Digital Marketplaces Unleashed

by Claudia Linnhoff-Popien

  The permissions by Fraport and infSOFT GmbH, SBB and Faveno GmbH to publish the screenshots are greatfully acknowledged.

  45.1 Introduction

  Large e‐tailers are disrupting the established shop retail business. Therefor these shops try to compensate their decreasing revenue by extending into online business [1]. Even so, shopping centers and malls have increased their market share recently and seem to play a distinct role [2]. To gain this momentum the upcoming possibilities of indoor positioning and indoor navigation may be a preferential way for mall managers (MM) to add value to customers and retailers within the mall.

  This paper gives a comprehensive overview of Indoor Positioning Systems (IPS) and Indoor Navigation Systems (INS) in large malls or similar places (large railway stations, airport terminals, etc.). Starting with an overview of the actual possible technologies the conceptual layer is illustrated by some examples. On the business layer the impact of an INS will be discussed, especially owing to an expected conflict of the stakeholders in a mall. The primary goal for implementing indoor navigation is to generate revenue by guiding the customer and providing shopping information using a smart‐phone app and optimizing this by acquiring customer behavior.

  The second section after this introduction explains briefly the latest most promising technologies to detect the current position of the customer in a mall. The section thereafter shows the possibilities to use the IPS as basis for INS and some options for the user interface. The fourth section will identify the stakeholders of IPS and INS. An intellectual stakeholder analysis reveals the different interests and possible benefits of the stakeholders from such an infrastructure. It identifies similar but also competing aims and interests of the stakeholders. It also shows the outstanding role of the manager of a mall. In this paper the MM is synonym for the executive person of the operative mall management to act in the interest of the mall’s owner. He is responsible for providing the technical infrastructure (TI) and thus the data owner of the position data. Due to lack of space, legal and contractual issues like data privacy as well as issues regarding business intelligence systems are not discussed in this paper.

  The last section is about the different risks of implementing and operating an IPS/INS and how to mitigate them. At the end a conclusion will summarize the results and give a short outlook for the future.

  45.2 Indoor Positioning Systems as Technology Enabler

  Outdoor positioning systems rely on Global Navigation Satellite Systems and are mature using e. g. the Geographical Positioning System (GPS). They are a commodity and are used not only in cars and ships but also by byciclists, joggers and hikers. When approaching semi‐outdoor or indoor environments the GPS‐satellite signals become unreliable or unavailable due to reflection and/or absorption by the building. Thinking about large buildings like (sport‑)stadiums, airports, railway stations, shopping malls or campuses with multiple buildings (fairs, universities, museums, etc.) there exists a desire to extend the outdoor navigation to indoor seamlessly. Gartner positions IPS in his technology hype cycle as “climbing of the slope of enlightening” and to reach “the platform of productivity” within the next two years [3]. Since smartphones are becoming more and more ubiquitous they are the perfect end user device (UD).

  The aim of IPS is to determine the whereabouts of a pedestrian in a mall by using sensors of the UD and a terrestrial, inertial TI that is suitable to be installed in a mall. Most modern smartphones have the following sensors integrated by default: camera, microphone, accelerometer, gyroscope, WLAN, Bluetooth, GPS, GSM and a 3D‐magnetic sensor. There are many publications about using all these sensors and the related physics for indoor positioning (light, radio waves, sound waves, magnetic fields and mechanical parameters) [4, 5]. However at the moment there are two sensors to favor for immediate practical use in malls namely WLAN and Bluetooth. Both of the technologies are also implemented in the UD as a transmitter which enables forward and reverse signal detection. Beacons are small transmitters of Bluetooth signals that frequently send a customable ID with a defined period and a defined power. Beacons use the Bluetooth Low Energy Protocol (released in 2009) that is an extension of Bluetooth. Since WLAN and Bluetooth work in the same frequency band they might interfere. Due to the frequency multiplex technology the interference can be minimized by skilled selection of the WLAN channels. The beacons may be installed as stand alone. Some types are manageable by beacon controllers or by using them as a mesh net. Controllers are still expensive and need LAN‐connection and power‐connection which leads to significant additional costs compared to the beacons themselves. The mesh net approach has often been the disadvantage of a reduced feature set and leads to addition power consumption. Since most of the beacons use batteries that have to be changed after a time (life‐time depends on radio power and transmission rate and the battery size and ranges between a couple of weeks and few years). The management facility also allows to set all parameters, to update the firmware and to detect defective or lost beacons or identify weak battery status for replacement demand.

