1. Sensor Networks: intro, overview, example Jim Kurose* Vic Lesser CMPSCI 791L Sensor Nets Seminar Fall 2003 Some slides used/adapted (with thanks) from D. Estrin, with permission

2. Today’s class: overview  sensor nets: motivation  system design themes  themes  time and space: synchronization, location, coverage  in-network computation  “data is king”  illustrative sensor net application, system structure

3. Embedded Networked Sensing: Motivation Imagine:  high-rise buildings self-detect structural faults (e.g., weld cracks)  schools detect airborn toxins at low concentrations, trace contaminant transport to source  buoys alert swimmers to dangerous bacterial levels  earthquake-rubbled building infiltrated with robots and sensors: locate survivors, evaluate structural damage  ecosystems infused with chemical, physical, acoustic, image sensors to track global change parameters  battlefield sprinkled with sensors that identify track friendly/foe air, ground vehicles, personnel

4. Embedded Networked Sensing Apps  Micro-sensors, on-board processing, wireless interfaces feasible at very small scale--can monitor phenomena “up close”  Enables spatially and temporally dense environmental monitoring Embedded Networked Sensing will reveal previously unobservable phenomena Seismic Structure response Contaminant Transport Marine Microorganisms Ecosystems, Biocomplexity

5. Imagine (the CASA version)…. Noontime: all clear  DCAS systems monitor 3D winds, 0 to 3 km high  “clear-air” winds provide basis for pollutant monitoring, migratory bird tracking Dense network of radars - distributed collaborative adaptive sensing (DCAS)

6. Imagine…. 2PM: solar ground heating  wind convergence zones form  DCAS pattern detection algorithms detect convergence  data archiving begins  radars automatically tasked to sample moisture fields around convergence zone  models generate predictions, provided to local emergency managers for planning

7. Imagine…. 3PM: severe weather  Clouds, precipitation develop in convergence several zones  DCAS radars adjust, provide fine-scale measurements, precipitation estimates in critical areas  skies to south clear, but DCAS systems monitoring 3D temperature, moisture to assess potential for future thunderstorms  rotational signatures cause nearby radars to enter tornado tracking mode  location, intensity, projected path provided to community, state organizations, industry. Because of 2PM predictions, officials prepared  spawned tornado destroys two radars, nearby DCAS radars reconfigure

8. Imagine…. 5PM: storms move south to Houston .. as predicted by continuously monitoring DCAS systems  rainfall begins, DCAS systems reconfigure to map precipitation at fine resolution  DCAS measurements feed hydrological models  local, state, organizational emergency response teams are in action and prepared well in advance of flood waters..

9. Embedded Sensor Nets: Enabling Technologies EmbeddedNetworked Sensing Control system w/ Small form factor Untethered nodes Exploit collaborative Sensing, action Tightly coupled to physical world Embed numerous distributed devices to monitor and interact with physical world Network devices to coordinate and perform higher-level tasks Exploit spatially/temporally dense, in situ/remote, sensing/actuation

10. Sensor Nets: New Design Themes  self configuring systems that adapt to unpredictable environment  dynamic, messy (hard to model), environments preclude pre- configured behavior  leverage data processing inside the network  exploit computation near data to reduce communication  collaborative signal processing  achieve desired global behavior with localized algorithms (distributed control)  long-lived, unattended, untethered, low duty cycle systems  energy a central concern  communication primary consumer of scarce energy resource

11. From Embedded Sensing to Embedded Control  embedded in unattended “control systems”  control network, and act in environment  critical app’s extend beyond sensing to control and actuation  transportation, precision agriculture, medical monitoring and drug delivery, battlefield app’s  concerns extend beyond traditional networked systems and app’s: usability, reliability, safety  need systems architecture to manage interactions  current system development: one-off, incrementally tuned, stove-piped  repercussions for piecemeal uncoordinated design: insufficient longevity, interoperability, safety, robustness, scaling

12. Why cant we simply adapt Internet protocols, “end to end” architecture?  Internet routes data using IP Addresses in Packets and Lookup tables in routers  humans get data by “naming data” to a search engine  many levels of indirection between name and IP address  embedded, energy-constrained (un-tethered, small-form- factor), unattended systems cant tolerate communication overhead of indirection  special purpose system function(s): don’t need want Internet general purpose functionality designed for elastic applications. 

