iMAS: an Intelligent Mobile Advertising System: Development and Implementation
Evans, C. Thomas, A., Moore, P.
Pavlemko, O.
Faculty of Technology, Engineering & the Environment Birmingham City University Birmingham, UK
[email protected]
Faculty of Technology, Engineering & the Environment Birmingham City University Birmingham, UK
[email protected]
Abstract—This Rapid expansion of wireless technologies has provided a platform to support intelligent systems in the domain of mobile marketing. Utilizing Location Based Services and Global Navigational Satellite Systems provides the capability for transport of real-time, scheduled, location-based advertising to individuals and business. This paper introduces location-based marketing and the related novel intelligent mobile advertising system (iMAS). Following and overview of the location technologies the iMAS prototype is presented. The evaluation is discussed with a discussion around the testing strategy. A discussion with results is provided; the paper closing with conclusions and open research questions.
and implementation is discussed. The testing and results are presented. The paper concludes with open research questions. II.
The intention of iMAS [6] is to provide location-based context dependent advertising on a personalized basis implemented using intelligent context [3, 6]. Therefore, appropriate technologies requiring consideration are threefold: accurate location determination, context-aware advertising and consumer detection. Accuracy of location determines the quality of contextrelated advertising for iMAS and can make the difference between provision of relevant, or irrelevant, content. Accuracy levels of location technologies can be classified as follows; ignoring the more sophisticated methods unavailable to most consumers, such as differential satellite-location with post processing utilizing: Global: Utilization of satellite signals for location, such as GPS, with widespread coverage. More sophisticated methods unavailable to most consumers, such as differential satellite-location with post processing are ignored here, but widely adopted augmented satellite location techniques (e.g. Assisted-GPS) are applicable. Wide-area: Utilization of wide-coverage radio signals for location, such as mobile phone cell signal strength and round-trip timing techniques. Local: Utilization of signal strengths from radio sources located in close proximity to advertising contexts, such as WiFi, Bluetooth, ZigBee and suchlike. This could involve positioning but could just as simply provide only proximity data. Space restricts a detailed discussion on these topics however for a detailed discussion on global, wire-area, and local location technologies see [7, 8, 10, 11, 12, 13, 14, 15, 16].
Keywords-intelligent context-aware systems; location-based services, inteligent mqarketing, personalization
I.
TECHNOLOGY REVIEW
INTRODUCTION
Mobile marketing is a relatively new field of study, which over its short historical period of evolvement has proved to be effective and profitable. Majority of success can be associated with rapid development of enabler technologies such as positioning systems GPS, GLONASS, COMPASS and wireless standards of IEEE 802.11 family, 3G, Bluetooth. Now this type of advertising competes with traditional media channels: television, radio and press. Advertisements are becoming increasingly personalised and context-aware to fulfill customer’s identified needs; this is natural, because “Retail has always fundamentally been a local business” [1]. Location-based advertisement not only helps to improve the sales in a particular area, it can also improve brand awareness among potential customers [2]. Location-tracking functionalities were introduced in Japan in 2001 and since then triggered the era of LocationBased Services (LBS) and advertisement [4]. Large companies such as McDonald’s, Coca-Cola, BMW and Nike are investing in proximity based intelligent mobile advertising [5], which is effectively location-aware advertising with ability to broadcast messages via wireless networks (Bluetooth, Wi-Fi, GSM Cell broadcast) or Internet connection using GPS coordinates. This paper addresses the novel location-based mobile advertising system (iMAS) which employs intelligent context to target advertisements to individuals and business based of their context [3]. The paper is structured as follows: following an overview of the iMAS system, development
III.
THE IMAS FRAMEWORK
The iMAS system (Evans et al, 2012) flow of data streams between the moving vehicles must be maintained through the iMAS framework. The design problem central to the functionality of iMAS has been implemented as a proof of concept model capturing and transmitting a range of information relevant to a particular advertising media stream. 1
The intent is to provide a platform for additional mobile units to interact synchronously in real-time. The problem domain is outlined in Figure 1.
Consumer
iMAS uses vehicle tracking of synchronous information to keep the system updated with the appropriate advertisement, using the scheduling algorithm as the information controller. Coupling static data feeds with realtime data using a location based algorithm is possible but with interoperability between multiple technological tools and techniques can sometimes be counter-productive. An example model is found in Figure 2 and this is indicative of the prototype operationally.
