Mesh WiFi Location and Context - Aware Mobile Services in
Transcription
Mesh WiFi Location and Context - Aware Mobile Services in
Mesh WiFi location and context-aware mobile services in stations D. Sanz SNCF, Paris, France Abstract This paper presents the research work carried out by SNCF within the domain of Indoor mesh WiFi location and the manner SNCF would ideally use the location information in order to provide contextual mobile services for passengers in stations. Firstly, the context of the project and its objectives are introduced. Secondly, the experimentation realized with a WiFi-based location system in Paris-Nord Station is presented. Lastly, different mobile contextual services imagined for passengers in stations are introduced. Introduction In 2003, SNCF deployed its first WiFi hotspot in Paris-Nord Station. Since then, the 52 more important train stations in France have been equipped with WiFi hotspots. In parallel to the deployment of these WiFi hotspots, SNCF’s Large Stations Department (Direction des Gares et de l’Escale), deployed the “GaresdeFrance.com” website, which allows SNCF customers to obtain information about each train station: services and shops in the station, real-time depart/arrival train schedules, access to the station and other practical information… This website was developed in order to deliver services to SNCF customers at home but also on the move and in particular in the stations that have a WiFi hotspot. This mobile version of the website was free of use and available to all the users of SNCF station’s WiFi hotspots equipped with WiFi laptops or PocketPC terminals. The deployment of this website represented the first SNCF effort dedicated to improve customer information in stations through the use of their own mobile terminals. Since then, a new version of the website has been created (called “gares-en-mouvement.com”), mobile phone information services have been deployed, and mobile phone ticketing services experimented… All these projects were created in order to attempt to follow the fantastic development and diffusion of mobile communication technologies observed during the last years: GSM, GPRS, EDGE, UMTS, HSDPA, 802.11b/a/g (WiFi) and soon HSUPA, 802.11n and WiMAX… These technology evolutions and their acceptance by the public demonstrate how much mobile services are destined to become “the services of tomorrow”. That’s why SNCF’s Innovation and Research Department launched in 2005 the research project “Context-aware mobile services in stations”, which aimed at: - evaluating the application of Mesh WiFi wireless technologies in order to deliver location based services (LBS) - through the experimentation of innovative applications for customers that take into account the context of the user in order to deliver better and more intelligent information services to the customers This paper presents the work carried out within the framework of this project in order to achieve these two goals. WiFi location and Mesh WiFi networks When IEEE 802.11 standard (WiFi) was created, the objective was to propose a standard to provide wireless connectivity to an Ethernet network. Today, WiFi can be found in all kinds of terminals such as laptops, PDAs, hybrid mobile phones… and also in various types of equipments such as printers, slide projectors, USB-keys, triple-play Internet “boxes” at home,… The norm continues to evolve, as well as the use of this powerful technology… Among these new uses, one of the most remarkable is the use of WiFi for indoor location purposes. Two main applications of this WiFi terminal location can be foreseen: - WiFi network security-related applications, in order to locate a pirate terminal or access point which attempts to connect to a private WiFi network. - The provision of WiFi-based LBS (Location Based Services) for users. In SNCF’s project, only the provision of LBS was considered. Two main location techniques can be used in order to geo-locate a WiFi terminal within a WiFi coverage area: triangulation and calibration. Triangulation-based technique contains the precise information regarding the position of access points of the network and calculates the position of the WiFi terminal from the signal level received from the terminal at each access point. The location server then calculates the position of the user by simple triangulation of received signals (at least three access points are needed). The second one uses a calibration procedure that memorizes in the location server the signal level observed by the mobile terminal coming from all the available access points when positioned at some specific “reference” points. Once the calibration is made, the client in the user terminal simply sends to the location server the power level measured for the different access points. Then the server compares the power levels with the calibration information and sends the location information to the user terminal client. Both techniques have advantages and