13. Is there an broader architecture Duck Island ME: habitat sensing Oklahoma: atmospheric sensing Can we define layered (Internet-like) architecture appropriate for wide variety of networked sensing systems? : stovepipes or layers?

14. Sample Layered Architecture Resource constraints call for more tightly integrated layers Open Question: What are defining Architectural Principles? In-network: Application processing, Data aggregation, Query processing Adaptive topology, Geo-Routing MAC, Time, Location Phy: comm, sensing, actuation, SP User Queries, External Database Data dissemination, storage, caching

15. Today’s class: overview  sensor nets: motivation  system design themes  themes  time and space: synchronization, location, coverage  in-network computation  “data is king”  illustrative sensor net application, system structure

16. Sensors  passive elements: seismic, acoustic, infrared, strain, salinity, humidity, temperature, etc.  passive Arrays: imagers (visible, IR), biochemical  active sensors: radar, sonar  High energy, in contrast to passive elements  technology trend: use of IC technology for increased robustness, lower cost, smaller size  COTS adequate in many of these domains; work remains to be done in biochemical

17. Fine Grained Time and Location  unlike Internet, node time/space location essential for local/collaborative detection  fine-grained localization and time synchronization needed to detect events in three space and compare detections across nodes  GPS provides solution where available (with differential GPS providing finer granularity)  GPS not always available, too “costly,” too bulky  other approaches under study  localization of sensor nodes has many uses  beamforming for localization of targets and events  geographical forwarding  geographical addressing

18. Coverage measures  area coverage: fraction of area covered by sensors  detectability: probability sensors detect moving objects  node coverage: fraction of sensors covered by other sensors  control:  where to add new nodes for max coverage  how to move existing nodes for max coverage S D x Given: sensor field (either known sensor locations, or spatial density)

19. In Network Processing  communication expensive when limited  power  bandwidth  perform (data) processing in network  close to (at) data  forward fused/synthesized results  e.g., find max. of data  distributed data, distributed computation

20. Distributed Representation and Storage  Data Centric Protocols, In-network Processing goal:  Interpretation of spatially distributed data (Per- node processing alone is not enough)  network does in-network processing based on distribution of data  Queries automatically directed towards nodes that maintain relevant/matching data  pattern-triggered data collection  Multi-resolution data storage and retrieval  Distributed edge/feature detection  Index data for easy temporal and spatial searching  Finding global statistics (e.g., distribution) K V Time

21. Directed Diffusion: Data Centric Routing  Basic idea  name data (not nodes) with externally relevant attributes: data type, time, location of node, SNR,  diffuse requests and responses across network using application driven routing (e.g., geo sensitive or not)  support in-network aggregation and processing  data sources publish data, data clients subscribe to data  however, all nodes may play both roles node that aggregates/combines/processes incoming sensor node data becomes a source of new data node that only publishes when combination of conditions arise, is client for triggering event data  true peer to peer system?

22. Traditional Approach: Warehousing  data extracted from sensors, stored on server  Query processing takes place on server Warehouse Front-end Sensor Nodes

23. Sensor Database System  Sensor Database System supports distributed query processing over sensor network Sensor DB Front-end Sensor Nodes

24. Sensor Database System  Characteristics of a Sensor Network:  Streams of data  Uncertain data  Large number of nodes  Multi-hop network  No global knowledge about the network  Node failure and interference is common  Energy is the scarce resource  Limited memory  No administration, … Can existing database techniques be reused? What are the new problems and solutions?  Representing sensor data  Representing sensor queries  Processing query fragments on sensor nodes  Distributing query fragments  Adapting to changing network conditions  Dealing with site and communication failures  Deploying and Managing a sensor database system

25. Performance Metrics  High accuracy  Distance between ideal answer and actual answer?  Ratio of sensors participating in answer?  Low latency  Time between data is generated on sensors and answer is returned  Limited resource usage  Energy consumption

26. Today’s class: overview  sensor nets: motivation  system design themes  themes  time and space: synchronization, location, coverage  in-network computation  “data is king”  illustrative sensor net application, system structure