Scheduling, vehicular location, time- frame slot, vehicular speed...
iMAS Report
Mobile Target Advert to consumer
Report Updates
Device
Internet Shop A
iMAS Scheduler
GeoLocation of vMAS
Shop B
iMAS
iMAS Interface v MAS Geolocation
Dis
Scheduler Server
pe
rsi o
n
Shop C
Shop D
Petrol Station Shop E
Figure 1. iMAS Design Problem
Shop F
The pace of change has created a number of innovative methods to capture real-time data. Gathering data for specific marketing campaigns could be seen as a step forward to context-aware personalization. Personalization, based on pattern behavior, is one of the techniques implemented with Data Warehousing and data marts.
Figure 2. Example iMAS Vehicular Prototype
The following are a number of technologies reviewed to inform a way forward and to understand the potential difficulties with transit feeds in a real-time context-aware system. These technologies include: AVL – Automatic Vehicle Location TAD – Travel Assistance Device (to assist the Disabled) GPS – Global Positioning System GLONASS – Global Navigation Satellite System – Russian OGC – Open GIS Consortium (geo-spatial services) OGSi – Open Grid Services Infrastructure Location Based Services offer an optimized service providing real-time data. Provide real-time information as a service to attract potential interest among the public by sending information in the format of text, video or interactive messages over mobile networks. e.g., DoCoMo (Japan, India) HTML5 – Has its own “geo-location api” with independent functions, and is compatible with legacy systems which were using earlier api’s and protocols in regards to regulating & isolating a target based on ID’s (ie., Mobile, system or service provider) GeoLocation Api with Google Maps, Bing Maps and many more Open/Closed source api’s are available in the market based on the type of platform to be implemented. Usage of Java Scripts, CSS3, XML, HTML and other web standards needs to be kept in mind while investigating the iMAS project.
A.
iMAS and Concept Technology Already, a number of technologies have investigated approaches that are platform independent, these provide a way forward. Designing and implementing an architecture that is adaptive to the changing context that identifies the geo-location of the vehicular device is central to the iMAS concept; to this end a number of technologies and tools have been reviewed to understand the most practical solution to this problem domain. General Transit Feed Specification – “The GTFS transit feed specification defines a common format for public transportation schedules and associated geographic information” (code.google.com). As a concept this opens up a number of possibilities for developing a prototype application architecture. However, Google expects a number of schemas to be followed: Data Format – GTFS has its own standardized format and one needs to follow the same in order to attain the set goal. Data provided within each data file has to be comma separated otherwise it won’t work. Naming convention of data files according to GTFS requirements might not suffice the data being inserted per our requirement. GTFS provides developers with a static layout of the testing area using Google maps and then based on the transit feed data the use of visualization provides a way to structure the content graphically. The main issue surrounding this is real-time synchronization between the GTFS layout with the vehicular transit data feed.
IV.
THE EVALUATIO
The implementation of the iMAS system has involved evaluation and testing in a ‘real-world’ environment. The selected environment is Birmingham city center in the UK.
2
The following sections present an overview of the testing with results.
such as longitude and latitude of every point of interest. In total, there was a list of 41 businesses (see figure 5) on both routes. The next step was finding relevant images and uploading them onto cloud storage (DropBox).
A. The Testing Strategy The testing took place in an actual live environment (Birmingham city center, UK) with the objective of evaluating and verifying the performance iMAS system. In total, the testing took place over three days during which setting and testing the system outside of usual test environment took place. T The commissioning client and the developers have agreed upon location of live tests on one of the weekly meetings arranged. Street (High Street) within a busy shopping area selected for the live tests (see Figure 3. It was called “Route A”, whereas “Route B” was picked on developer’s initiative to confirm the results of the evaluation and testing process (see Figure4).
Figure 5. Process of testing in live environment
2)
Day 2
On the day two a number of tests were conducted based on the identified points of interest. The weather conditions varied between day 1 and day 2. This was identified ad relevant and significant as: (1) when the weather was sunny the GPS signal was found much faster with approximately a minute difference between sunny and cloudy conditions, and (2) where cloudy conditions prevailed, such conditions affected measurements and the system was performing with minor errors on max relative accuracy setting. The information was updated dynamically by modifying latitude and longitude values in iMAS database. Three circles of the routes were made to ensure advertisements are shown correctly. The ‘eye detection’ feature of iMAS prototype was also tested under various conditions including locations in shade and illuminated areas. In both cases the eyes were detected with high success rate, which proved OpenCV library being a great tool for such task.
