embracing localization inaccuracy - a case study
DESCRIPTION
In recent years, indoor localization has become a hot research topic with some sophisticated solutions reaching accuracy on the order of ten centimeters. While certain classes of applications can justify the corresponding costs that come with these solutions, a wealth of applications have requirements that can be met at much lower cost by accepting lower accuracy. This paper explores one specific application for monitoring patients in a nursing home, showing that sufficient accuracy can be achieved with a carefully designed deployment of low-cost wireless sensor network nodes in combination with a simple RSSI-based localization technique. Notably our solution uses a single radio sample per period, a number that is much lower than similar approaches. This greatly eases the power burden of the nodes, resulting in a significant lifetime increase. This paper evaluates a concrete deployment from summer 2012 composed of fixed anchor motes throughout one floor of a nursing home and mobile units carried by patients. We show how two localization algorithms perform and demonstrate a clear improvement by following a set of simple guidelines to tune the anchor node placement. We show both quantitatively and qualitatively that the results meet the functional and non-functional system requirements.TRANSCRIPT
Usman Raza
Amy L. Murphy
Gian Pietro Picco
Embracing Localization Inaccuracy:
A Case Study
1
2
Localization in wireless sensor networks has been studied for
a decade now
Increasing complex localization techniques to achieve tens of
cm accuracy
Use costly specialized hardware (UWB, Antenna arrays)
Experiences from the real environments are still limited!
Motivation
Goals of This Work
3
Not to propose yet another localization technique
Evaluate localization techniques in a real-world nursing home
Requirement are representative of diverse localization systems
Unveil the relationship between system level performance and application level objectives
What a WSN geek want? vs. What an end user want?
To give guidelines to improve both the system level performance and end user satisfaction
4 4
Higher quality of life for impaired and elderly in nursing homes
Monitoring to the medical support staff
A single floor of a nursing home in Trento, Italy
• 10 Public spaces, 20 Patients, 4 Nurses
Monitoring in Nursing Homes - Services
5
Doctors
Offline evaluation of patient movement
To assess the general health of the patient
To diagnose the progression of Alzheimer's disease
Nurses
Real time use of approximate patient location
To find the patient
To raise an alarm if patient leaves the facility
6
WSN Localization - Architecture
Anchors(x,y)
Mobile
Proximity Detection Localization Technique
Localization
7 7
WSN Localization - Architecture
Anchors(x,y)
Mobile
Proximity Detection Localization Technique
Localization
Localization Technique
8 8
WSN Localization - Architecture
Anchors(x,y)
Mobile
Proximity Detection Localization Technique
Localization
Localization Technique Proximity Detection
(x,y) (x,y)
(x,y)
9
WSN Localization - Architecture
Anchors(x,y)
Mobile
Proximity Detection Localization Technique
Localization
Localization Technique Proximity Detection
(x,y) (x,y)
(x,y)
Visualization
Mobile(x,y)
10
WSN Localization - Architecture
Anchors(x,y)
Mobile
Proximity Detection Localization Technique
Localization
Localization Technique Proximity Detection
(x,y) (x,y)
(x,y)
Visualization
Mobile(x,y)
WSN geek – How far is the estimate to the actual position?
Nurse – Does the visualization allows me to find the patient?
11
Objective
Identify the proximity of the mobile
patient to an anchor
Requirement – Low maintenance
Energy Efficiency
Localization
Step 1– Proximity Detection
Sleep Box-MAC-LPL Anchor
Mobile Sleep Sleep
Visualization Refresh Interval
Localization Technique
Lifetime Anchors - 45 days (2 AA batteries)
Mobile - 5 days (1 coin battery)
System Design
A custom proximity detection protocol
12
Max-RSSI Localization
Localizes patient at the anchor detecting proximity with maximum signal strength
Patient
Proximity Detection Localization
Step 2– Localization Techniques
13
Max-RSSI Localization
Localizes patient at the anchor detecting proximity with maximum signal strength
Relative Span Exponential
Weighted Localization (REWL)
Localizes patient at weighted
centroid of all anchor coordinates
Patient Patient
Proximity Detection Localization
Step 2– Localization Techniques
Experimental Evaluation
14
System level: WSN geek’s perspective
Accuracy: Percentage of time the patient is correctly detected
in its current area/room
Application level: Nurse's Perspective
Satisfaction level : moderate, fair, excellent!
System Level Evaluation - Accuracy
Max
-RSS
I R
EW
L
82%
87% 77%
89%
89%
94%
92% 72%
85%
95%
78% 72%
91%
97%
95%
91% 60%
88%
Living Area
Corridor
Bedroom
Bathroom
Living Area
Living Area
Corridor
Bedroom
Bathroom
Living Area
- Reasonable accuracy
- No clear winner
- REWL outperforms in
key areas e.g. living and
exit areas
15
Application Level Evaluation -
Methodology
16
Nurse
Operator was asked to evaluate multiple localization and
visualizations
Find the patient
Raise an alarm
Patient Icon
Patient Tracking GUI
Application Level Evaluation - Results
17
Quality of localization system depends heavily on the
visualization
Moderate Excellent! Fair
Max RSSI - at anchor REWL - at x,y Max RSSI- at room center
18
Low Accuracy ≠ Unacceptable solution to end user
REWL 72%
65% 52% 84% 79% 57%
91%
Living Area
Corridor
Bedroom
Bathroom
Living Area
34%
Experience with our Initial deployment:
Low accuracy
Positive qualitative evaluation
Application Level Evaluation - Results
Guidelines for Anchor Placement
19
Maximize the distance between anchor nodes deployed in the adjacent areas
Place the nodes near the center of the monitored area
Use the radio shieling of
obstacle to your advantage!
Principle: Minimize the likelihood of signal reception across the
monitored areas
Conclusion
20
Simple low cost localization system is enough
for many applications
Low accuracy can still be acceptable to end
user
21
Usman Raza, Amy Murphy, Gian Pietro Picco, “Embracing Localization Inaccuracy – A Case
Study ”, April 2013, IEEE International Conference on Intelligent Sensors, Sensor Networks
and Information Processing (ISSNIP), Melbourne, Australia
Full Text (Pre-print) : http://disi.unitn.it/~raza/Papers/DISI_TR_12_038.pdf
Thank You!