drawbacks. The first technique is slightly less efficient as it is unknown if the received signal has followed reflections before arriving to the access points. The second is more efficient but there is a need of installation for dedicated software in the mobile terminal. The first one is more of a hardware-oriented solution, as it needs specific infrastructures (access points). The second one is more of a software-oriented solution. SNCF stations have been equipped with WiFi networks. As a result SNCF could not choose a specific hardware (access points) and selected the Ekahau calibration-based location system for the experimentation. One of the advantages of this solution is that it can be easily interfaced with an application server in order to use location information for LBS purposes. Figure 1 shows the service architecture. When a user wants to access a contextual service on the terminal, first (1) the location client in the user terminal measures the power level received from the different access points detected (A, B, C), and sends the information to the location engine, which uses the calibration data to transform the information sent by the user terminal client into coordinates on a map (x, y, floor). Then (2), the location engine gives these coordinates to the application server. Finally (3), the application server customizes the information demanded by the user in order to provide him with the desired contextual service. SWITCH 2 APPLICATION SERVER 3 1 LOCATION ENGINE (A,B, C) ( x , y, floor ) AP WiFi 2 y AP WiFi 1 B A PERSONNALIZED CONTEXTUAL CONTENT C x AP WiFi 3 MOBILE TERMINAL WITH PARIS-NORD STATION HALL LOCATION CLIENT Figure 1: Service architecture Figure 2 shows the calibration work carried out at Paris-Nord station’s hall. Figure 2: calibration points at Paris-Nord station hall Each calibration point was recorded every 5 to 10 meters. These points will be the reference for the location engine, which will evaluate the real-time position taking into account this calibration measures, also taking into account the previous positions of the user terminal and the access points estimated position (to estimate the position of an access point is easy if the calibration phase is correctly made). SNCF experimentation demonstrated that, with this calibration-engine, the location accuracy was 1 meter, in a laboratory, where optimal conditions are respected (that is, in which the environment has not changed between the calibration phase and the moment of the measure). In Paris-Nord Station however, the measured accuracy was on average about 10 meters… and errors larger than 20 meters were observed. The explanation for this large error regarding location performance comes from the fact that there were not enough access points in the station (4 AP but located in only two locations), but also due to the strong variability of the physical environment: train movement, presence of trains and other vehicles such as supply trolleys, crowd movements, mobile stands, presence of kiosks… In order to improve the location accuracy, a large number of access points must be added, in order to ensure that every location in the station is covered by the signal of at least 3 access points (the higher the access points are placed the better, in order to avoid reflections of the signal). However, WiFi deployments in a large train station reveals complicated and expensive… resulting in SNCF deciding to analyze WiFi Mesh techniques. Mesh WiFi Networks Originally, IEEE 802.11 standard was created to allow wireless connectivity to an Ethernet LAN network. Though WiFi continues to evolve since its creation, its operating modes remain rudimentary, and within 99% of the applications, the basic “infrastructure mode” is used. In the “infrastructure mode”, an access point is wired to a backbone network (LAN, MAN, WAN), allowing mobile terminals to access the wired network resource. As a result, these access points, in general, have to be physically connected by a wire to a backbone network. When deploying a WiFi hotspot within a large site such as for example a Paris railway station, the cabling constraints result in much effort in terms of complexity of installation/deployment nad costs. The next WiFi revolution will be that of the Mesh Networking techniques. Mesh Networking is based on the implementation of intelligent and dynamic routing algorithms and auto-configuration techniques (both coming from mobile ad-hoc networks’ domain) inside the WiFi access points, so that the access point will be capable to discover the network and to connect to another access point wirelessly, creating a kind of "multi-hop wireless infrastructure": the "meshed network". A Mesh WiFi network is, in general, composed of 3 kinds of nodes: - “Gateway” nodes, which are wired to the backbone network, - “Mobile router” nodes, which play the role of wireless access points (allowing the user terminal to connect to the network), and are connected to the rest of the network