An AGPS-Based Elderly Tracking System

Transcription

1 An AGPS-Based Elderly Tracking System Albert Kai-sun Wong, Tim Kam Woo, Albert Ting-Leung Lee, Xiaoming Xiao, Vincent Wing-Hei Luk, Kwok Wai Cheng Department of Electronic and Computer Engineering Hong Kong University of Science and Technology Clearwater Bay, Hong Kong Abstract The design of an experimental AGPS-based (Assisted Global Positioning System) elderly tracking system is described. The system includes: a wearable AGPS terminal with HSPA two-way communication capability and designed for 10 days of continuous battery operation, a GPS assistance data server with reference GPS stations, location database and server, application server, and web server and client. Assistance data is retrieved by the wearable AGPS terminal using the SUPL protocol (Secured User Plane Location). This paper describes the design of each component based on key considerations such as accuracy, availability, battery life-time, and user behavior. Keywords AGPS (Assisted Global Positioning System), SUPL (Secure User Plane Location), elderly tracking I. INTRODUCTION GPS [1] is a mature technology for accurate localization in clear outdoor environments. One short-coming of GPS for people tracking, in addition to its unavailability indoor, is its large time-to-first-fix (TTFF). In cold and warm starts, the TTFF is generally in the 40 seconds to 2 minutes range. It is fundamentally limited by the data rate at which navigation messages are broadcasted from the GPS satellite 0 seconds for each complete navigation message and the additional time required for performing frequency searches when the satellite Doppler frequencies are unknown. In people tracking applications, with battery size and life-time often a key consideration, operating GPS in hot mode is often not an option. AGPS is naturally fitted here. For people tracking applications, two-way communication capabilities is generally required for location information to be reported, so the terminal is naturally equipped with the capability to access GPS assistance data from the network. With AGPS, the TTFF can be reduced to matter of seconds, greatly reducing the power consumption and improving the tracking performance and response speed. Because of the avoidance of frequency search, the AGPS standard specification allows a reduction of 18 db in the signal required for the first fix compared to GPS. This added margin enhances the success rate of location fixes in a variety of environments. Like many aging societies in the developed world, Hong Kong is facing an increasing need to enhance the quality of life for the elderly, many of which living independently. The need exists for a location-aware emergency detection device and service to assist care providers or relatives to support the living elderly persons. With inputs and support from a non-profit elderly service organization in Hong Kong serving over 10,000 members, a leading electronic manufacturing firm, and a leading mobile operator in Hong Kong, we developed an AGPS based prototype elderly tracking system and began field test of the system in year-end The system include the wearable G-enabled AGPS terminals, GPS reference stations for generating GPS assistance data in Hong Kong, SUPL (Secured User Plane Location) server, web-engine, web-client, and centralized database. The retrieval of the assistance data is referenced to the known cellular base-station ID detected and provided by the wearable AGPS terminal. We have made use of base-station ID s provided by the mobile operator, and we have also undertaken an effort to scan base-station ID s across the territory. In the initial phase, staff members and healthy elderly persons serve as volunteer carriers of the AGPS terminal. In later phases, we plan to extend the trial to a larger set of elderly individuals. In the paper, we will describe the architecture of the prototype system and the different considerations that we have made concerning different sampling intervals, battery lifetime, strategies for dealing with the unavailability of GPS in indoor environments, and different geo-fencing, alerting and alarming algorithms. II. OVERALL ARCHITECTURE Figure 1 below illustrates the components and overall architecture of our system. Flash Web Client HTML / Javascript Dojoroot Web Engine Web Server AJAX Django (PYTHON) Location / Users Groups / Geofencing Wearable AGPS Terminal IMS GPS RF Location/Cell Location Server ANS.1 ELC Server AGPS / Locator MYSQL Database SUPL GPS Reference Station GPS Receiver ARM7 Processor AGPS Update Collector Server (Java) AGPS Data Store Server (Java) Figure 1. Components of the AGPS Elderly Tracking System G RF G Baseband + Processor Proprietary AGPS Protocol UDP SSL /09/$ IEEE 100 ICUFN 2009

