Outline � • � Medical applications using sensor networks � • � Design requirements for wireless clinical monitoring � Robust Wireless Clinical Monitoring for • � Existing approach to wireless clinical monitoring � Ambulatory Patients � • � Initial prototype � • � 4 - bit link estimator � Octav Chipara � • � CTP � • � Impact of mobility � • � Initial prototype revisited � • � Multi-user study � Medical applications using WSNs � Assisted living - crowded space � behavior � memory � • � AlarmNet, UVA � • � Lace, Rochester � • � Lace,Rochester � • � Smart Homes � • � Intel � • � Honeywell � • � Motorola � • � Philips � fall prevention � vitals � • � AlarmNet, UVA � more projects at http://www.agingtech.org/techdemo.aspx � Smart homes � Disaster recovery problem � • � Best practice: � • � use color-coded tags to indicate severity � • � constant reassessment of patients (5-15 mins) � • � verbal coordination between different organization (e.g., first responders, hospitals, etc.) � • � Issues with current practice � • � limitations of colored tags � • � the tag may not accurately reflect a patients condition � • � unclear how to prioritize patients with same color � • � no room to leave notes � • � hard to coordinate activities between rescue workers and hospitals, ambulances, etc. �
Wireless clinical monitoring � Requirements � • � Early detection is fundamental to reducing mortality and length of • � Designed for non-ICU settings � stay in hospitals (e.g., in respiratory/cardiac arrests, septic shock) � • � inexpensive � • � Changes in vital signs may indicate clinical deterioration � • � reliable � • � automatic early detection systems are in the works � • � works for ambulatory patients � • � The accuracy and sensitivity of these systems depends on having • � lifetime of 3-5 days (avg. hospitalization duration) � up-to-date vital sign information � • � continuos monitoring once per minute � • � low maintenance cost � • � Integrate with electronic patient records to enable automatic early detection records � Existing solutions � • � GE, Phillips,CISCO provide wireless medical telemetry systems and what wireless have limited success in cardiology departments; However, � • � closed systems � technology should you • � high cost of infrastructure, relies on LAN � • � ad hoc deployment of nodes � pick? � • � high maintenance cost � Wireless devices � Selecting a wireless technology � • � Criteria: � • � Low-power operation (3 - 5 day requirement, small packets) � Passive Bluetooth � Bluetooth � Device � WLAN � 802.15.4 � RFID � (class 1) � (class 3) � • � Small bandwidth requirements � • � Minimize electro-magnetic interference � • � Degree of miniaturization (weight) � Frequency � 125 KHz � 2.4-2.5 GHz � 2.4 GHz � 2.4 GHz � 2.4 GHz � • � Inexpensive radios � • � Our choice: TelosB with 802.15.4 � Power � 2-4W � 100mW � 1mW � 100mW � 1mW �
A wireless pulse-oximeter � hardware � System components System components � • � Base-station node � • � has access to wired network � • � powered � • � integrates with electronic patient records � system architecture � • � Relay nodes � • � deploy to ensure coverage � Relay node � • � impractical to change batteries => powered � • � no access to wired network => reduce cost � • � Patient nodes � • � integrates sensing modalities with TelosB � • � battery operated � Wireless � pulse-oximeter � 9 � System architecture System architecture � initial software prototype � 10 �
Software prototype � • � Patient node: � • � OxylinkSensorC - device driver for pulse-oximeter � • � PacketLoggingC - logs received/transmitted packets � • � CollectionC - implements CTP � 4-bit link estimator � • � PatientAppC - implements the actual patient application � • � Base-station & relay nodes: � • � CollectionC � • � PacketLoggingC � Link Quality Estimation � ETX Estimation Example � • � Identify good links � A B • � ETX: Expected Transmission Count � TX [Mobicom 2003] � Beacons ReTX 1.0 3.0 1.0 ETX Estimate 1 2.0 1.8 2.04 1.83 (alpha = 0.8) t1 t2 t3 ETX(L) = ACK PRR(f) * PRR(b) 22 � State of the art � Scope � • � Not all information used � • � Identify the information • � Coupled designs � Network Layer different layers of the stack • � Physical layer (LQI) � can provide � • � Coupled implementation � LE • � Define a narrow interface between the layers and the link estimator � • � Describe an accurate and efficient estimator implemented using the four bit interface � 24 �
