Location Polling Algorithm for Alerting Service Based on Location

Transcription

1 Location Polling Algorithm for Alerting Service Based on Location Byung-Ik Ahn 1, Sung-Bong Yang 1, Heui-Chae Jin 2, and Jin-Yul Lee 3 1 Dept. of Computer Science, Yonsei Univ Seoul, Korea biahn@pointi.com, yang@cs.yonsei.ac.kr 2 Dept. of E-Business, Cheonan Univ Cheonan, Korea hcjin@paran.com 3 LBS Lab., POINT-I Co, Ltd., Seoul, Korea jinylee@pointi.com Abstract. Location Alerting Service is a push service that informs users whether predefined conditions be satisfied based on periodic acquisitions of location data from mobile terminals for several purposes such as safety. Polling the location of terminal is one of the basic functional requirements of this service. In this paper, an analysis of the location polling technique will be given, we will propose an efficient location polling algorithm and present the results of experiments with this algorithm. 1 Introduction Location Alerting Service is a service that notifies users through SMS or provides a specific service that already has been defined by the users in the case of event such as entering, being and outing occurring in a specific region or established territory by monitoring the location of terminal users in the mobile communication network[1][2]. Location Alerting Service is a very personalized push type service. As for the examples of such service, there are Location Advertising Service, L-Commerce, Location Meeting/Matching Service, Polluted Area Alerting Service, Disaster Sensing Service, and Distribution Control Service. An important technique of such location alerting service is either concerned about which method the location generation equipment uses or about scheduling for location acquisition. In relation to the type of location generation equipment, whether the location determination equipment is a server or a terminal becomes an issue and in relation to the scheduling for location acquisition, kind of method used in determining the location of mobile terminal either periodically or by calculation within the established time becomes an issue. Generally, the location determination equipment inspects and reports trigger condition by determining the location at the server or terminal at a fixed interval, for example, every five minutes, regardless of each method. However, a system that has large number of users, such as mobile communication, causes excessive location determination, and ultimately increases the network load. Therefore, this paper will examine techniques that could accurately perceive and report trigger conditions while reducing the number of position acquisition by K.-J. Li and C. Vangenot (Eds.): W2GIS 2005, LNCS 3833, pp , Springer-Verlag Berlin Heidelberg 2005

2 Location Polling Algorithm for Alerting Service Based on Location 105 dynamically changing the position acquisition period and based on such examination, will present and evaluate an algorithm of efficient location determination method for alerting service. Section 2 examines location polling technique and method for alert service and Section 3 derives strengths, weaknesses and problems of location polling method and techniques. Section 4 develops a new AlertPoint algorithm based on the existing problems and Section 5 evaluates this method through an experiment. 2 Location Polling Technique and Method A tracking technique for the location of mobile terminal in the location alerting service is the location polling technique. Location polling technique is divided into network polling technique(server polling technique) and intelligent polling technique (terminal polling technique) based on whether the location determination agent is a server or terminal in accordance with the location determination method. Server polling technique uses determination method of Cell-ID, TOQQ, AOA, TDOA, E-OTD and MS-Assisted GPS, and terminal determination technique uses determination method of MS-Based GPS and S-GPS[8][9]. 2.1 Server Location Polling Technique Server location polling technique is a technique that polls the location information using the internal server of mobile communication and wireless network. A method that uses MSC/HLR (Mobile Station Center / Home Location Register)[1] and a method that uses location acquisition polling server have been used to materialize the existing location alerting service[4]. The method that uses MSC/HLR is a method of acquiring location from the process of shifting the location of terminal user to HLR through sensing the update of VLR (Visitor Location Register). This requires a change of software directly on MSC/HLR. The method that uses position acquisition polling server is a method that acquires the location of user through requests to MSC and MPC of mobile communication companies by dynamically scheduling the interval of position acquisition time. The former of these two methods possesses a security level that is difficult for an ordinary service company to approach and also has the problem of having to directly adjust MSC/HLR. On the other side, the latter method is easily accomplished through a contract with each communication network company. Consequently, majority of existing location alerting service applies the latter model. Location polling server is composed of location acquisition model engine, location request scheduler and alerting engine. The location acquisition interface is the standard for location request/reply between the location polling server and wireless mobile communication and mobile RAN network and is designated as MLP(Mobile Location Protocol) in the case of mobile communication network. Push interface is the standard for alerting engine to transmit messages. The location acquisition model engine is an important factor that calculates the location acquisition time intervals through algorithm. The location request scheduler schedules all of the calculated location acquisition time intervals. Lastly, the alerting engine inspects the trigger conditions with the location information of the mobile terminal and alerts through the push interface when satisfied. This paper describes a location acquisition model, which is an important technology of location acquisition model engine.

