While visiting the Louvre museum, Brooks had several experiences he wanted to share. He took a number of photographs and leaves some of them at that location for others who visit the place. A couple of years later, Red visits the Louvre and chances upon the photographs left by Brooks. Red reads Brooks’ experiences and finds out about a lesser known section of the castle. When leaving, Red leaves some of his own pictures and anecdotes there. People who visit the Louvre are able to look at it at different points in the past by browsing through blog posts, pictures and videos left by others. In essence, people leave their memories which others can view and feel when they visit that location.
1. Overview To enable the scenario described above, in this paper we present a novel approach for automatically tagging information published by a user with location data. The ability to search on location tagged information like news, blog posts, pictures, videos and ads is going to have a strong presence in the future. Our scenarios show the compelling applications that can be built on top of location tagged information.
The reach of current localization techniques such as GPS and war driving is not sufficient to support large scale services. In this paper, we present a novel approach to use existing Wi-Fi and GSM infrastructure around the world as location landmarks. For most of the applications we discuss the absolute location of the user is not needed. Our approach is readily deployable because it works with existing mobile devices and does not require modification of clients or infrastructure. These location based services will give Microsoft unique competitive advantages in mobile platform space and online advertising.
2. More Scenarios
Red visits a restaurant in the mall and is able to see reviews and comments that other people left at the restaurant. He finds that one of the dishes is loved by many people and he decides to try it.
As Red walks through a busy city street, he searches for nearby coffee shops. He is instantly served with ads for nearby coffee shops. He also receives a discount coupon from the nearby Starbucks Café.
Brooks is sitting at his house and wants to know about a tourist place. He locates the place in the map and searches for blog posts, photos and videos people uploaded from the place.
Red is waiting at the airport for his flight. He searches for people nearby, he can play a game with. He discovers that Brooks is also searching for people to play with.
Red visits a new city. He wants to talk to people who visited the same location in the past. He finds that Brooks visited the place some time ago. He is able to query him and ask about good hotels and places to see.
3. Introduction Mobile devices and contextual information retrieval are now allowing us to travel with information. But software is still stuck in the old paradigm that expects users to explicitly declare context and location. A major step forward would be to organize, tag and link information to specific locations . Such location based services can provide Microsoft with an advantage in the growing advertising business. This year, online advertising is a $40 billion business. It will grow to $80 billion by 2010 and will continue to increase in the years beyond . One of the key reasons for success in this market is that the ads are targeted – the ads delivered to the end-user are relevant to the website visited by the user, or the search query typed by the user.
Location based advertising is currently facing problems of reach. The number of GPS or GSM localization enabled mobile devices remains relatively small. Indeed, many players are already testing the waters in this market. For example, Google has announced to offer ad-supported free Wi-Fi access in San Francisco. MSN has introduced the “Locate Me” feature, which can determine the approximate location of the user based on what Wi-Fi networks the user can hear. Google CEO Eric Schmidt recently expressed the company's interest in location based ads as the future in mobile advertising . Last year Google launched a mobile version of AdSense and their own mobile OS. We can definitely expect location aware ads from them in the near future. Microsoft is appropriately placed to become a leader in this space.
Microsoft's strong presence as a mobile platform and its large user base in the online messaging space can be leveraged to provide compelling situated experiences. The Live Messenger and Hotmail together form a massive network of more than 12 billion relationships . New context rich, location aware services built on top of this existing social network will allow users to extend their interactions in rich mediated-spaces.
We propose location based services that will allow Microsoft unique competitive advantages in the messaging and mobile platforms space. One obvious advantage is that MSR has already filed a number of relevant patent applications on these ideas; we believe that Microsoft has two inherent structural advantages.
4. Technology Existing implementations of location based services [8, 9] require additional functionality to track the device’s location. This usually means additional hardware requirements like GPS or extensive war driving to map GSM (using triangulation) and Wi-Fi signature to latitude and longitude. GPS is extremely power hungry and does not work inside buildings such as malls.
Features like MSN’s “Locate Me” and Google’s location apps use war driving data and locates users geographically using nearby Wi-Fi Access Points and GSM towers. For the scenarios and applications we discuss we do not require the absolute location of a user and hence war driving is unnecessary.