Layers and Information � PHY Info Not Sufficient � • � Better estimator with information from different layers? � Unacked Network Layer • � Physical Layer � LE • � Packet decoding quality � Network Layer • � Link Layer � • � Packet Acknowledgements � LE PRR • � Network Layer � • � Relative importance of links � Network Layer LE LQI 25 � 26 � Network Layer Network Layer LE LE Physical Layer � Link Layer � • � Decoding Quality � • � Outcome of unicast packet transmission � • � Agile � • � Higher quality links � • � Free � • � Successful TX � A • � Asymmetric (receive) quality � • � Successful ACK reception � • � Radio-specific � ACK DATA • � Examples � • � Example � • � LQI, RSSI, SNR � • � EAR [Mobicom 2006] � B 27 � 28 � Network Layer LE Network Layer � Interface Details � • � Is a link useful? � PIN COMPARE Keep this link in the • � Keep useful links in the table � Is this a useful link? SRC table • � Network layer decides � • � Geographic routing � A • � Geographically diverse links � ACK WHITE • � Collection � A packet transmission Packets on this • � Link to the parent � on this link was channel experience acknowledged • � Link on a good path � few errors DST 29 � 30 �
Collection Tree Routing � • � Link dynamics: uses 4-bit link estimator � • � Control plane � • � Data plane � collection tree routing � Control plane � Route selection � • � Responsible for link quality estimation and route selection � • � We strive to select the route with minimum ETX � • � ETX is used to path quality � • � switch routes only if its ETX is > 1 better than current route � • � uses both data and beacons for estimation � • � it is dangerous to route on a path for which we have a little statistical data � • � Route updates are disseminated in beacons � • � Loops are detected using feedback from the data plane � • � beacons transmitted on a variable timer � • � the beacon period is reset when an inconsistency is detected � • � the beacon period increased if route costs remain stable � Data plane � • � Responsible for forwarding packets � • � Issues: � • � self-interference - a child’s transmission may interfere with its Is CTP suited for mobile parent’s transmission � • � duplicate suppression - result of lost ACKs � users? � • � loop detection � • � ETX should monotonously decrease as a packet is forwarded to the sink � • � long routes � • � duplicates �
Impact of mobility on reliability Impact of mobility on reliability � Experimental setup � • � Initial prototype used Collection Tree Protocol (CTP) � • � de facto data collection protocol in TinyOS � • � Reliability components � Data delivery to the base station � relay reliability � end-to-end reliability � Data delivery to first relay � first-hop reliability � • � Setup: � • � 1 volunteer � • � data rate: 1 reading/min � 11 � CTP under normal activity CTP under normal activity � DRAP � DRAP end-to-end reliability: 82.39% � • � Approach - separate the routing problem in two parts � first-hop reliability: 85.38% � • � first-hop delivery: DRAP handles it � relay reliability: 96.49% � • � relay delivery: CTP handles it � • � Advantages of this approach: � • � reliable data collection from mobile users without changes to routing � • � reduces footprint and bandwidth overhead on patient nodes � • � Challenge: optimize for no mobility react quickly to mobility � • � no mobility => feedback from physical & link layer � • � Normal activity experiment showed � • � mobility => n broken links � • � most packet losses occur on the first hop � • � packet drops tend to follow user movement � • � Hypothesis: � poor routing table management => use of outdated entries � 12 � 13 � Efficient one-hop collection � DRAP’s reliability DRAP’s reliability � DRAP � CTP � end-to-end reliability: 82.39% � end-to-end reliability: 99.33% � one-hop reliability: 85.38% � one-hop reliability: 100% � relay reliability: 96.49% � relay reliability: 99.33% � 14 �
Recommend
More recommend