3 106 B.-I. Ahn et al. Fig. 1. Structure of Server Location Polling Service 2.2 Terminal Location Polling Technique Terminal location polling technique is a method that could be applied to MS-based GPS and S-GPS determination techniques. This technique allows possible location determination by the terminal itself based on the initial information and data communication occurs between the server and the terminal to attempt synchronization with the sever when it diverges from the established threshold or satisfies the optional conditions. The attempt of synchronization between the server and terminal is a process of synchronizing the conditions calculated based on the territory of spatial trigger requested from the server to the terminal. Reliability on sensing the event occurring at the time when the trigger conditions have been satisfied can be acquired through such synchronization. Since this technique does not use the network determination method, there is an advantage of not having to consider the network load from the determination, but also possesses a disadvantage where network load communication for attempting synchronization exists. The greatest weakness is not being able to load GPS chip of high accuracy due to the function of mobile terminal and many of the mobile terminal don t mount GPS chip. As shown in Figure 2, there exist Location Assistant of the terminal, Location Cash Database of the server, Tracking Module and Alerting Engine in the terminal location polling server and is divided into Synchronization Interface between Location Assistant and Tracking Module and Push Interface between Alerting Engine and Alerting Application. Location Assistant of the terminal carries out location determination and obtains present geographical location information of the terminal by communicating with artificial satellite and also evaluates conditions for synchronization with the server and communicates with the server. Tracking Module of the server calculates and transmits new condition to the terminal based on the location information obtained from the terminal. Alerting Engine evaluates trigger condition based on location information and transmits message to Alerting Application.

4 Location Polling Algorithm for Alerting Service Based on Location 107 Fig. 2. Structure of Terminal Location Polling Service 3 Examples of Position Acquisition Method and Service 3.1 Position Acquisition Model When storing location information of the terminal or acquiring location information, Position Acquisition Model aims to decrease overhead on network load and communication cost by requesting location information on the mobile communication network and LBS platform and reducing the number of acquisition within the range of not reducing the reliability of the location information. Position Acquisition Model is classified into a method that does not request unnecessary location information by analyzing mobile pattern and a method that dynamically reduces the number of inquiries on location information by appropriately adjusting the intervals of location acquisition time. The flowing methods are being used[6][7][10]. Static Acquisition Model. Acquires location by setting the position acquisition time at regular intervals Distance-Based Acquisition Model. Adjusts position acquisition time interval through changes in amount of distance based on standard distance Respective Distance-Based Acquisition Model. Adjusts position acquisition time interval through differences in the standard distance of each terminal user and continuous condition changes of the terminal location Group-Based Acquisition Model. Adjusts position acquisition time interval based on the amount of change in MBR(Minimum Boundary Rectangle) to the recent distance