2 Initially, our prototype system functions in the MS-based (Mobile Station-based) localization mode in which the mobile terminal s positioning is computed by the mobile terminal itself. In the long-term, we expect that the system will operate in the MS-assisted mode where positioning will be computed in the network employing hybrid capabilities (for example, mixed AGPS, cellular, and WiFi positioning). In the following, we will provide a summary of each component in the system. 1. The AGPS Terminal - Consists of a G baseband chip, a G RF chip, and a GPS RF chip. The G baseband chip provides the terminal s HSPA (High Speed Packet Access) connectivity with the network. The processor within the G baseband chip is the master of the AGPS terminal, being where all the operation and application logics of the terminal reside, and where the GPS positioning computation is performed using input signals from the GPS RF chip. We implement different applications algorithms and power management strategies through the programming of this processor. The AGPS terminal is powered by a 770mAhr lithium-ion battery, and equipped with an emergency buzzer and a LED status display. 2. Mobile Location Server (MLS) - It is a SUPL server that provides a single point of contact to the AGPS terminals. It serves the GPS assistance data to the AGPS terminals. Communications between the AGPS terminal and the MLS is based on SUPL v1 with ASN.1 encoding. We implemented also a proprietary protocol for experimental purposes. On a pre-set schedule or upon command from the MLS, the AGPS terminals perform location fixes and report their locations.. GPS Reference Station (GRS) - It is a simple device consisting of a GPS receiver and an ARM7 processor. It receives the navigation messages from all of the observable satellites and forwards these messages to the AGPS collector and AGPS data store, which reformats the information into GPS assistance data. Initially, one GPS Reference Station (GRS) is built and placed in a central Kowloon rooftop location sufficient to provide the geographic coverage for our initial trial. Pictures of the AGPS terminal and the GRS are shown below. 4. The AGPS Collector and Data Store - The AGPS collector is located within close physical proximity to the GRS. It relays the GRS data into the network. The AGPS data store captures and reformats data from multiple GRSs into AGPS assistance data. 5. MySQL Database - The system is based on a data-centric design and all operation and historic data are stored in a MySQL database. 6. Web Engine - It implements various applications in support of the user Web Client: the web server, terminal locations and display, geofencing, user group management, etc. The Web Engine is developed using the Django framework. Django is an open source web application framework written in Python in support of the Model-View-Controller (MVC) design pattern [4][5]. The MVC design pattern decouples the application logic and data storage from user interface considerations. Django allows us to develop the system in an incremental fashion without finalizing the entity-relationship model at the very beginning. 7. The Web Client - The Web Client provides the care providers with the web interface to the various tracking applications. It makes use of AJAX (Asynchronous JavaScript and XML) and the Dojo Toolkit. AJAX is a collection of web programming techniques for creating interactive applications running on a web browser. It is adopted by Google Map [6] which is used by the Web Client for various map applications. The Dojo Toolkit [6] is an open source JavaScript library for DOM (Document Object Model) manipulation, asynchronous communication, and client-side/server-side data storage management. It provides a set of tools for handling common issues experienced in cross-platform web site developments. III. OPERATION MODEL The AGPS terminal communicates with the MLS normally through the Open Mobile Alliance (OMA) defined SUPL protocol and is known as a SUPL Enabled Terminal (SET). The principle of AGPS operation is to supply the SET with GPS assistance data through the wireless network as a side channel so that the SET will not have to obtain this data directly from the GPS satellite. There are four major parameters included in the assistance data. First is the satellite ephemeris which is a set of the latest orbital data of the satellites in view for precisely determining each satellite exact position at a given time. Knowledge of which satellites are in view and their Doppler frequencies allows the SET to avoid frequency searches. Second is a reference time so that the SET can have an immediate estimate of its own clock offset. Third is a reference location which gives the SET a good starting estimate of its position for the SET s position and clock offset computation which is typically done through iteratively solving a set of four or more non-linear equations derived from the pseudo-range measurements. Fourth is the atmospheric correction data to account for variations in signal propagation speed through the atmosphere at different times. The atmospheric correction data can give an improvement in positioning accuracy by up to 50%, or 7 meters. With AGPS, the GPS terminal can obtain a position fix from cold start within seconds instead of the tens of seconds usually needed for pure GPS. The signal level required for initial position fix is also reduced by -18 db from -142 dbm to -160 dbm. 101

