Context-aware services based on spatio-temporal ...

Behaviour & Information Technology

ISSN: 0144-929X (Print) 1362-3001 (Online) Journal homepage: http://tandfonline.com/loi/tbit20

Context-aware services based on spatio-temporal zoning and crowdsourcing Akhlaq Ahmad, Md. Abdur Rahman, Mohamed Ridza Wahiddin, Faizan Ur Rehman, Abdelmajid Khelil & Ahmed Lbath To cite this article: Akhlaq Ahmad, Md. Abdur Rahman, Mohamed Ridza Wahiddin, Faizan Ur Rehman, Abdelmajid Khelil & Ahmed Lbath (2018): Context-aware services based on spatio-temporal zoning and crowdsourcing, Behaviour & Information Technology, DOI: 10.1080/0144929X.2018.1476586 To link to this article: https://doi.org/10.1080/0144929X.2018.1476586

Published online: 04 Jun 2018.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at http://tandfonline.com/action/journalInformation?journalCode=tbit20

BEHAVIOUR & INFORMATION TECHNOLOGY https://doi.org/10.1080/0144929X.2018.1476586

Context-aware services based on spatio-temporal zoning and crowdsourcing Akhlaq Ahmad a,b, Md. Abdur Rahman Abdelmajid Khelil f and Ahmed Lbath

c

*, Mohamed Ridza Wahiddin

b

, Faizan Ur Rehman

d,e

,

e

a

College of Engineering, Umm Al Qura University, Makkah, Saudi Arabia; bKICT, International Islamic University, Kuala Lumpur, Malaysia; College of Computer and Information System, Umm Al Qura University, Makkah, Saudi Arabia; dScience and Technology Unit, Umm Al Qura University, Makkah, Saudi Arabia; eDepartment of Computer Science, LIG, University of Grenoble Alpes, France; fTU Darmstadt, Darmstadt, Germany c

ABSTRACT

ARTICLE HISTORY

Crowdsourcing offers great opportunities to recognise user context and prescribe relevant services for both offline and real-time activities. In this work, we present a zoning model that leverages spatio-temporal dimensions and then employs different contexts to recommend necessary customised services. The context model takes into consideration three context sets: fully restricted, fully unrestricted and semi-restricted with respect to both spatial and temporal dimensions. As a proof of concept, we apply this zoning model in a scenario where a very large crowd get together to perform spatio-temporal activities. The user context of the heterogeneous crowd is captured using the carried smartphones, i.e. via crowdsourcing. Depending on the context sets and zone, the system can recommend a set of services to each user. The system has been deployed since 2014 to support the spatio-temporal activities of a very large crowd. We present our implementation details and the user feedback, which is very encouraging.

Received 3 July 2016 Accepted 8 May 2018

1. Introduction Crowdsourcing is a generic approach for formulating and thereby solving distributed problems by the direct, indirect or event-based participation of users themselves (Chatzimilioudis et al. 2012; Zambonelli 2011). Social media, Wikipedia, Amazon Mechanical Turk (AMT) and E2 (Everything2, a web-based interlinked community) are a few examples of crowdsourcing technologies (Pan and Blevis 2011). Furthermore, fundraising, electronic voting, micro-works etc. commonly use crowdsourcing applications (Chatzimilioudis et al. 2012). Analysing crowdsourced data from a very large crowd may deliver useful and new insights regarding crowd behaviour (Xu et al. 2016). Despite the remarkable advances in crowdsourced sensing, capturing the context of each individual within a large crowd is still a challenging task since it requires modelling the crowd by considering components and attributes related to both individuals and groups and their preferences (Hara et al. 2014; Murturi, Kantarci, and Oktug 2015). Hajj is an example of an extremely large and heterogeneous crowd. More than 6 million pilgrims from all over the world accumulate in Makkah city to perform a series of spatio-temporal rituals in about one week’s time duration (Ahmad et al. 2014b; Ahmad et al.

Crowdsourcing; spatiotemporal zones; contextaware services; system usability; Hajj and Umrah

2015b). This crowd is highly diversified in terms of different cultures, languages, literacy rate, ages, etc., which possesses managerial challenges for the local authorities in providing suitable services. Different ministries of Saudi Arabia are actively serving pilgrims and always struggle for improvement and modern solutions. As in a Hajj crowd, inter- and intra-user context is different, it is necessary to capture each pilgrim’s context at any given time and location and provide the appropriate set of services. Supplemental challenges for the Hajj crowd include the restrictive spatio-temporal mobility. For example, on the 9th day of Zulhijjah (12th month of the Islamic calendar), attendance of all pilgrims in a small place called Arafat is mandatory until after sunset; failing to do so would invalidate their pilgrimage. Other than Arafat, pilgrims have to move within other places such as Mina, Muzdalifah, Jamarat and Haram in a specific order and with certain relaxed temporal constraints. Similarly, Umrah is an event that restricts pilgrims from performing a few of the rituals similar to Hajj, but only in the Haram area, with no temporal restrictions (Ahmad et al. 2014a). During the Hajj event, other than complete ritual guidance, pilgrims need supervision in locating their fellow pilgrims and family members, navigation to their points of interest

CONTACT Akhlaq Ahmad [email protected] College of Engineering, Umm Al Qura University, Makkah, Saudi Arabia * Present Address: Forensic Computing and Cyber Security Department, University of Prince Mugrin, Madinah, Saudi Arabia © 2018 Informa UK Limited, trading as Taylor & Francis Group