  Since the position and the physics of the TI is well known there are different possibilities to determine the unknown position of the UD. There are mainly two dimensions of the IPS issue. The first one distinguishes if the UD is the transmitter and the TI determines absolute or relative physical parameters or if the TI is the transmitter and the UD has to measure the physical parameters. The second dimension is the mathematical way of how to calculate the UD’s position. There are mainly three possibilities. First is the trilateration that needs at least three distances that are determined either by the signal amplitude and a simple propagation model of the radio waves or by using the time of arrival or time difference of arrival. The second possibility is the triangulation that measures the angle of arrival by using amplitude or phase differences of antenna arrays. The third possibility is to model the radio topology. This can be done by empirical investigation of the signal strength by grid measurements and/or by using a parametrical radio wave propagation model. By having the radio map and the indoor map joined it is possible to compare a received signal strength (RSS) by the UD with the offline determined radio map of the related area in real time [6]. The relative RSS will then be correlated with the most probable and reasonable position of the UD. This method is called fingerprinting or received signal strength identification [7]. When changing the TI or the building the radio map needs to be updated which is an additional effort. Whereas parametrical models do not need the cost to establish the empirical radio map, since they are not as accurate. On the other hand using an empirical radio map leads to better spatial accuracy but needs much more storage and computing time on the UD and increases the power consumption. For further reading about IPS see [5] and references therein.

  45.2.1 Wireless LAN for Indoor Positioning

  Wireless LAN seems to be a good possibility to implement IPS since WLAN is often already installed. However it has some weaknesses to be mentioned. Apple’s iOS does not allow to detect the amplitude of the received WLAN. So the use of UD as the receiver is not suitable. The use of the UD as sender is suitable and commercial solutions of different vendors are available in the market. However this solution is cost intensive since you need to have many access points. For accurate position at any point in time the receiver should have at least 4 conventional access points. Furthermore WLAN‐IPS is only possible for connected UD. The standard period of the access‐point‐request by the UD is about 2 min in the case that the UDs are not connected to the WLAN. In this case it is only possible to determine the position every 2 min which is inappropriate.

  However practical experience shows that WLAN‐IPS have good accuracy (see e. g. [8]). They are suitable for semi‐indoor environments and large halls like terminals where few access points may cover a large area.

  45.2.2 Beacons for Indoor Positioning

  Recently more and more malls have started to implement beacon‐based indoor‐positioning‐infrastr
ucture. A thorough planning of the position of the beacons is necessary depending on the requirements. When only push notifications are required it is easy to implement single beacons at the required points of interest (POI) and to use a simple app to receive notifications or to display a website. Sometimes it is necessary to detect the transition between areas (geo‐fencing) e. g. to detect if the customer is in front of or already in the shop. In this case the number of necessary beacons doubles. The change of floors by escalators seems to be very difficult due to the metal and the interference of beacons located in different floors. For indoor navigation a dense distribution of beacons is necessary to get the required accuracy of the position to be able to navigate the customer through the building. In this case the number of beacons increases dramatically. The issue is that in many cases it is not easy to put beacons in place due to optical or mechanical issues. The large height of a hall is also disadvantageous. As a rule of thumb the distance between beacons should be double the spatial resolution requirements. However depending on the topological circumstances this varies strongly. Best practice is to start with a standard distribution and to generate an empirical “heat map” that shows the special resolution and to add more beacons where necessary.

  45.2.3 Improving Indoor Positioning by Sensor Fusion

  Empirical measurements show that the radio level of beacons fluctuates significantly. This can be due to moving people or objects, or modifications of buildings. The RSS also depends on the specific UD and the spatial orientation of the device because of the anisotropy of the receiving antenna. The relative position of the user and the UD influences the RSS as well. These fluctuations and uncertainties can be reduced by mathematical methods. The most prominent and efficient ones are Bayesian statistical analysis and Kalman filtering.

  In practice it turns out that due to the topology of the mall the different technologies differ in implementation costs as well as in accuracy of the error of the detected position of the device. To improve accuracy one possibility is to combine the detected position of two different IPS (mostly WLAN and Beacon‐based). Different holistic algorithms exist to join the position results of different IPS to decrease the deviation error of the position. This is called Indoor Positioning by Sensor Fusion (IPSF). The factors of the conjunction metrics may depend on topology, time, possible error factors (like moving people or trolleys etc.), the device type or signal‐to‐noise ratio. The use of IPSF is not trivial because it may lead to a jitter of the position. This is not caused by the IPS but by artefacts of the IPSF algorithm, so intense testing is required. However if the coverage of WLAN and/or beacons’ signals are not comprehensive and the user maybe switched off either Bluetooth or WLAN on the UD, IPSF is the only chance to achieve a comprehensive IPS/INS over the complete area. IPSF is also necessary when seamless navigation between outdoor semi‐indoor and indoor navigation is required.

  45.3 Indoor Navigation Systems as Business Enabler

  This chapter introduces the “conceptual layer” based on an IPS as “infrastructure layer” to implement an INS. An INS gives the possibility to guide the customer dynamically through the building which adds the first value to the customer. Second it enables the integration of business applications to provide location‐based services and to navigate the customer to required locations and enables business opportunities.

