cenwits: a sensor-based loosely coupled search and rescue system using witnesses jyh-how huang saqib...
TRANSCRIPT
CenWits: A Sensor-Based Loosely Coupled Search and Rescue System Using Witnesses
Jyh-How Huang Saqib Amjad Shivakant MishraDept. of Computer Science, University of Colorado at Boulder
Presented by: Yi Zhang
Most slides taken from Jyh-How Huang’s slides athttp://csel.cs.colorado.edu/~huangjh/Cenwits/CenwitsSlides.ppt
IntroductionIntroduction
Goal: To build a search and rescue system that can pinpoint missing person’s last seen point in wilderness areas Lost hikers, stranded climbers, injured
skiers, … Difficult because of lack of timely
information about the current location “Last seen point” is critical for
search and rescue actions2
Current Search and Rescue Current Search and Rescue Technologies Technologies
The Old School Way – Ask Personal GPS receiver and Satellite
transmitter – Power greedy; Must operate manually to send your location
Localization system and GSM transmitter – Need GSM network coverage
Avalanche beacon/RFID reflector – Limited usage
Need a better, cheaper, reliable system4
Design GoalsDesign Goals
Self-Operate, long life time Small and light weight Non intrusive; no infrastructure
needed Power and memory efficient Cheap($20~$50) Meets security and privacy
requirements5
CenWitsCenWits
A Connection-less Sensor-Based Tracking System Using Witnesses
Comprised of RF sensors GPS receivers Access points Location points Control center
6
How it Works (I)How it Works (I)
Node ID Coordinate Time
10
3
Node ID Coordinate Time
8
6
8
6
2
10
3
1
1 2
7
How it Works (II)How it Works (II)
Node ID Coordinate Time
10
3
8
6 x3, y3, z3 16:58
2
1
Node ID Coordinate Time
…
6 x1, y1, z1 12:31
…
…
6 x2, y2, z2 14:09
…
6 x3, y3, z3 16:58
8
How it Works (III)How it Works (III)
Hiker 6 is reported missing at23:59
Node ID Coordinate Time
…
6 x1, y1, z1 12:31
…
…
6 x2, y2, z2 14:09
…
6 x3, y3, z3 16:58
x1, y1, z1
x2, y2, z2
x3, y3, z3
Inferred location at 23:59
Hot Search Zone
9
Memory ManagementMemory Management
MAX RECORD COUNT Replace the old record with new one of the
same node and keep the total records # of a node less than MAX_RECORD_COUNT
MAX HOP COUNT Don’t forward a packet that has high
probability of reaching AP already MIN RECORD GAP
When 2 records are recorded in a time gap < MIN_RECORD_GAP, replace the old one with new one. 12
Power ManagementPower Management
Beacon frequency adjustment based on speed, time of day, etc.
4-phase hand shake protocol Only transmit as much as the
receiver is willing to take
13
Grouping to Save EnergyGrouping to Save Energy
One active leader at any time; others sleepLeadership time-multiplexed
14
More Memory More Memory ManagementManagement
Partitions: sub-groups containing >= 2 nodesEach of the K partitions receives/sends 1/K of total dataWhen nodes dies: split large partition or merge with small partition
15
Prototype ImplementationPrototype Implementation
MICA2 sensors 900 MHz; 4 KB SDRAM; 128 KB
flash; and 4-512 KB EEPROM Mantis OS 0.9.1b MTS420CA GPS module
Successfully conducted a number of experiments in a indoor environment
16
Experiment 1 (Direct Contact):Experiment 1 (Direct Contact): One hiker One hiker starts from A , goes to B and C, and returns to Astarts from A , goes to B and C, and returns to A
17
Experiment 2 (Indirect Inference):Experiment 2 (Indirect Inference): Hiker 1: A to B and beyondHiker 1: A to B and beyondHiker 2: Does not come in contact with any APHiker 2: Does not come in contact with any APHiker 3: C to B and beyondHiker 3: C to B and beyond
Path of hiker 2 is drawn successfully
18
1
3
Hiker 1 to 5 walk on designated trails while hiker 6 does a random walk. We successfully depict path of hiker 6
Experiment 3 (Identifying Experiment 3 (Identifying Hot Search Areas)Hot Search Areas)
19
ConclusionsConclusions
CenWits has several advantages over other search and rescue systems
No need for a connected network of any kind Power and storage efficiency Cost effective Non intrusive
Suitable for deployment in wilderness areas Applications: Hiking; skiing; wildlife
monitoring; vehicular network
20
Discussion (1)Discussion (1)
Simplistic approach to the resource/inference quality tradeoff How to set memory management
parameters? What’s the impact on inference quality? Ultimate goal is not to save
energy/memory, but lives!
21
Discussion (2)Discussion (2)
Energy really saved by transmitting as much as needed? 4-phase hand shake protocol itself is an
overhead Or maybe 3-phase? (piggyback)
Lack of analysis/quantification Transmission energy when changing
group/partition leader? Energy/memory costs in experiments?
Security and privacy issues22