5 108 B.-I. Ahn et al. Predict-Based Acquisition Model. Adjusts position acquisition time interval by predicting the next mobile location through a verified prediction model using direction and speed, which are the mobile information of terminal users in the past Dynamic Acquisition Model. Operates the same with Static Acquisition Model but adjusts position acquisition time interval from system load Among the position acquisition models, Static Acquisition Model, Distance-based Acquisition Model, Respective Distance-based Acquisition Model, and Group-based Acquisition Model are designed to minimize the number of position acquisition and maximize perception rate of event occurrence by adjusting position acquisition time interval based on important parameter of each model. However, on the other hand, Predict-based Acquisition Model and Dynamic Acquisition Model are models that stabilize the system by appropriately adjusting or apportioning time intervals from system load. Fig. 3. Concept of Position Acquisition Model Position Acquisition Model generally used up until now is Static Acquisition Model. It is a method that carries out location determination of mobile terminal at a fixed interval of time. The number of position acquisition is 144 times in 10 minute intervals based on 24 hour standard and it corresponds to 140 million times if service request for total of 1 million people. Such would cause network traffic and reduce economical adequacy of provided service. Therefore, the purpose of location polling technique is to perceive a significant point of time when event will occur to provide service with the minimum number of position acquisition. The minimum number of position acquisition and the event perception rate are closely related to one another. In other words, as the number of position acquisition increases within a specific time, the period of position acquisition reduces and then causes more accurate perception of the point of time when the event would occur. However, this brings decline in the function of overall system with the load increase of mobile communication network. On the other side, when the number of position acquisition decreases, the perception rate for event occurrence reduces and causes decrease in the reliability of service. Thus, it is necessary to minimize the number of position acquisition within the range where the perception rate of event occurrence could be trusted.

6 Location Polling Algorithm for Alerting Service Based on Location Service Examples WaveAlert Model. WaveMarket uses a method that supports the search for nearest alert request through Scheduler that adjusts position acquisition time by calculating MATT (Minimum Alert Triggering Time) and EAUT (Earliest Available Update Time) and territory division grouping of alert request area[3]. MATT is a future time value acquired from dividing the calculated distance (Euclid distance, Shortest Path distance, etc.) from the position of terminal user to the user s aimed location by the calculated maximum speed. EAUT is a renewal time with reception for the data acquired from request to renew a new MATT. In such case, previously acquired data is disregarded. In other words, as shown in Figure 4, the location data determined from the time between EAUT and MATT is valid and such data is used in the renewal of next MATT or EAUT. Fig. 4. Concept of MATT & EAUT This solution has an outstanding function in the number of minimum position acquisition and event perception rate. On the other side, there is a disadvantage of high calculation cost of the software since algorithm for regional division of alert request is complex. As a solution to such problem, WaveAlert realizes division algorithm with hardware. Cell ID Synchronization Location Polling Technique. Cell ID Synchronization Location Polling Technique is a method that minimizes the system load of mobile communication network by distinguishing a point of polling time on its own using location assistant software and receiving trigger assistant information from the trigger server by expanding the function of mobile terminal[4]. There is an advantage of minimizing the network load by carrying out precise position acquisition through mobile communication network when it exists in Cell ID list once the Location Assistant software within the terminal is transmitted to currently communicated Cell ID by transmitting base station Cell ID information, which includes Alert Zone, to the terminal. However, Cell ID has the network load where the server has to transmit Cell ID of alert zone to the terminal when a new alert request is received and needs support in the hardware part of the terminal equipment in order to bring the communicated base station ID, which is Cell ID. Moreover, Location Assistant software must exist in memory. The most important disadvantage is the fact that the number of precise position acquisition increases only when the zone that includes the registered Cell ID in the terminal is large or numerous cell IDs exist.