  Assuming an IPS‐System is providing an accurate and precise estimate of the actual position of the customer, it is possible to assign this position to distinct areas or to form routes and trajectories. Trajectories are directed graphs that contain digitized positions parameterized by time. By forming conceptual trajectories (ways and stops) and attaching geographical content and annotating semantic information it is possible to compute a semantic behavior of the consumer [9]. By aggregation of the trajectories of multiple customers, it is possible to learn about customer behavior and to group them into so called personas. This enables marketing, offerings and business opportunities.

  45.3.1 Geo‐Fencing Using Indoor Positioning Systems

  A geo‐fence is a virtual perimeter for a real‐world geographic area. The use of a geo‐fence is called geo‐fencing. One example of usage involves a location‐aware device of a user entering or exiting a geo‐fence. This activity could trigger a push message to the device’s user as well as a message to the geo‐fence operator [10].

  Geo‐fencing in INS is to locate the customer in a specific area e. g. a specific shop or a waiting area or service area etc. If the exact position is not a requirement geo‐fencing is easier than to determine the exact point of position, e. g. in airports it may be sufficient to detect if a flights passengers are already waiting in the waiting area of the flight gate. This can easily be solved by implementing a single beacon with a simple geo‐fencing algorithm at each gate. To provide simple location based services this solution is also suitable. This may be a good cost effective starting point for implementing an IPS‐infrastructure even if it does not allow to guide the customer through the mall.

  45.3.2 Navigating by Indicating the Direction

  One of the biggest issues of the customer is to find a POI in a large building environment. By calculating a route to the requested POI, it is possible to display the direction to walk on an indoor map (Fig. 45.1 right). The indoor map is needed as pictures in different sizes and resolutions as well as in a parametrized way to allocate the position onto the picture before displaying it on the UD. However you have two main issues with this approach. The customer expects to have the map in the direction of his front view. Therefor the UD needs to know the direction of the cardinal points. In buildings there are numbers of disturbances of the geomagnetic field so that this can’t be detected by the UD’s magnetic sensor in a constant reliable manner. The other issue is that the route has taken the non‐trivial geographical topology into account. However since the assumption of the accurate and precise estimate of the position is still in dispute, this way of INS needs to be evaluated and tested properly at the moment.

  45.3.3 Navigating by Following Virtual Walkways

  A possibility to overcome the issues of the navigation by indicating the direction is to create virtual walkways like airways. The position of the IPS is then projected onto a position of the next walkway (Screenshots of different examples of Indoor‐Navigation apps: Fig. 45.1). The routing is then along the walkway. The overall route is easy to calculate by having a repository of the routes in the mall and by using the Dijkstra Algorithm [11] to find the best route. This is a proven technology and also used in car navigation systems although the streets are real there. This solution therefor is expected to have a high user acceptance.

  Fig. 45.1 Left Frankfurt Airport app, the starting point was detected using WLAN‐IPS, middle SBB (Suisse Railway) Zurich railroad station using iBeacon IPS, right Prototype of INS using i‐Beacon IPS, map oriented in viewing direction

  Using this approach you may get trouble with the calculation of the route. In many cases it is not likely to indicate the shortest and/or fastest way to the customer. Either you have issues in your mall like construction areas, moving stairways out of order or you have decided to take the elevators or stairways. If you have dynamic routing of customers (e. g. in airports) the routing of your INS needs to have an interface to that system to dynamically change the routing accordingly. Otherwise the customer gets different directions from the static signs, from the dynamic signature and from the mobile app. In this case the customer will become confused and may de‐install the app. Dynamic routing is also useful for marketing or revenue reasons.

  45.3.4 Guiding by Description

  Routing by following virtual walkways may not be accepted by people who have trouble with spatial imagination and abstraction. It may also lack of usability for certain consumer profiles (e. g. elder people). For these cases it may be better to visualize the route by one or more overview m
aps and to provide dynamically textual information about the route like: “follow the hallway for 30 m”, “now, when you pass the service office on the right side, turn left”, “walk 20 m straight” “use the escalator up …” (Fig. 45.1 left, middle). This leads to the necessity of having a repository of route segments in place, their routing messages, together with (changing) POIs and an engine to dynamically compile routes together with route‐descriptions. By using route descriptions it is also possible to use text to speech software to enable blind people to navigate [12]. The best solution is to combine the virtual walkways with semantic annotations. Systems using virtual reality are under way but still not regarded as a solution.

  45.4 How to Add Value to the Stakeholders

  In a mall you have at least 3 stakeholders. These are consumers, shops (and chain stores) and the MM. In railway stations and airports you may have the so‐called “meeters and greeters”. These “customers” are summarized by consumers since they are potential consumers in most cases. In this section the aims and requirements for an INS by the stakeholder from the business point of view are summarized. It reveals that some requirements are similar but some of them are conflicting.