by wireless means. - Mobile terminal nodes. The user terminals. INTERNET Wired Network R GW GW : Gateway Access Point R : Router Access Points : User terminal R R R R Figure 3: Mesh network architecture A gateway node can manage up to 7 or 8 wireless routers, but generic deployments use no more than 3 or 4 wireless routers per gateway, due to the throughput loss generated by each hop in the wireless nodes. In average, a wireless access node divides by a factor 2 the available throughput since the wireless node generates and listens traffic in both directions: towards the wired network and towards the rest of the wireless network. This fast reduction on the available throughput can still be accentuated when a single radio is used for both the infrastructure links and the user’s access... That’s why many of the proprietary solutions in the market of Mesh WiFi technologies use two different radios, one for the infrastructure links and one for the users’ access, thus improving the throughput and the global capacity of the network. Mesh WiFi networks use the same frequency bands and power limitations of 802.11bg and 802.11a networks, that is: either the 2,45 GHz ISM (Industrial, Scietific, Medical) frequency band, either the 5 GHz ISM band, either both of them. In terms of security, Mesh WiFi networks deliver the same level of security of standard WiFi networks (WEP, WPA, WPA2 TKIP + AES, 802.1x, 802.11i) and in addition, they implement in general, VPN tunnels in the infrastructure segment of the network, in order to avoid pirate connections at the infrastructure level. In addition to the advantages previously described, intelligent routing on wireless access points allows, when the multi-hop Mesh Network is dense enough, to implement multiple routes between a source and a destination, resulting in self-healing capacity (if an access point breaks down, the rest of the network will automatically re-organise in order to create new routes). This characteristic of wireless Mesh networks is invaluable in the context of large deployments and industrial "sensitive" applications. WiFi Mesh standard is in the process of development (802.11s). However, there are many proprietary products already existing and in use for large scale deployments, mainly in America and Asia. SNCF has decided to analyze and experiment these technologies immediately, without waiting for the standard, in order to evaluate their advantages and drawbacks within a large on field deployment (this time in Paris-Montparnasse Station), and to be prepared to utilise the standard version as soon as it will be available. Figures 4 and 5 show the results of the site survey (radio coverage study) realized at ParisMontparnasse station. In Figure 4 it can be observed that 20 access points are needed if we want a dense coverage of the station’s hall. Figure 5 shows the perfect coverage of the station’s hall (green colour means that WiFi coverage is obtained at the best throughput level (54 Mbps for 802.11g) taking into account a receptor sensibility of -66dBm. Only 3 of these 20 access points (AP12, AP15 and AP19) will be wire connected to the stations’ LAN backbone, hence avoiding network cabling installations for 17 of them! In addition to such reduction in complexity and cost, the density of the proposed mesh WiFi architecture will represent a double interest: - to be able to react to access points breakdowns, topology changes, etc… - to optimise the accuracy given by a location engine using this infrastructure in order to provide the user with LBS. The accuracy in presence of the required access points (correctly positioned) is estimated to 3 to 5 meters. Figure 4: Paris Montparnasse Mesh WiFi site survey results. Access points’ locations [1] Figure 5: Paris Montparnasse Mesh WiFi site survey results. Radio coverage @ -66 dBm. [1] Context-aware mobile services in stations Once the coverage and the location precision issues of the WiFi wireless network were studied, the works of the project were focused on the development of the context-aware applications SNCF has imagined. But what has to be understood by “context-aware applications”? Applications that use all the available information about the actual situation of the customer and about who he is (profile concept), in order to deliver him the more possible targeted information. Two large sets of information have to be considered then: - Context elements: time of the day, day of the week, day of the month, in which station is he? Where in the station is he? Is it the first/second time he enters this station today? Are we in a rush hour of this station? Is there any particular event in the station at the moment (renovation work, an exhibition…)? - Personalisation elements (profile) (see Figure 6 bellow): email address (to identify him and to keep in touch with him), password (to allow him to manage his own profile for the service), first and mid name, gender, age, civil state (is he married and/or has he some children? and if yes, can we know if they’re also with him at the moment?), is he a disabled person?, favourite train routes and stations (during the week and during the weekend), does he want to receive automatic push information/offers from a list of specific service providers?