HartFi: An Energy-Efficient Localization System

Song Han, Tianji Li, Chen Qian, Douglas Leith, Aloysius K. Mok, Simon S. Lam

Introduction

2

• Location-based applications (LBA) are becoming increasingly popular– Foursquare, Facebook Places, Gowalla, MicroBlog, TrafficSense,

Pothole Patrol, PeopleNet …– Most LBAs are running on mobile devices.– Many LBAs benefit from having Loc info available all the time.

• Tradeoff among existing localization techniques– Energy efficiency, estimation accuracy, coverage …

Introduction (Cont.)

3

• Pros and Cons of existing localization techniques– GPS:

• Most accurate localization• Energy hungry, does not work in indoor environment

– Cellular network:• Less energy consumption• Coverage in both indoor/outdoor environment• Estimation error is large

– Wi-Fi:• Estimation accuracy and energy efficiency lie between those of GPS

and cellular network

Motivation

4

• Use low-power interface to propagate location information

• Features of 802.15.4-based interface:– Radio consumes up to two orders of magnitude less energy than

Wi-Fi . (1mW for 802.15.4-based radio and 32-200mW for Wi-Fi).– Data rate (250 kbps) is sufficient for location-based applications.– Tree or mesh structure can help achieve large coverage.

Motivation

5

• HartFi: An energy-efficient localization system – Collaborative:

• Devices have Wi-Fi and WirelessHART interfaces combined together– Energy efficient:

• Only a few devices acquire location info through Wi-Fi directly• Using WirelessHART interface to propagate <loc, rss> info

– Accurate: • A novel approach for reducing error propagation

– Coverage:• A multiple-hop mesh network is formed among WirelessHART interfaces

HartFi Infrastructure

6

• A device in HartFi system has multiple interfaces– GPS, Wi-Fi, Bluetooth, WirelessHART…– WH interfaces form a mesh network – All other interfaces work as usual

• A unique coordinator in the system– Manages the network and schedules

all the transmissions– As an option, reports location

information to the location server for other applications

Localization Algorithm

8

HartFi defaults to Wi-Fi localization system if there is no WH interface. Support incremental deployment.

9

Reducing Estimation Error Propagation

• Using confidence-based iterative localization– Run localization algorithms for multiple rounds– In each round, each node, say u, computes a confidence value for

every three devices that have locations.

la

lb

lc

ld

u

Devices a, b, c ,d have locations. Hence there are 4 combinations and 4 confidences values for u.

10

Reducing Estimation Error Propagation– Confidence of a trilateration:

– Node u uses the three devices with max confidence value for trilateration. Then it sets that value as the confidence of its estimated location.

– Confidence value <= 1. An AP’s confidence is set to 1. If device u computes its location using node v’s location, Cu<Cv.

Product of confidences of references

Error of the distance measurement

HartFi Hardware and Testbed Setup

11

• Asus netbook (Atheroes AR5007EG wireless network adapter)

• WirelessHART interface

Energy Consumption Measurements

12

• Experiment Setup– Connect the netbook to an AP through Wi-Fi interface. Ping the AP every

1-second.– Connect the netbook to a WH coordinator through WH interface. Send

“hello” messages with the same size as the ping packet.– Measure the time (in minutes) taken to consume half of the full battery

capacity.

Both Off WH On Wi-Fi On Saving

Netbook 1 92 82 58 70.588%

Netbook 2 115 107 84 87%

Netbook 3 123 117 88 80%

Netbook 4 91 81 61 63.33%

Using WirelessHART interface leads to 75.23% power saving

Inferring Distance From RSS (Cont.)

13

Comparable coverage

More accurate RSS model

Simulation Results

– 100 nodes in 100m x 100m square area and 12 Wi-Fi APs– Pure WiFi localization also requires node sharing location

information, because not every node can find 3 nearby APs.– Iterative localization of HartFi runs 5 rounds.– Errors of distance measurements are based on experiment traces

– A unit of energy cost is the cost to send a WirelessHART msg. – When the number of nodes increases, less portion of nodes

requires Wi-Fi localization. Hence the average cost decreases.

Future Work

16

• Enhance WH network with Ad-hoc mode– How to select the new coordinator in face of dynamics?– How to make communication/configuration consistent?– How to keep synchronization when the coordinator changes?

• Simulations in large-scale HartFi system

• Testbed experiments in outdoor/indoor environment– Comparison of localization accuracy between Wi-Fi and HartFi system– Performance on HartFi incremental deployment