1.1 Motivation
The vast development of wireless network and mobile communication technologies, the widely use of handheld and laptop devices, and the popularity of Global Positioning System (GPS) make Location-Based Services (LBS) attractive to a lot of people. In recent years, many powerful mobile phones are available to the market. The tiny devices have 3G, Wireless LAN, and Bluetooth connectivities, and many are equipped with two cameras and GPS receiver. User can not only use them to take pictures but also can use LBS to enjoy the intelligent lifestyle.
LBS services are capable of providing information based on the user’s location [1-3]. The application of LBS services is ubiquitous in our life. Every morning, we can get local weather information based on our location. If you have a meeting with somebody, you can just follow the directions of your phone and know where the person is. When driving, users can get turn by turn navigation to places. The traffic information can be used to help us not to get stuck on traffic, which saves time and be more environmental friendly. On the way to destination, a user can get the information of nearby gas stations and the lowest gas price on the route. Before arrival, you can use LBS to help you find a parking place. If you want to use public transportation, you also can get the schedule and its real-time status. After work, you see e-coupons on the phone sent by the restaurant in the city you lives. You can choose whatever you what and make a reservation. On the way home, you can see where your children are on the phone and pick them up. When got home, you can feel free to relax because the air conditioner has been ready. All this can be done from a
2
handheld device.
LBS Queries are submitted by users of LBS Services. The server receives a query request, processes it, and then sends result back to the user. There are many varieties of LBS Query [4]. The type of query can be range query [5], nearest-neighbor query [6], and many variants. The range query in LBS is that given a region area, we can retrieve the interested information in the area. The nearest-neighbor query is to find the closet point of interest in a space. The duration of the query can be snapshot or continuously. The snapshot query is one-shot. The result is static. The continuously query continuously update its result until the user terminate the query request. The time of the query can be past, present, and future. The query and object in the query can be stationary or moving. For example, the user at home sent a query request for the nearest hospital. The type of this query is nearest-neighbor query. The duration is snapshot. The time is present. The query and objects in this case are stationary. The police officer sent a query request for continuous report the number of cars on the freeway. The type of this query is range query. The duration is continuous. The time is present and future. The query is stationary and the objects in this case are moving. The user driving on the highway may be interested in what are the nearest gas stations for the next hour. The type of this query is nearest-neighbor query. The duration is continuous [7]. The time is future. The query is moving and the objects in this case are stationary. The other user at home may be interested in what is the closest distance between him and his friends yesterday. The type of this query is closest point query [8]. The duration is snapshot. The time is past. The query and the objects are moving.
LBSs are usually implemented in client-server architecture [9]. The server processes the requests from the client side. The client side is usually mobile hosts.
3
Mobile hosts usually have limited computation ability, battery capacity and wireless bandwidth. To ensure the correctness of query results, the client side usually periodically reports its position.
When the user population is huge and many requests are sent to the server. The server will be a bottleneck. The scalability of the LBS will also become a problem.
There are also other architectures for implementation of LBS. The idea of distributed client-server architecture is to let clients share their computing ability. Divide the work into server site processing and client site processing. The server managers and maintains queries, and act as a mediate of requests and clients. The clients monitor the nearby queries and send the updates of the query result. Another approach is decentralized peer to peer architecture [10-11] . This type of LBS system does not have central control or knowledge of all nodes.
In this thesis, the architectures we used to implement LBS Services is assumed to be distributed client-server architecture. The type of query is range query. The duration is continuous. The time is present. The query is stationary and the objects are moving. This query is also known as range monitoring query. The definition is: given a set of rectangle regions, we want to know the mobile hosts inside them. And we can get real-time update when the mobile hosts move in and out of these regions.
1.2 Research Questions
In detail, the primary research question that we address is as follows. The result of range monitoring query is dynamic. Mobile hosts may frequently move in and out of query region. The result may only valid for a very short time. Continuous update the results or periodically updates make heavy and unbalanced load on the server. The mobile hosts also need to communicate to the server through Wireless LAN or Mobile
4
communication interface, which is battery consuming. There are also huge messages on the network. Decreasing the location update requests not only can lessen the server load but also save the battery power of mobile hosts. Although the range monitoring query process was introduced in many papers [12-16], we present a more effective algorithm based on the location of mobile hosts.
In distributed client-server architecture of LBS, the division of work into server side is processing and client side processing is still a problem to resolve. Using the client recourse better can make the LBS more scalable. The distributed client-server architecture of LBS was used in previous work. However, we present an algorithm based on the capability of the mobile host.
Another related question is that mobile hosts have different capability. How to allocate suitable work that meets the capability of every mobile host is important. We proposed the algorithm that tries to fit every mobile host’s capability.
1.3 Related work
To deal with the problem of huge location update in range monitoring query, Prabhakar et al. proposed the concept of Safe Region[17]. A safe region is defined to be a circular or a rectangle region that contains mobile host’s location but not overlap with any query boundary. If mobile host is inside the safe region, it does not affect any query in the system. If it moves out of the safe region, it instructs the server and requests a new safe region. Because the safe region cannot overlap with query boundary, the size of safe region is usually small when the query density is high.
Unfortunately, computing a rectangular safe region takes from O(n) to O(nlog3n), where n is the number of queries. When adding new queries to the system, the server may also need to compute all safe regions for the mobile hosts because the new query
5
may affect existing safe regions.
Ying et al. proposed a scalable and adaptive technique, Monitoring Query Management (MQM)[13], for real-time processing of range monitoring queries. The resident domain concept they proposed is that a server assigns each mobile host a resident domain (region) that contains a number of queries based on mobile host’s current location and ability. The server also notifies the mobile host the queries that overlap with that domain. When a mobile host detects that it has cross over some query boundary, it contacts the server to updates the query results. When a mobile host leaves the resident domain, it reports to the server to get a new resident domain.
They allow the mobile host to monitor its nearby queries and report the result to the server.
To leverage the computing capability of mobile host, MQM proposed a BP-Tree (Binary Partitioning Tree) spatial access method in the server for efficient query management. BP-Tree can be used to find the resident domain with a number of queries that meets the capability of the mobile host. The detail design of MQM will be described in Chapter 2.
1.4 Overview
This dissertation is organized as follows. Chapter 2 presents related work.
Chapter 3 presents the algorithms and system architecture. Chapter 4 discusses the results in the system design. Finally, Chapter 5 concludes this thesis and describes the future work.
6