Figure 3. Test Route A (High Street, Birmingham City Centre)
Figure 4. Test Route B (New Street, Birmingham City Centre)
3)
Day 3 Another further day of testing was necessary to evaluate the system and ensure predictable results given a set of coordinates that do not change over the time. The system performed as documented without significant issues. Figure 4
1) Day 1 The relevant businesses’ (points of interests) were identified on High Street and New Street that would be used for testing purposes. Appropriate measurements were taken
3
shows the process of testing in live environment on High Street. V.
implementing location-based systems utilizing GPS, rather than examples of implementing with help of Cell-ID, WiFi and other technologies. The path of GPS implementation also does not require any extra hardware to be provided by the client other than already specified in Table 3.1. For these reasons GPS was chosen as the technology for positioning determination. Series of observations were conducted as a part of preliminary research for IMAS project. They were carried out in different environments: busy town centre with tall buildings obstructing the view of sky, residential area with clear sky view and indoors. The measurements were taken on a sunny day with clear sky and in cloudy weather to preserve unambiguity. The outcomes of this research as foreseen suggest that the best results were achieved in residential area on a sunny day, whereas the worst were obtained in a building on cloudy (rainy) day. In the first case the average accuracy hit the mark of between 1 meter and 3 meters; under more complicated weather conditions the average accuracy was between 3 meters and 6 meters, in some occasions reaching to 59 meters. Measurements were made with help of Android smartphone using free utility called ‘GPS Test’. A-GPS function was deliberately switched off in phone settings for obtaining results close to those provided by equipment used for the project. The interface of the program is simple yet informative as seen on Figure 6.
DISCUSSION AND RESULTS
The purpose of the iMAS system is to provide intelligent context-aware or “smart” advertisements of local businesses via displaying visual information, including video and graphics through a large digital screen of a moving vehicle. The software requirements fall into two categories: functional and non-functional; every significant criterion was identified and associated with one of these categories during the process of requirement gathering and customer negotiation. However, despite of this, additional changes could be possibly made to the requirements as the project continues. A. Functional Requirements The system should be able to identify the position of a moving vehicle in busy city environment at all weather conditions. The location of the vehicle should be determined with accuracy level of less than 10 meters. Initialisation of the client software should not exceed three minutes until the first advertisement message is shown after its start. Depending on location, the system should display relevant graphical (image or video) advertisements on vehicle screen. The vehicle should be able to change its route at any time without losing advertisement capabilities. iMAS should be able to count number of people starring at vehicle’s screen (not original requirement, was introduced later) B. Non-Functional Requirements Open Source implementation is preferable although not essential. The system should be easy extendable The system should be able to run on the following hardware: please refer to Table xx C. Choice of Positioning Method The range of appropriate technologies for location determination include (individually and used in combination): GPS, GLONASS, WiFi, Bluetooth, Cell-ID. GPS and GLONASS provide the best outdoor accuracy, whereas WiFi and Bluetooth are used mainly for indoor navigation. Although, Cell-ID and similar technologies theoretically cheaper and simpler for implementation, they do not satisfy the needs of iMAS. Depending on the cell size, the accuracy may vary significantly, and even in best case, it reaches around 100-150m mark. In contrary, GPS positioning gives a control of accuracy level: positioning data of high accuracy can be rounded for abstraction purposes. Additionally, the hardware specification given by client has GPS receiver, simplifying implementation of the end product and making this positioning method preferable. Academic and technical literature is also more supportive and information rich when it comes to examples of
Figure 6.
UI of GPS Test tool for Android
D. Accuracy Scope One of the design questions that had to be answered was GPS accuracy scope. As stated previously, the best average results that were obtained ranged from 1 meter to 6 meters under various weather conditions. Sometimes, the GPS signal was weaker resulting in less precise positioning
4
information (up to 59 meters). However iMAS system does not require this level of precision. It is not navigation software, where 1 meter plays a significant role. The vehicle location can be approximate for advertising purposes, because the system will be looking at surrounding area with radius of 5, 10, 50 or 200 meters. More precision in GPS data will limit the geographical area to a small point, making advertising almost impractical e.g. vehicle will advertise a shop or business in 1-2 meters radius, which is irrational, because presumably this business already has some form of a street banner. Less precision in GPS data means that advertising will be carried out on a city level, eliminating the need of location information e.g. it can be done in traditional way. It was decided keeping the accuracy of GPS information in the range between 10 and 200 meters, which is the optimal value.