7 110 B.-I. Ahn et al. Update Policy Synchronization Location Polling Technique. Update Policy Synchronization Location Poling Technique has the server transmit update policy to the terminal and the terminal acquires position using GPS determination technique at a fixed time interval at the time of initial operation. The acquired position continues with the position acquisition at a fixed interval of time if it does not satisfy the server and the synchronized update policy. There are distance-based policy, vector-based policy, and street network-based policy and they differ accordingly to each service circumstances. The strength of this technique is the relatively great reduction in network load from the updated polices but with the burden on the terminal from the continuous location determination and the consumers tendency to refuse, it can t be applied as a location polling technique for alert serve in a vast range of zones. 4 AlertPoint Location Polling Algorithm AlertPoint adjusts the position acquisition time based on Distance-Ratio based Acquisition Model that bases on moving distance ratio and proposes algorithm for insert process of grouping and new request for one to multiple requests. Fig. 5. Entering the Alert Zone Set α is the ratio of distance (d current ) of current time (t current ) and before time (t before ) to distance (d before ) of before time (t before ) and former time (t 0 ). When 0 < α < 1, position acquisition time (t next ) increases by t (1/α t unit ) and α 1, position acquisition time (t next ) decreases by t (α t unit ). This technique is similar to the concept of Respective Distance-based Acquisition Model but it carries out dynamic time adjustment and controls the infinitely increasing t by calculating static valid time section accordingly to the distance with the Buffer Zone for high perception rate of event occurrence. Moreover, the grouping process from alert request reduces the number of position acquisition by distinguishing whether or not to carry out the position acquisition for the alert request. As shown in Figure 6, when the location of terminal user is included in the Buffer Zone, time adjustment is set as the time value to minimize the possibility of the user s location getting out of the alert zone.

8 Location Polling Algorithm for Alerting Service Based on Location 111 Fig. 6. Leaving the Alert Zone Fig. 7. AlertPoint Position Acquisition Model

9 112 B.-I. Ahn et al. Figure 7 shows a diagram of AlertPoint s Position Acquisition Model. When the ratio of moving distance used to obtain the time interval of position acquisition is very dramatic, the value comes out to be very large. This problem is resolved by adjusting the minimal moving distance (D min ) and maximum moving distance (D max ) by searching for the minimum error in position acquisition through the mobile communication network and the distance affected by base station. AlertPoint Solution supports the grouping process of alert request. Accordingly, it reduces the number of position acquisition from unconditional position acquisition when a new request is inserted. 5 Experiment and Test 1,000 sample data have been generated for the performance test on data distribution per moving distance on AlertPoint. The alert zone entering of this data is 95% and the alert zone non-entering is 5%. 16 hours of location data has been traced. As for the characteristic of the data, 50% was the case (for instance, housewives, students, office workers) that repeatedly stop and move at 2-3 locations, 25% was the case that continuously move to various locations after making stops, 15% was the case that move to various locations without long stop and 10% was the case that move to a wide zone without any stop. The following Figure 8 shows the speed distribution of the sample group. Fig. 8. Mobility Distribution of Sample Group Let s experiment the optimal performance function by setting the standard radius of Alert Zone to be 1 km and using 5 minutes for Tmin, which is generally used in such experiment. Find the optimal value from the test and Vmax values of 20, 40, 50, 60, 70, 80, 90, 100, 120 Km when the Buffer Size is 1, 2, 3, 4, 5, 6 Km. The evaluation of the overall function at this time is set as, Performance = (Total number of Alert engine location inquiry 100)/total number of periodically location inquiry (5 minutes) and Alert Event Perception Rate is Alert Event = (Total number of Final Alert Request(100)/Total number of Alerting Request. The results of experiment in accordance with each variable are shown by Figure 9 and 10.