The system described leverages GPS or war driving data when available to improve the overall experience. However, in its absence we use the following approach for localization.
The density of Wi-Fi Access Points (APs) has increased rapidly . Within the last five years, there has been remarkable growth in the number of Wi-Fi hotspots and in the usage of Wi-Fi enabled devices – recent research has shown that this trend is likely to continue .
Every Access Point has a 6-byte unique ID called the BSSID. A Wi-Fi client can scan for Access Points around and get their BSSIDs. Access points once deployed are usually immobile and hence their BSSIDs can be used as the landmark for that location. It should be noted here that clients can scan for Access Points around even when they are completely disconnected.
Fig. 1 Architecture If Wi-Fi Access Points are not available, we rely on GSM towers to function as landmarks. The wide scale deployment of GSM towers around the world has enabled users to be connected all the time. We use the GSM tower ID as a landmark in the same way we use APs. Using GSM tower as landmark info even in the presence of Wi-Fi enables us to build smarter location services as explained in the Applications section. In the scenario shown in Figure 1, AP1 and GSM1 form the landmarks.
When a user publishes data to a central server, we automatically tag the information with relevant landmarks. Once location tagged information is published to the server, other users who visit the same location can retrieve it by reporting the current location’s landmarks.
The approach is readily deployable as we do not rely on absolute location. Neither war driving nor modification to clients, APs or GSM towers is necessary. We only make the assumption that the client device (PDA or laptop) has a Wi-Fi card or a SIM card and can connect to the Internet.
We could also adopt this mechanism to passively perform much richer war driving of location information. Users with GPS enabled devices could opt in to provide specific location information to map Wi-Fi signatures and GSM signatures to exact location (latitude, longitude).
This technique of publishing and retrieving information from the server with location information enables us to build several applications as discussed in the Applications Section.
Wi-Fi cards are power hungry and hence people often switch them off and prefer GPRS or EDGE connections. When the current location’s Wi-Fi signature is required we turn on Wi-Fi, scan for Access Points around and turn off Wi-Fi. Since the Wi-Fi card is used only to scan on demand, our approach is not power hungry.
It is not mandatory that a client needs to be connected to the Internet to publish information from a location. The location’s Wi-Fi and GSM signatures can be obtained even if the client is not connected and hence can the information can be tagged. The client can upload it later when he regains Internet connectivity.
5. Applications Post-Its in the air
Our application enables users to annotate physical locations with media. Users can post text, images and videos on their blog. Behind the scene the post is tagged with the GSM and Wi-Fi signatures at that location. This will enable users to tag locations with their memories and experiences. Other people passing through the same location can serendipitously chance upon the located media, and leave further comments and anecdotes.
Some scenarios describing such interactions are - People can post reviews about a restaurant after they eat and others visiting the restaurant later can read and these reviews to choose a popular dish. Users might like to leave photos they took at a beach for others to comment on. Users can leave directions at a hotel to a nearby lagoon that is not shown on maps.
The application allows users to leave behind traces of their experiences and memories in-situ. At any place, users can navigate and comment on media submitted by other users. They can continue to track posts tagged to that location and also follow posts by a particular user.
Location Sensitive Search
Over time, a knowledge base of information tagged with location will be formed. The Internet search of tomorrow can be in touch with the physical world. People can search for information filtered by the contents’ physical location as against the location in which the content is stored. "M.G. Road" for instance appears in most of the major cities of India; however, the disambiguation would be easy when the location on the map is specified with search queries.
Location Sensitive Advertisements
Online advertising is the one of the fastest growing markets with a net worth of over 40 billion USD . Microsoft realizes this fact and has invested large amounts of money in the advertising potential of networks such as FaceBook and MySpace. The high hit to view ratio of context sensitive ads can be improved phenomenally by taking the location of the user into account. Location sensitive ads are the new wave in the online advertising industry.
We look at a scenario wherein, advertisers would bid for ads to be broadcast to users who see a particular Wi-Fi Access Point or GSM Tower. Location sensitiveness can be combined with existing search engine marketing techniques like bidding on keywords, etc to provide more targeted advertisements to the user.
