incoming events The node may simply send the received data message to the appropriate neighbors. else The node may downconvert to the appropriate gradient which has the lower.

  13. Directed Diffusion for Wireless Sensor Networks Reinforcement for Path Establishment and Truncation When a source detects a matching target, it sends exploratory events, possible along multiple paths, toward the sink. After the sink starts receiving these exploratory events, it reinforces one particular neighbor in order to “draw down” real data. Path Establishment Using Positive Reinforcement - How to use positive reinforcement? Reinforce that neighbor only if it is sending exploratory events. To positive reinforce this neighbor, the sink resends the original interest message but The sender’s interest specifies a higher data rate than before. - How to choose a good path, an empirically low-delay path? Choose neighbour form which the most events have been received which consistently sends events before other neighbors

  14. 2) Local Repair for Failed Paths event source Gradients sink • Intermediate nodes on a previously reinforced path can apply the reinforcement rules. • How node C detects failed or degraded paths? • Noticing that the event reporting • rate from its upstream neighbor is now lower. • Realizing that other neighbors • have been transmitting previously unseen location. • Eventually, intermediate node C positively reinforces the direct link to the source. C Fig.2. Repair for failed paths Source [1] Directed Diffusion for Wireless Sensor Networks

  15. 3) Path Truncation Using Negative Reinforcement event B source sink • When to use? • If the path through B is consistently better (i.e. B sends events before A does), we need a mechanism to negatively reinforce the path through A. • What mechanism? • Negatively reinforcing that neighbour from which fewer new events have been received. • How to use? • The negative reinforcement is the interest with the lower data rate. • When node receives this interest , it degrades its gradient toward the sink. A Fig. 3. Negative reinforcement for path truncation Multiple path Source [1] Directed Diffusion for Wireless Sensor Networks

  16. 4) Loop Removal Using Negative Reinforcement E D source Sink A B C • Suppressing high-delay, negative reinforcement is also used for loop removal because the looping paths never deliver events first. • Loop {B  E D} eliminated Fig. 3. Negative reinforcement for path truncation (b) Loop Removal Source [1] Directed Diffusion for Wireless Sensor Networks

  17. Directed Diffusion for Wireless Sensor Networks Evaluation Implications of design choices Comparaison with ideal schemes: Flooding Omniscient multicast (~IP-based network) Impact of dynamics (node failure)

  18. Directed Diffusion for Wireless Sensor Networks Simulation Sensors randomly placed in a square Five sources, five sinks Metrics: Average dissipated energy Average delay Regime far from overload Study impact of dynamics

  19. Comparative evaluation • Directed Diffusion better than Omniscient multicast: • In-network aggregation Flooding Omniscient multicast Directed diffusion Fig. 4. Comparative evaluation (a) Average dissipated energy Directed Diffusion for Wireless Sensor Networks Source: [2]

  20. Comparative evaluation Shortest path ~ lowest delay path Fig. 4. Comparative evaluation (b) Delay Directed Diffusion for Wireless Sensor Networks Source: [2]

  21. Impact of dynamics • A fixed fraction of nodes is turned off for 30s, some on the shortest path Fig. 5. Impact of dynamics (b) Delay Fig. 5. Impact of dynamics (a) Average dissipated energy • Reinforcement rules are conservative Directed Diffusion for Wireless Sensor Networks Source: [2]

  22. Directed Diffusion for Wireless Sensor Networks Simulation – discussion Potential for significant energy efficiency With sub-optimal path selection Stable under the network dynamics Need for more practical evaluation

  23. Testbed • 14 PC/104 nodes, 13kb/s modems • 1 sink, multiple sources, ~ 4 hops apart Fig. 6. Testbed Source: [3] Directed Diffusion for Wireless Sensor Networks

  24. Directed Diffusion for Wireless Sensor Networks Studied features Filtering: Opportunistic aggregation: Sensor selection and tasking through app-level attributes Data cached as it traverses App filters act on data Fig. 7. Nested queries • Nested queries: • Optimize correlation triggers • Example: light change triggers audio sensor Source: [3]

  25. Benefits of in-network processing Fig. 8. Bytes sent by diffusion Source: [3] Directed Diffusion for Wireless Sensor Networks

  26. Benefits of nested queries Fig. 9. Audio events delivered Source: [3] Directed Diffusion for Wireless Sensor Networks

  27. Directed Diffusion for Wireless Sensor Networks Conclusion Main features: Attribute-named data Avoids multiple level name binding Enables in-network processing with filters Reinforcement Robustness Directed diffusion can significantly decrease energy consumption

  28. Directed Diffusion for Wireless Sensor Networks References [1] C. Intanagonwiwat, R. Govinda, D. Estrin, J. Heidemann and F. Silva, “Directed Diffusion for Wireless Sensor Networking”, in Networking, IEEE/ACM Transactions. vol.11, issue 1, pp. 2-16, February 2003. [2] C. Intanagonwiwat, R. Govinda and D. Estrin, “Directed diffusion: a scalable and robust communication paradigm for sensor networks”, in Proceedings of the Sixth Annual International Conference on Mobile Computing and Networks (MobiCOM 2000), August 2000, Boston, Massachusetts. [3] J. Heidemann, F. Silva, C. Intanagonwiwat, R. Govindan, D. Estrin and D. Ganesan, “Building Efficient Wireless Sensor Networks with Low-Level Naming”, in Proceedings of the Symposium on Operating Systems Principles, Lake Louise, Banff, Canada, ACM. October 2001. [4] D. Estrin, R. Govindan, J. Heidemann and S. Kumar, “Next Century Challenges: Scalable Coordination in Sensor Networks”, in Proceedings of the Fifth Annual International Conference on Mobile Computing and Networks (MobiCOM '99), Seattle, Washington, August 1999. [5] Y.Yu, R. Govindan and D. Estrin, “Geographical and energy aware routing: A recursive data dissemination protocol for wireless sensor net-works”, UCLA Computer Science Department Technical Report UCLA/CSD-TR-01-0023, May 2001.

  29. Directed Diffusion for Wireless Sensor Networks Thanks for your attention...

More Related