Wireless Sensor Networks Real- Time Systems

Transcription

1 Wireless Sensor Networks Real- Time Systems Chenyang Lu Computer Science and Engineering h5p://

2 Wireless Sensor Network Processor + Sensors/Actuators + Wireless Interface Miniature hardware manufactured economically in large numbers Smart Dust (UCB) 2

3 WUSTL Sensor Network Testbed 3

4 WUSTL Sensor Network Testbed 4

5 Promise of Sensor Networks Wireless Sensor Networks will change the world. 5

6 Promise of Sensor Networks Wireless Sensor Networks will change the world. Well, maybe not the whole world. 6

7 Promise of Sensor Networks Wireless Sensor Networks will change the world. Well, maybe not the whole world. But how about Smart Energy Healthcare and wellness (e.g., body networks) Process monitoring and control (e.g., oil rigs and refineries) Smart civil infrastructure PolluUon monitoring 7

8 Research Areas Wireless Sensor Networks ApplicaUon- driven Systems Research Clinical Monitoring Fall detecuon Structural Health Monitoring and Control PolluUon Monitoring Low- power Wireless Networking Home- Area Networks MAC Layer Architecture Real- Ume Wireless Control Networks Real- Time Systems RT- Xen: VirtualizaUon for Real- Time Systems CyberMech: Parallel MulUcore Pla_orm for Hybrid TesUng 8

9 Research Areas Wireless Sensor Networks ApplicaUon- driven Systems Research Clinical Monitoring Fall detecuon Structural Health Monitoring and Control PolluUon Monitoring Low- power Wireless Networking Home- Area Networks MAC Layer Architecture Real- Ume Wireless Control Networks Real- Time Systems RT- Xen: VirtualizaUon for Real- Time Systems CyberMech: Parallel MulUcore Pla_orm for Hybrid TesUng 9

10 Two- Tiered Clinical Early Warning Tier 1: idenufy at- risk pauents from exisung medical record data + Tier 2: predict pauent s outcome from exisung medical records and real- Ume vital sign monitoring 10

11 Clinical Monitoring First deployment of clinical monitoring system using wireless sensor networks. 7- month clinical trial with 46 pauents. Much higher frequency than current pracuce 1 reading/min vs. several readings/day. Highly reliable network Developing early detecuon system of clinical deteriorauon Chipara, Lu, Bailey, Roman, Reliable Clinical Monitoring using Wireless Sensor Networks: Experience in a Step- down Hospital Unit, SenSys 10. Chipara, Brooks, Bha5acharya, Lu, Chamberlain, Roman and Bailey, Reliable Real- Ume Clinical Monitoring Using Sensor Network Technology, AMIA

12 System Architecture Base stauon Laptop connected to Wi- Fi Relays Plugged into wall outlets Redundant deployment coverage fault tolerance Portable pulse oximeter Pulse ox + microcontroller + radio Ba5ery operated 12

13 Reliable Network Architecture Problem: PaUents in general hospital units are ambulatory. SoluUon: Two- Uer architecture for end- to- end data delivery. 1 Dynamic Relay AssociaUon Protocol (DRAP): PaUent - > 1 st relay Dynamically associate the pauent node with a relay Single- hop protocol handles pauent mobility Simplify power management in pauent nodes (send only) 2 StaUonary relay network: 1 st relay - > - > base stauon Reuse exisung mesh rouung protocol: CollecUon Tree Protocol (TinyOS 2.x) Isolated from pauent mobility Wall- plugged => no need to worry about energy 13

14 Clinical Deployment Step- down cardiac care unit 16 pauent rooms, 1200 m 2 Network 18 relays: redundant network Longest path: 3-4 hops Channel 26 of IEEE Pulse and oxygenauon collected every 30 or 60 seconds 46 pauents enrolled >41 days of monitoring 2-68 hours per pauent 5 pauents excluded from analysis Up to 3 pauents at a Ume 14

15 PotenEal for DetecEng Clinical DeterioraEon #$%& #$%&!"!" Pulmonary edema Sleep apnea #$%&!" Bradycardia 15

16 System Reliability Network reliability >95% for all pauents. EffecUveness of DRAP+CTP Median sensing reliability > 80%. But 29% of pauents with sensing reliability < 50% System reliability dominated by sensing reliability. 16