… These lists are, of course, not exhaustive, and even if some people may think that accessing all this information can be a little bit intrusive, it has to be considered that each customer will allow the system or not to locate him, or to know his profile details in a voluntary way, depending on how much interesting the proposed services are for him. It has to be pointed out that customers are already used to give their profile to Internet websites towards whom they are confident of. Perhaps one day, customers will be ready to share even their agenda information (for example: I have a meeting at 3pm at Paris Congress Centre), but of course, only if the proposed services worth it! Figue 6: profile definition The way customers will access the service must also be taken into account. Mobile terminals like PDAs or hybrid telephones (GSM/PDA) are often not much practical in terms of interface and typing an URL is quite complicated… That’s why, within the framework of this project, the Appear Networks context aware service discovery platform was used. This service discovery platform can be linked with the Ekahau location engine in order to, through a single client in the user terminal, implement both the service discovery mechanism and the LBS client. When a customer enters the station, the client in the user terminal automatically contacts the platform, which identifies it and push it the list and icons of available services. Figure 7 shows the different steps in that process. 1 2 4 3 5 6 Figure 7: service discovery seen by the customer In (1) the client in the terminal is launched. The interface remains empty if no service is available (2). When the customer enters the station (3) a message appears “You’re entering the Hall of the Station”. And 3 services are automatically displayed (4). A “category menu” allows the user to pass from one category to another (5) in order to access to another set of services (6). Figure 8 shows how the services will change dynamically when the user passes from one zone of the station to another one. Figure 8: dynamic “zone-dependent” adaptation of available services Here bellow, some of the developed context-aware mobile applications studied in the project are presented. In Figure 9, the welcome service is presented. The user receives a personalized welcome message (if the user passes many times by the station in the same day, this message is sent just the first one) and has the possibility, by a single click, to display the next train schedule for his preferred destination from this station. Figure 9: Welcome message Figure 10 shows one of the “alert” services dedicated to disabled people who want to be helped by a station employee in order to get on the train. When the alert icon is clicked, the user receives a message “your call and position are being set to station welcome employees”. At Welcome desk, SNCF employees will see an alert message displaying the position of the person having sent the alarm. Figure 10: Alert message Figure 11 shows a commercial service which says to the user “did you know there’s a Monoprix supermarket in this station?” and displays the position of the user and the position of the supermarket in a single map of the station. Figure 11: a context aware commercial service Figure 12 shows the service/boutique research service. The user can choose among a complete list of services/boutiques in the station to display, again on a single map, the place where the service is and his own position. Figure 12: research of service / boutiques in the station Conclusions and perspectives This SNCF research project has experimented in two large Paris railway stations both a calibration-based WiFi location system and a service discovery platform able to provide personalized and context-aware mobile services for customers. WiFi Mesh networking technologies have also been studied in order to evaluate the complexity and cost reduction made possible by these new techniques. Finally, the project has also allowed to study how the use of the location of the user terminal and the use of profiling techniques can allow railway operators like SNCF to personalize and contextualize the information delivered to each customer, so the information will be more intelligent and targeted. Acknowledgements I would like to thank all the persons and teams having collaborated with us within this projet: - at SNCF: SNCF Paris-Nord Station, Paris-Montparnasse Station, DDGE, DSIT-T and IGTL ; - external companies: Appear Networks, Ekahau, Neotilus, NextiraOne. References [1] RAPPORT ETUDE RADIO WIFI GARE MONTPARNASSE v1.0; SNCF Document; Contractor: NextiraOne; Pierre-Yves Lorand, Peine Akiana and Yann Chauvière. 14 January 2008.