3 Parameter Satellite Ephemeris Reference Time Reference Location Atmospheric Correction Valid Period 2 4 hours Up to 20 seconds (assuming no handovers) Minutes and depends on velocity of SET Hours Both the SET Initiated Reporting and the Network Initiated Reporting methods are implemented in our system. In SET Initiated Reporting, the AGPS terminal periodically initiates contact with the MLS with a request for assistance data while informing the MLS of its current cellular network CELL ID. The MLS retrieves the assistance data from the MySQL database and forwards the data to the AGPS terminal, which then detects the satellite signal, compute its own position, and report the position to the MLS. A simplified model of the interaction between the MLS location server and the AGPS terminal (SET), without an authentication procedure, is shown in Figure 2 below. Location Server Request (CELL ID) SUPL END (Result Position) SET In our initial trial, we have focused on collecting initial performance data and user feedbacks. Our system has primarily been operating in the SET Initiated Reporting mode. IV. AVAILABILITY AND ACCURACY In this section, we present availability and accuracy data collected through initial trials with healthy elderly individuals as carriers of the AGPS terminal. The data is compared against stationary AGPS testing results in Hong Kong urban outdoor settings and against controlled pedestrian environments. The comparison allows us to understand what may contribute to degradation in availability and accuracy in a real application environment. In Figure 4 below, we show the testing results for the case when the AGPS terminal is placed in an open field environment on the HKUST campus. Position fixes were repeated 100 times at a regular interval using forced cold start (the AGPS terminal was forced to ignore any previously acquired data). The result shows that there were successful position fixes 100% of the time. The mean position error is meters, and the mean response time is second. The response time measures the complete AGPS delay, from the time the AGPS wants to determine its position, to the time the AGPS terminal has successfully done so. The maximum position error and the maximum response time are meters and 25.8 seconds respectively. The result also shows the average number of satellites that should be in view, and the percentage of the in-view satellites that were detected at different signal levels. Figure 2. Operation Model of SET Initiated Reporting In Network Initiated Reporting, a client may triggers a position fix of the AGPS terminal by initiating contact through the location server, as shown in Figure below. Client Location Server SET SET Location Request SUPL INIT Request (CELL ID) SET Location SUPL END (Result Position) True position = ( , ) type, yield min max 95% 67% 50% mean accur = 100/ times = 100/ #SVs in view: (avg) not detected -160: :-10-10: % 6.% 42.1% 4.8% Figure. Operation Model of Network Initiated Reporting Figure 4. AGPS Performance in Open Sky Environment 102

4 In Figure 5 below, we show the stationary test results when the AGPS terminal was placed on Lyndhurst Street in Central, one of the most GPS adverse urban canyons in Hong Kong. The results show that even though many of the satellites (47.99%) are not detected, the terminal successfully fixed its position 98% of the time. The mean position error was increased to.8 meters, and the mean delay to second. In one controlled experiment, we invited a group of 1 healthy elderly individuals to take walks around a predetermined route around an elderly activity center. The walk took anywhere from 1 to 2 hours. The individuals were told to wear the AGPS terminal with a neck string. They might stop and rest whenever necessary. The area was surrounded by 25- storey high-rise residential buildings, as shown in Figure 6 below. Figure 6. Walk Path in a Controlled Tracking Experiment True position = ( , ) type, yield min max 95% 67% 50% mean accur = 98/ times = 98/ #SVs in view: (avg) not detected -160: :-10-10: % 22.12% 7.8% 22.09% Figure 5. AGPS Performance in Deep Urban Canyon It should be noted that the positioning accuracy can be significantly improved if we allow the AGPS terminal to operate in the tracking mode (hot start), in which it makes frequency position determination based on previous positions. In a GPS friendly environment (Location A1) and an adverse environment (Location A2), a mean position error of 10 meters and 16 meters can be achieved respectively as shown in Table 1 below. The results were obtained with the terminal left stationary while performing position fixes in the tracking mode. The positioning delay is around 1 second. A systematic bias was observed that brought the inaccuracy in the result. Again, the AGPS terminal was programmed to perform SET Initiated position fixes at intervals of 40 seconds to 60 seconds with forced cold start. The results in Figure 7 show that a terminal failed to fix its position in a large percentage of times, and that the failure rate varied greatly across different individuals, who could be identified by the IMSI (International Mobile Subscriber Identity) of the SIM card that was placed into the AGPS terminal that they wore. For two individuals, the success rate dropped to around 0 percent, probably because of how the AGPS terminal was worn. For one individual, the AGPS terminal was placed inside a pocket and the success rate dropped to zero. But in all the cases with nonzero success rate, the successful position fixes were sufficient to provide an indication of the tracks of the individuals, as shown in Figure 8. This depends, however, on the fact that the positioning intervals were relatively short (mostly 60 seconds) and the individuals were not moving at a high speed. Accuracy (m) Location A1 Location A2 Mean Standard Deviation Table 1. AGPS/GPS Accuracy in Tracking Mode Figure 7. Success Rates of the Controlled Tracking Experiment 103