17 Research Areas Wireless Sensor Networks ApplicaUon- driven Systems Research Clinical Monitoring Fall deteceon Structural Health Monitoring and Control PolluUon Monitoring Low- power Wireless Networking Home- Area Networks MAC Layer Architecture Real- Ume Wireless Control Networks Real- Time Systems RT- Xen: VirtualizaUon for Real- Time Systems CyberMech: Parallel MulUcore Pla_orm for Hybrid TesUng 17

18 Fall DetecEon DetecUng falls of elderly. ConUnuous monitoring Beyond homes. ConnecUng body sensors with smart phones Working with real elderly users User acceptance is criucal Falls during everyday acuviues of elderly Rigorous clinical studies CollaboraUon with medical researchers at Alzheimer s Center 18

19 Research Areas Wireless Sensor Networks ApplicaUon- driven Systems Research Clinical Monitoring Fall detecuon Structural Health Monitoring and Control PolluUon Monitoring Low- power Wireless Networking Home- Area Networks MAC Layer Architecture Real- Ume Wireless Control Networks Real- Time Systems RT- Xen: VirtualizaUon for Real- Time Systems CyberMech: Parallel MulUcore Pla_orm for Hybrid TesUng 19

20 Wireless Smart Civil Structures Develop smart structures (with monitoring and control) to prevent Minneapolis Bridge Collapse Freeway aner 1989 San Francisco Earthquake 20

21 Structural Health Monitoring >26% of the nauon's bridges are either structurally deficient or funcuonally obsolete. [ASCE 2009] Wireless monitoring and control systems for civil infrastructure Damage localizauon on real structures Advantages over centralized approach reduce latency by 88% x3.4 increase in ba5ery life under an hourly schedule G. Hackmann, W. Guo, G. Yan, C. Lu, S. Dyke, Cyber- Physical Codesign of Distributed Structural Health Monitoring with Wireless Sensor Networks, ICCPS'10. G. Hackmann, F. Sun, N. Castaneda, C. Lu and S. Dyke, A HolisUc Approach to Decentralized Structural Damage LocalizaUon Using Wireless Sensor Networks, RTSS

22 Research Areas Wireless Sensor Networks ApplicaUon- driven Systems Research Clinical Monitoring Fall detecuon Structural Health Monitoring and Control PolluUon Monitoring Low- power Wireless Networking Home- Area Networks MAC Layer Architecture Real- Ume Wireless Control Networks Real- Time Systems RT- Xen: VirtualizaUon for Real- Time Systems CyberMech: Parallel MulUcore Pla_orm for Hybrid TesUng 22

23 Home Area Networks (HANs) Smart energy: Power meters, thermostats, HVAC and appliances. Assisted living: Body sensors, environmental sensors, medical devices. Source: AT&T Labs 23 23

24 Wireless CharacterisEcs in Real Homes Measurements in >10 apartments. Noisy environment severely affect reliability and power efficiency. ARCH: ReacUve channel hopping for Home Area Networks (HANs). 2.4 GHz spectrum in six apartments and an office Sha, Hackmann, Lu, MulU- Channel Reliability and Spectrum Usage in Real Homes: Empirical Studies for Home- Area Sensor Networks, IWQoS'11. Sha, Hackmann, Lu, ARCH: PracUcal Channel Hopping for Reliable Home- Area Sensor Networks, RTAS

25 Research Areas Wireless Sensor Networks ApplicaUon- driven Systems Research Clinical Monitoring Fall detecuon Structural Health Monitoring and Control PolluUon Monitoring Low- power Wireless Networking Home- Area Networks MAC Layer Architecture Real- Ume Wireless Control Networks Real- Time Systems RT- Xen: VirtualizaUon for Real- Time Systems CyberMech: Parallel MulUcore Pla_orm for Hybrid TesUng 25

26 Problem: Monolithic Radio Stack Hard to develop new MAC protocols No clear separauon of concerns Need inumate knowledge of enure stack Hard to maintain muluple MAC stacks as OS evolves Hard to port protocols across radio/processor pla_orms