10 Location Polling Algorithm for Alerting Service Based on Location 113 Fig. 9. Alert Event and Performance by Buffer Size Fig. 10. Alert Event and Performance from Vmax Change According to the result of this experiment, it has been verified that Alerting Event Perception Rate demonstrates over 95% of performance in most cases while improving the performance by reducing the number of location inquiry for alert by over 60% in general. The result advices that using AlertPoint method, which could perceive 95% of event, is efficient if the importance of alert is not especially an urgent matter. Moreover, AlertPoint method has been recognized that it can be used appropriately in the reduction of network load when a large number of mobile terminal users request location alert since it can overcome the weakness of fixed time interval method of Static Acquisition Model, which has been generally used up until now. 6 Conclusion Location-based Alerting Service has been developing as diversified and advanced personal service based on accurate location determination with the advancement of wireless determination method and the improvement in terminal performance. Consequently, the number of subscribers using the wireless network will increase and the wireless network load will gradually increase as well. Accordingly, mutual comparative evaluation between the reduction in the number of location determination and the reliability of service was necessary for the purpose of minimizing the network load.

11 114 B.-I. Ahn et al. This paper aims to use more efficient location polling method by analyzing location polling technique and presenting a new method to resolve problems. The location polling method presented in the paper proposes a method of appropriate scheduling and does not fall behind Static Acquisition Model, which has been developed previously and used up until now, in performance such as location inquiry and Alert Event Perception. If there is a gradual increase in the use of location-based service and the necessity of alerting service, AlertPoint method seems efficient in managing the system and the network load. References 1. Ki Jun Han. : Standardization and Research Development of Location-Based Service (LBS). Informationization Policy Volume 10, 4 th Edition (2003) 2. Korea IT Industry Promotion Agency. : Location-Based Service (LBS), Software Market News of Korea IT Industry Promotion Agency (2003) 3. WaveMarket Incorporated. : A System that Provides Alert-Based Service to Base Stations from Wireless Communication Network. Korea s Public Patent Notification, Application # (2004) 4. Min, K. W., Cho, D. S.: Moving Object Position Acquisition Technique for LBS. Thesis Magazine Vol.1 (2003) of Korea Information Processing Society 5. Yun, J. K., Kim, D. O. & Han. K. J.: Development of Real-Time Mobile GIS supporting the open Location Service. Proc. of Geotec Event Conference.(2003) 6. Dieter Pfoser, Christian S. Jensen. : Capturing the Uncertainty of Moving-Object Representations. SSD (1999) pp Ouri Wolfson, S. Chamberlain, S. Dao, L. Jiang, G. Mendez. : Cost and Imprecision in Modeling the Position of Moving Objects. Proceedings of the Fourteenth International Conference on Data Engineering (ICDE14) (1998) 8. OGC. : OpenGIS Location Services (OpenLS): Core Services, (2004) 9. Green, J., Betti, D. & Davison, J. : Mobile Location Services: Market Strategies, White Paper., Ovum, (2000) 10. A. Prasad Sistla, Ouri Wolfson, Sam Chamberlain, Son Dao. : Querying the Uncertain Position of Moving Objects. Temporal Databases, Dagstuhl (1997) T-S. Yeh, B. de Cambray. : Modeling Highly Variable Spatio-Temporal Data. PRiSM TR, 1994/50(1994) 12. Guttman, : A.:R-trees: a Dynamic Index Structure for Spatial Searching. In Proceedings of the ACM-SIGMOD Conference on the Management of Data, pp (1984) 13. Luca Forlizzi, Ralf Hartmut Guting, Enrico Nardelli, Markus Schneider. : A Data Model and Data Structures for Moving Objects Databases. SIGMOD Conference (2000) GPP : TS V chapter 9.5, (2004) 15. Yufei Tao, Dimitris Papadias. : Efficient Historical R-trees. IEEE SSDBM, (2001) 16. Yannis Theodoridis, Michael Vazirgiannis, Timos Sellis. : Spatio-Temporal Indexing for Large Multimedia Applications. ICMCS (1996) 17. Dieter Pfoser, Christian S.Jensen, Yannis Theodoridis. : Novel Approaches to the Indexing of Moving Object Trajectories. VLDB (2000) 18. Dieter Pfoser, Yannis Theodoridis, and Christian S. Jensen. : Indexing Trajectories of Moving Point Objects. CHOROCHRONOS Technical Report CH (1999)