5 Figure 8. Tracks Generated by the Successful Fixes Again, the positioning performance can be significantly improved when we operate the AGPS terminal in the tracking mode. In Table 2 below, we show the results when four terminals were set to operate in the tracking mode and compute a position fix every 5 seconds. A fix was considered a failure if it failed to complete in 5 seconds. We see that a position fix was achieved within 5 seconds around 85% of the time. AGPS is sufficient for providing an accurate indication of the location of the individuals as long as they are outdoor, even in urban canyon environments. network at anytime. Over extended periods when no activity is expected, the terminal can be set into the SLEEP mode, where the G module must first be awaken by the internal clock for it to initiate contact with or be contacted by the network. However, the reduction in the background current by going into the SLEEP mode is found to be very small. So in any event, the maximum continuous operation interval is limited to around 1 days, or 11 days with an allowed safety margin factor of 1.2. When the GPS module or the G module is turned on, additional current is consumed: 100 ma and around 00 ma respectively for the two modules. An AGPS position fix consumes on average two seconds of operation of each device for a total of 0.22 mahr. Allowing a margin factor of 1.2, we can compute: 770 Hours of Battery Operation = n where n is the number of minutes between position fixes. The hours of continuous operation as a function of the AGPS positioning intervals are shown in Figure 9 below. Figure 9. Hours of Battery Operation versus Positioning Interval V. BATTERY CONSUMPTION It is desirable to have a terminal that can support 10 days of continuous battery operation to accommodate the typical house visitation schedule in local elderly service applications. The AGPS terminal is equipped with a 770mAhr.7V lithium-ion battery and can be set into one of three modes: GPS Module G Module SLEEP off sleep STANDBY off standby OPERATING on on In the STANDBY mode, the terminal consumes a measured current of 2.55 ma, which includes the current to sustain a real-time clock and a status LED. In the STANDBY mode the G module is ready to be contacted by the G To increase the battery capacity further will bring size and safety issues. Seven days of continuous operation can be achieved with the current battery capacity if we limit the positioning interval to no less than 10 minutes. In the future, we will investigate different algorithms for adjusting the positioning interval in a dynamic and intelligent way. Also, since a large part of the energy consumption is in the G RF front-end, energy consumption can be reduced by avoiding AGPS assistance data retrieval when possible. For example, for the most important part of the assistance data which is the satellite ephemeris, the validity period is as long as 2-4 hours. The terminal can obtain position fixes with inaccurate reference time and reference location at a penalty of an increased position error [9]. There could also be an increased in the GPS computation time which may in turn use additional energy but the impact of this requires further study. 104

6 VI. ADD-ON APPLICATIONS AND FEEDBACK FROM USERS We have also developed a geo-fencing application. An administrator may conveniently draw a geo-fencing area that is any arbitrary polygon on the map. The system can be configured to generate a notification when the AGPS terminal carrier is outside of the geo-fencing area. A screen capture of this geo-fencing application is shown in Figure 10 below. varies across elderly individuals. Some do not mind carrying of this terminal in a conspicuous manner. Others dislike the stigma that wearing such a device may bring. For elderly individuals who are mobile phone users, a GPS-enabled phone would eliminate the need of a dedicated AGPS tracker device. Further study is needed to understand the target elderly user group for this dedicated AGPS terminal and to enhance its features and design. ACKNOWLEDGMENT The authors would like express our gratitude to Giant Wireless Technology, PCCW, and the Evangelical Lutheran Church Social Services of Hong Kong for their support of this work. REFERENCES Figure 10. Geofencing Application Screen Capture Our trial is being conducted at a gradual pace. Our initial results show that for people tracking, AGPS will work quite well in the outdoor or semi-open environments. Augmenting solutions may be needed to provide position information when the individual is indoor. Acceptance of the AGPS terminal [1] GPS, Theory, Algorithms and Applications, 2 nd edition, Guochang Xu, Springer 2007, ISBN [2] OMA-AD-SUPL-V1_ D, Secure User Plane Location Archiecture, June 2005 [] OMA-AD-SUPL-V1_ D, Secure User Plane Location Requirements, June 2005 [4] [5] [6] [7] [8] [9] M D Karunanayake, M E cannon and G Lachapelle, Analysis of on AGPS performance, Measurement Science and Technology, 18 (2007)