27 MLA: MAC Layer Architecture SeparaUon of power management from radio core Components for sleep scheduling protocols Reusable ease development and maintenance of MAC protocols Pla_orm independent reduce porung effort MLA: encapsulates MAC power management techniques into reusable components in a unified and flexible architecture Power Management Timers Radio Core Klues, Hackmann, Chipara, Lu, A Component Based Architecture for Power- Efficient Media Access Control in Wireless Sensor Networks, SenSys 07. Klues, Xing, Lu, Link Layer Driver Architecture for Unified Radio Power Management in Wireless Sensor Networks, ACM TransacUons on Embedded CompuUng Systems, 9(4),

28 Research Areas Wireless Sensor Networks ApplicaUon- driven Systems Research Clinical Monitoring Fall detecuon Structural Health Monitoring and Control PolluUon Monitoring Low- power Wireless Networking Home- Area Networks MAC Layer Architecture Real- Eme Wireless Control Networks Real- Time Systems RT- Xen: VirtualizaUon for Real- Time Systems CyberMech: Parallel MulUcore Pla_orm for Hybrid TesUng 28

29 Real- Time Wireless Control Networks WirelessHART Standard for wireless process monitoring and control. Commercial deployment. Real- Ume scheduling for WirelessHART Mapping to real- Ume CPU scheduling Handle conflicts between wireless transmissions General theory for real- Ume scheduling for wireless Saifullah, Xu, Lu, Chen, Real- Time Scheduling for WirelessHART Networks, RTSS 10. Saifullah, Xu, Lu, Chen, End- to- End Delay Analysis for Fixed Priority Scheduling in WirelessHART Networks, RTAS 11. Saifullah, Xu, Lu, Chen, Priority Assignment for Real- Time Flows in WirelessHART Networks, ECRTS'11. 29

30 Research Areas Wireless Sensor Networks ApplicaUon- driven Systems Research Clinical Monitoring Fall detecuon Structural Health Monitoring and Control PolluUon Monitoring Low- power Wireless Networking Home- Area Networks MAC Layer Architecture Real- Ume Wireless Control Networks Real- Time Systems RT- Xen: VirtualizaEon for Real- Time Systems CyberMech: Parallel MulUcore Pla_orm for Hybrid TesUng 30

31 RT- Xen VirtualizaUon enables flexible resource allocauon among systems. Cloud compuung: Xen, Vmware Real- Ume virtualizauon - > integrauon of real- Ume systems Preserve real- Ume performance guarantees across system hierarchy. RT- Xen: virtual machine manager for real- Ume systems Hierarchical scheduling [EMSOFT 11] Real- Ume I/O scheduling MulUcore Distributed systems Open- source: h5ps://sites.google.com/site/realumexen/ S. Xi, J. Wilson, C. Lu and C.D. Gill, RT- Xen: Towards Real- Ume Hypervisor Scheduling in Xen, ACM InternaUonal Conference on Embedded Sonware (EMSOFT), October

32 CyberMech High- fidelity validauon is criucal for system acceptance. Empirical tesung of full- scale structures is infeasible. Time and cost for comprehensive test Scale of civil structures CyberMech: run- Ume substrate for real- Ume hybrid tesung ComputaUonal + Physical Models Real- Ume system for parallel compuung over mulucore pla_orms A. Saifullah, K. Agrawal, C. Lu and C.D. Gill, MulU- core Real- Time Scheduling for Generalized Parallel Task Models, IEEE Real- Time Systems Symposium (RTSS'11), December

33 Take Home Message World- class research program on sensor networks and real- Ume systems From theory to real systems deployed in the field ImpacUng both the academia and real world Right area at the right Ume! Academic placement O. Chipara (2010) Assistant Professor, University of Iowa. G. Xing (2006): Assistant Professor, Michigan State University. X. Wang (2006): Associate Professor, Ohio State University. Talk to me and my team about opportuniues Rahav Dor, Yong Fu, Greg Hackmann, Bo Li, Abu Saifullah, Mo Sha, Chengjie Wu, Sisu Xi. h5p:// 33