An Analysis of a Large An Analysis of a Large John Anderson, - PDF document

The Problem The Problem An Analysis of a Large An Analysis of a Large � John Anderson, College of the John Anderson, College of the � Scale Habitat Monitoring Scale Habitat Monitoring Atlantic Atlantic Application Application • Studying distribution and abundance of • Studying distribution and abundance of sea birds on Great Duck Island, Maine sea birds on Great Duck Island, Maine � Wanted to: Wanted to: � Presented by: Mike Leib Presented by: Mike Leib • • Measure occupancy of burrows Measure occupancy of burrows • Understand role of micro Understand role of micro- - climactic climactic • factors in habitat selection factors in habitat selection The Problem The Problem Goals Goals 1. System capable of producing animal 1. System capable of producing animal density GIS plots density GIS plots 2. Continuous real 2. Continuous real- - time data collection time data collection 3. Minimal impact on birds and surrounding 3. Minimal impact on birds and surrounding environment environment The Solution – – A Wireless Sensor A Wireless Sensor The Solution The Solution - - Continued Continued The Solution Network Network 2. 2. Continuous real- Continuous real - time data collection time data collection 1. System capable of producing animal 1. System capable of producing animal • • Real- Real - time sensor readings time sensor readings density GIS plots density GIS plots • • Internet Connectivity Internet Connectivity Record temperature of burrows Record temperature of burrows � � Continuously updated logs available via web Continuously updated logs available via web � � • • Hot- Hot - spots are likely caused by a resident bird spots are likely caused by a resident bird 3. Minimal impact on birds and surrounding 3. Minimal impact on birds and surrounding environment environment • • Remote m onitoring Remote m onitoring Both habitat and instrumentation can be monitored Both habitat and instrumentation can be monitored � � Minimizes observer effects and alteration of Minimizes observer effects and alteration of � � environment environment • Long life of sensor network means fewer • Long life of sensor network means fewer visits by humans visits by humans 1

System Architecture System Architecture Mote Networks Mote Networks Patch Sensor Node � Mica2Dot motes Mica2Dot motes Network � Sensor Patch • • Repackaged Mica2 w/ 1 Repackaged Mica2 w/ 1” ” diam eter diam eter Gateway � 2 kinds of sensors 2 kinds of sensors � Transit Network • • Burrow motes Burrow motes � Detect occupancy via infrared temperature sensors Detect occupancy via infrared temperature sensors � Client Data Browsing Basestation � Ambient temperature/ humidity sensors Ambient temperature/ humidity sensors � and Processing • Weather motes Weather motes • Base-Remote Link � Monitor surface microclimates Monitor surface microclimates � Internet � Temperature, humidity and barometric pressure Temperature, humidity and barometric pressure � sensors sensors Data Service Motes Motes Mote Networks Mote Networks � 2 Networks 2 Networks � • Single hop (433 MHz) Single hop (433 MHz) • • • Multi Multi- - hop (435 MHz) hop (435 MHz) � In both networks, data was sent in In both networks, data was sent in � streaming fashion over an unreliable link streaming fashion over an unreliable link • Required maintaining very little information Required maintaining very little information • about network state about network state • • Oversampled Oversampled environment to m itigate effect of environment to m itigate effect of dropped packets dropped packets Mote Networks Media Access and Routing Mote Networks Media Access and Routing � Single hop Single hop � Low Power Listening (LPL) Low Power Listening (LPL) � � • Discussed last week in B Discussed last week in B- - MAC presentation MAC presentation • Sent packets directly to gateway (no • Sent packets directly to gateway (no • routing) routing) • • Node periodically wakes up to sample radio for Node periodically wakes up to sample radio for activity activity • Sampled sensors every 5 minutes Sampled sensors every 5 minutes • � If idle, returns to sleep If idle, returns to sleep � � Multi Multi- - hop hop � � Else wakes up and receives packet Else wakes up and receives packet � • Routed packets toward gateway • Routed packets toward gateway � Preamble must be longer than sleep interval Preamble must be longer than sleep interval � • Sampled sensors every 20 minutes • Sampled sensors every 20 minutes � Gateway always awake and listening Gateway always awake and listening � • Gateway sent out routing beacons to Gateway sent out routing beacons to • • • Single hop network can send normally sized Single hop network can send normally sized seed network discovery seed network discovery packets packets 2

Media Access and Routing Media Access and Routing Deployed System Deployed System � Stages Stages � Multi Multi- - hop hop � � • • First deployment: June 8 First deployment: June 8 th th , 2003 , 2003 • Nodes selected parent based on most • Nodes selected parent based on most � Single hop, 1 week deployment Single hop, 1 week deployment � • • Second deployment: July 8 Second deployment: July 8 th th reliable link – – Window Mean with Window Mean with reliable link � Multi Multi- - hop, 4 week deployment hop, 4 week deployment � Exponentially Weighted Moving Average Exponentially Weighted Moving Average � Peak deploym ent Peak deploym ent � • • Single hop: 49 motes Single hop: 49 motes (WMEWMA) (WMEWMA) • • Multi- Multi - hop: 98 motes hop: 98 motes • Link quality broadcast to neighbors Link quality broadcast to neighbors • � 115 days of operation yielded 650,000 115 days of operation yielded 650,000 � observations observations every 20 minutes every 20 minutes � Hurricane Isabel forced base station shutdown Hurricane Isabel forced base station shutdown � • Long packets required to support LPL Long packets required to support LPL • • Sept 15- - Oct 9 Oct 9 • Sept 15 � Operation continued through Oct 20 Operation continued through Oct 20 th th � Power Consumption Power Consumption Results vs. Estimates Results vs. Estimates Median Lifetime of Motes (in days) Median Lifetime of Motes (in days) Single Hop Single Hop Estimated Estimated Actual Actual Weather 140 120+ Weather 140 120+ Burrow 127 52 Burrow 127 52 Multi- - hop hop Multi Weather Weather 90 90 63 63 Burrow 80 34 Burrow 80 34 Analysis Multi- -hop Analysis hop Analysis Analysis Multi � LPL lowers cost of listening LPL lowers cost of listening � Why were estimates so far off??? Why were estimates so far off??? � � • • But at higher cost of sending/ receiving But at higher cost of sending/ receiving • SH burrow • SH burrow � Long preamble means longer transmission times Long preamble means longer transmission times � � Overhearing is very costly Overhearing is very costly � Power drawn exceeded battery rating! Power drawn exceeded battery rating! � � • No way to reject packets early • No way to reject packets early • Multi • Multi- - hop motes hop motes • All neighboring nodes receive the packet All neighboring nodes receive the packet • � Ignored forwarding & overhearing cost Ignored forwarding & overhearing cost � • Possible improvements: Possible improvements: • • Power draw from overhearing 8x more than cost Power draw from overhearing 8x more than cost • � Leaf nodes don Leaf nodes don’ ’t need to generate routing beacons t need to generate routing beacons � of single transmission! of single transmission! • Could explicitly reject packet forwarding • Could explicitly reject packet forwarding • • 32% of motes were leaf nodes 32% of motes were leaf nodes • Bad seals • Bad seals � Burrow motes heavily taxed by routing & overhearing Burrow motes heavily taxed by routing & overhearing � • Significant base station outages Significant base station outages • • Prevent them from routing packets • Prevent them from routing packets • Reduce overhearing Reduce overhearing • 3