We make use of relative distance between Access Points (APs) to provide a richer experience. For instance, if a user is browsing for information on a product, we could link him to nearby stores selling similar products. Technical details of how we infer nearness using only Wi-Fi APs and GSM Towers without war driving data is explained in Appendix A.
Discount and offer coupons from participating businesses can also be delivered to a client. A user with the client application walking near a coffee shop receives a virtual coupon from the shop. The user can then walk down to the coffee shop and present the coupon. Delivering location sensitive advertisements with coupons greatly adds to the users’ experience.
Friends near and who were here
Current social networking sites, model their contents as a tree of friends and their friends. Many of these sites also provide us options to updates of their friends. We introduce a mechanism to provide additional context to the social networking field in order to make it more useful. Users of our products regularly upload their GSM and Wi-Fi signatures to the central server (to subscribe to location sensitive material of that region).
A by-product of knowing the current location of users is that, users will be able to know if any of his friends are currently nearby. Microsoft has multiple revenue generating possibilities in this space. Users can also fine friends who were in the same region at some point in the past. He will be able to contact them for additional information about the place. Users will also be allowed to contact the authors of Post-Its left at that place. This can be restricted using incentive based schemes to reduce spam issues. Also it will be made possible to optionally disable random people from contacting the Post-It authors.
In order to overcome privacy issues, it will be possible to turn off regular updating of user's location.
Mobile gaming is expected to be a $10 billion market by 2009 . A major portion of this growth is going to be in location-based games. Using our technology users can search for others around them interested in playing games. These games can be played over peer-to-peer connections, or over the internet. This can be extended to other scenarios where a user is looking for assistance with a particular application.
Our location services can also identify users that are reading the same web page or book. Mobile devices are fast becoming the preferred platform for reading books. The reading experience can be enriched by finding other users in the vicinity, interested in the same topics, web pages or books.
6. Conclusion Location sensitive content distribution is an upcoming field and Microsoft can become an early player in the market. Our technology is readily deployable. As compared against existing solutions which attempt to provide location sensitive information, we place no specific demands on the client or infrastructure. Removing the requirement of GPS receivers also enables us to attract lower end smart phones. We can leverage Microsoft's existing technologies to easily build our solution. The LiveID platform provides an effective means to perform authentication. Post-Its in the air will add rich context to Virtual Earth and Live Search for location specific search. MSN Messenger and Internet Explorer are foundations on which the technology can be deployed. Services such as finding nearby friends will add great value to the large user base of these technologies. Given Microsoft's existing technological prowess, we have a painless path to dive into the field of location sensitive services.
7. References  Worldwide PDA and SmartPhone Forecast. http://www.etforecasts.com/products/ES SP-PDA.htm.
Appendix A - Inferring nearness using only Wi-Fi APs and GSM Towers
GSM Cell base stations typically have a range of few hundred meters to few kilometers. We have found that in developing countries like India, there is a very dense deployment of GSM cell towers making localization using base stations very effective. Wi-Fi on the other hand provides fine grained localization ranging up to 250 meters.
Figure 2 - Relative range of GSM vs. Wi-Fi We need not know the absolute location (in terms of latitude and longitude) of Wi-Fi APs in order to infer their nearness. Given that the GSM cell tower range is much more than the range of a Wi-Fi Access Point, we can safely assume that the coverage of a single GSM tower would include multiple Wi-Fi APs within itself (Refer to Figure 2). Due to existing limitations in Windows Mobile, we can programmatically only infer the currently associated GSM tower. Also, the same operator could have multiple cell towers with over lapping ranges. Given that users would be periodically uploading their GSM and Wi-Fi signatures to the central database, over time a 2 level tree (Refer to Figure 3) would be associating Wi-Fi APs with GSM cell towers.
Figure 3 – Wi-Fi and GSM landmarks association tree Now when a user reports to the server a Wi-Fi, GSM signature, we can obtain from the tree, other GSM towers in the range of the Wi-Fi AP, and the other Wi-Fi APs associated with this set of GSM towers. We can thus infer that these Wi-Fi APs would be physically near.