In the evaluation the system has been shown to be free of critical errors and logical flaws in its design. Version 1.0 Beta of iMAS system can be practically used and in general matches the original product requirements. Although the vast majority of errors were identified before final documented testing iteration has taken place, few issues were found. T In testing, it was revealed that a delay of a second when “seconds looked” counter was reset to zero after showing advertisement for a new business. This error was classified as minor and was not fixed in current version of iMAS prototype, however it has been considered in the development of the system and in future builds. A further issue was found where the iMAS Client was started and the iMAS Web Services were offline. This resulted in a occurrence of an {InvalidOperationException()}. This is due to the client application attempting (and failing) to send data to the web service (which was offline). After surrounding few lines of code with “try/catch” block the issue was partially resolved. Unfortunately, eye detection would not start when there was no connectivity to web services. One could argue it is a minor issue, as eye detection is pointless when advertisements are not show, but it should be corrected in the next version of the iMAS prototype. There was also a significant issue identified where the web services on IIS server was installed with a different port number, the client application would not resolve the location of web services and therefore work offline. This is the issue that has to be resolved.
E. Scheduling procedures Scheduling is required to address conflicts in the iMAS system when a conflict of interests occur. For example, where one or more clients (businesses contract to use the iMAS system wish to display advertisement simultaneously in a specific location the iMAS system implements 2 steps to manage such situations. handles such situations. They are simple yet effective and may be extended or enhanced in the full version of the program. Initially, when a list of results is obtained (mappings between places and geographical location) the scheduler checks whether this date is “active”. For instance, if today’s date is 28/08/2012 and the advertisement is active between 25/08/2012 and 30/08/2012, it will be selected for further processing. Otherwise, if it is too early or too late to show an advertisement, it will be ignored. Note, not only the dates are taken into account, but time too.
G. Conclusions The evaluation has been exhaustive and has involved both laboratory testing and testing in a ‘real-world’ environment. Notwithstanding the issues identified in the testing process the iMAS system has bben shown to be generally robust and usable in a ‘real-world’ location-based marketing application. All issues were documented and in a number of cases eliminated. Overall, the system has proven to be free of critical errors and logical flaws in its design. Version 1.0 Beta of iMAS system can be practically used and in general matches the original product requirements.
F. Results In the testing and evaluation (as discussed section IV) exposed testing strategy and procedures to the reader. Testing was accomplished in accordance with the preplanned testing strategy. A combination of white box and black box techniques were employed. The results derived from the testing conducted in day 3 (see Figure 4) when take with the results from the testing conducted in days 1 and 2 support the conclusion that the solution (the iMAS system) functions as intended including in situations where when direction on the route was changed rapidly or information was updated in the database in real time. Going off the route resulted in displaying default advertisement image. The relative accuracy figure was also changed to enlarge the radius of coverage, e.g. on minimal accuracy the device with iMAS system running would still advertise businesses in 100-150 meters radius (on a different street nearby). The prototype was also tested in university premises. In some occasions it would not find GPS signal indoors due to the reflections caused by metal constructions. The same behaviour is expected in road tunnels, which is typical for GPS technology in a whole.
VI.
CONCLUSION
This paper has introduced the novel intelligent mobile advertising system (iMAS). An overview of the available location technologies has been presented along with the iMAS framework and concept technologies. An evaluation of the iMAS system has been provided with a discussion, conclusion and an overview of the results derived from the testing. Location-based marketing is gaining traction based on the desire to enable effective personalization and the targeting of services to individuals based on expressed preferences and systemic constraints. The iMAS system implemented using intelligent decision-centric context middleware provides an effective basis upon which personalization in service provision can be achieved in a commercial location-based marketing system
5
There remain challenges in the implementation of location-based systems generally which also relate to the iMAS system principal among which is the ability to effectively profile individuals. Such challenges relate less to the technical specifications (this aspect of the iMAS system has been resolved in the system’s development) than to the effective identification and codification of individuals’ preferences and the motivation for such preferences influenced by emotion, mood, and environmental considerations. Addressing these advanced concepts is beyond the scope of this paper however investigations form the basis for ongoing research. Realizing these concepts falls within the realm of artificial intelligence and is an open research question.
[7]
[8]
[9]
[10]
REFERENCE [1] [2] [3]
[4]
[5]
[6]
Martinez, J. (2011) Marketers add location-based strategies to digital arsenal. DM News, Issue December 2011, p.4 Emerson, B. J. (2011) Using Location-Based Services to Get Customers. Franchising World, Vol. 43 Issue 7, p.9 Moore. P., & Pham. Hai. V. (2012). Intelligent Context with Decision Support under Uncertainty. Second International Workshop on Intelligent Context-Aware Systems (ICAS 2012), In Proc of The 6th International Conference on Complex, Intelligent, and Software Intensive Systems (CISIS 2012), 4-6 July 2012, Palermo, Italy, pp 977-982. Dhar, S. and Varshney, U. (2011) Challenges and Business Models for Mobile Location-Based Services and Advertising. Communications of ACM. Vol. 54, No. 5, pp 121-129 O’Reily, P. and Duane, A. (2010) Smart Mobile Media Services: Consumer Intention Model. Proceedings of The 8th International Conference on Advances in Mobile Computing & Multimedia (MoMM): Paris, ACM pp.188-194. Evans, C., Moore, P., & Thomas, A.M. (2012). An Intelligent Mobile Advertising System (iMAS): Location-Based Advertising to Individuals and Business. Second International Workshop on Intelligent Context-Aware Systems (ICAS 2012), In Proc of The 6th International Conference on Complex, Intelligent, and Software
[11]
[12]
[13]
[14]
[15]
[16]
6
Intensive Systems (CISIS 2012), 4-6 July 2012, Palermo, Italy, pp 959-964. Yeh, S., Hsu, W., Su, M., Chen, C. and Liu, K. (2009) A Study on Outdoor Positioning Technology Using GPS and WiFi Networks. Proceedings of the 2009 IEEE International Conference on Networking, Sensing and Control: Okayama, Japan, March 26-29, 2009. IEEE, pp. 597-601 Soubielle, J., Fijalkow, I., Duvaut, P. and Bibaut, A.(2002). GPS positioning in a multipath environment. IEEE Transactions on Signal Processing. Vol. 50 Issue 1, pp 141-150 Gao, G. X. , Heng, L. Walter, T. and Enge, P. (2011). Breaking the ice: Navigating in the Arctic. Proceedings of the 24th International Technical Meeting of the Satellite Division of the Institute of Navigation (IONGNSS 2011), Portland, OR, September 2011, pp. 3767–3772. Weyn, M. and Schrooyen, F. (2008) A Wi-Fi Assisted GPS Positioning Concept. The Third European Conference on the Use of Modern Information and Communication Technologies, Gent, Belgium. Spiekermann, S. (2004) General Aspects of Location-Based Services. In: Schiller, J., and Voisard, A. (eds.) Location-Based Services. Morgan Kaufman, p.12 Borkowski, J., Niemela, J. and Lempiainen, J. (2004). Performance of Cell ID+RTT Hybrid Positioning Method for UMTS Radio Networks. Proceedings of 5th European Wireless Conference Rashid, O., Coulton, P. and Edwards, R. (2005) Implementing Location Based Information/Advertising for Existing Mobile Phone Users in Indoor/Urban Environments. Proceedings of the International Conference on Mobile Business. ACM, pp. 377-383 Deblauwe, N. and Ruppel, P. (2007). Combining GPS and GSM CellID positioning for Proactive Location-based Services. Proceedings of Fourth Annual International Conference on Mobile and Ubiquitous Systems: Networking & Services, 2007: Brussel, Brussels. pp. 1-7 Rodriguez, M., Pece, J.P. and Escudero, C.J. (2005) In-building location using Bluetooth. Proceedings of the International Workshop on Wireless Ad Hoc Networks. Weyn, M. and Schrooyen, F. (2008) A Wi-Fi Assisted GPS Positioning Concept. The Third European Conference on the Use of Modern Information and Communication Technologies, Gent, Belgium.