KEYWORDS

2

A. AHMAD ET AL.

(POIs) such as their hotels of residence, tents in Mina/ Arafat/Muzdalifah, restaurants, prayer places and timings, weather update, locating shopping malls and currency exchange, to name a few, for their daily needs. Smartphones coupled with high-speed Internet services are useful tools for capturing useful sensory data to understand a user’s context. As an illustration, a user’s position coordinates read by a GPS sensor and their movement speed read by an accelerometer sensor together serve in deducing a user’s context, such as a user is ‘sitting’, ‘walking’, ‘running’, ‘sleeping’ etc., at the considered location (Ahmad et al. 2014a; Rahman et al. 2012). Our current work leverages the work done by certain authors (Ahmad et al. 2014a, 2014b; Ahmad, et al. 2015a; Estrin 2010; Rahman et al. 2012; Yan and Chakraborty 2014; Yurur, Liu, and Moreno 2014), where the authors have proposed the context-capturing techniques. A user’s context is classified into three categories: fully constrained, fully unconstrained and semi-constrained contexts. For example, visiting a bank for a financial transaction would be considered as a fully constrained activity due to the bank’s fixed location and banking hours. Buying grocery items is a fully unconstrained activity because one can buy from any store at any time. Selecting a filling station on the way can be considered as a semi-constrained activity because one can select any station but within the next few minutes before the petrol gets completely consumed. Understanding these three types of contexts, a user can be

offered navigation services and information on the time and distance remaining to reach his destination. Facilitating a Hajj crowd with best possible services first needs an understanding of pilgrims’ individual and collective behaviour. It needs to first model pilgrims’ contexts as per their activities. It is a challenging activity because unlike Facebook and Twitter which take years to build, the Hajj spatial-temporal activities need to be captured in a period of about a week. It requires the collection of data of a particular type and quality to finally extract pilgrims’ contexts and provide a suitable service. In this article, we present the Hajj event as a series of spatio-temporal activities, where pilgrims’ mobility is constrained by some spatial and temporal zones (Figure 1). Each spatio-temporal zone falls within the fully constrained, semi-constrained or fully unconstrained ritual context of a pilgrim. To the best of our knowledge, this is the first work where a massive crowd such as that in the Hajj has been modelled in the context of spatio-temporal zoning. We present a zoning model that elaborates pilgrims’ activities within the spatial zones of Arafat, Mina, Muzdalifah, etc., with time constraints. These zones together with smartphones’ raw sensory data (for example, pilgrims’ position coordinates), users’ generated contents both in social data (during chatting services) and e-forms (for example, traffic updates, registering complaints and updating health forms), captured through smartphones carried by the pilgrims help in construing their context. The zoning algorithm allows

Figure 1. Temporal zones TZ1–TZ5 associated with spatial zones SZ2–SZ6. (Tawaaf, Saee and Jamarat are different rituals that need to be performed by each pilgrim).

BEHAVIOUR & INFORMATION TECHNOLOGY

mapping of the spatio-temporal context of each pilgrim with a subset of our developed incentivised services and provides inter-user and intra-user personalised services. For example, on Arafat day, pilgrims’ attendance is mandatory; however, temporal relaxation to reach there within the defined duration would require interpreting their contexts as semi-constrained contexts. Accordingly, the system recommends needful services, such as reminders to reach Arafat, navigation and obligatory supplications. The zoning algorithm also allows the framework to be in touch with the pilgrims within a zone in cases such as emergency evacuation or other crowdsourcing scenarios. Moreover, the zoning model allows the framework to support crowdsensing applications such as pilgrims take part in sharing incidents of interest with pilgrims of nearby zones or with the city. The remainder of the paper is structured as follows. In Section 2, we present related works by other researchers. In Section 3, we first present a sample Hajj scenario to highlight the possible activities and needed services, and later describe how we associate different Hajj activities with multiple spatial and temporal zones. Furthermore, this section shows user context modelling by taking into consideration different spatio-temporal zones and constraints. This section also explains the high-level architecture of our proposed system, the role of a context manager, our focus on data quality and different data types involved. In Section 4, we detail the implementation of both cloud and crowd-sensing environments. In Section 5, we present the evaluation results of the deployed system in terms of usability, and users’ feedback to streamline our research for potential improvement for future gatherings.

2. Related work To facilitate the large crowd with best possible services, first, there is a need to interpret individuals’ and collective contexts through a specific mechanism. In this section, we have summarised the research work proposing different techniques to crowdsource useful information that can manipulate users’ contexts, their needs and respective solutions. We extend our study with the perspective of the Hajj event, which is an exclusive example of the culturally diverse crowd, where pilgrims are constrained to specific spatio-temporal zones and are in dire need of real-time guidance and services. Crowdsourced data include sensory data, personal data, social data, users’ feedback, e-health forms, ondemand surveys as well as other kinds of user-shared contents (audio, video, text and images). Varieties of sensor-enabled smartphones leverage their usability as exceptional crowdsourcing tools (Campbell et al. 2006;

3

Höller et al. 2014; Miluzzo et al. 2008a; Ziefle and Bay 2005) and specially designed people-centric applications can deliver a handsome amount of information which can be processed to get back to users with multiple real-time and on-demand services (Ahmad et al. 2014a; Alt et al. 2010; Dey, Abowd, and Salber 2001; Miluzzo et al. 2008b). For location-based services such as finding POIs and finding optimised routes for them, map-based solutions have been presented by Abowd et al. (1996), Cheverst et al. (2000), Yan et al. (2009), Alt et al. (2010), Muller et al. (2011), Shah et al. (2011), Korthaus and Dai (2012), Mitchell et al. (2013), Ahmad et al. (2015c) and Rehman et al. (2015), by providing smartphone applications and developed web applications for managing authorities. Mostly, the applications crowdsourced, collected user-generated MM (multimedia) data and after processing made them available for other users. For tracking pilgrims during the Hajj event to better understand crowd behaviour, Rahman et al. (2015) have presented an agent-based crowd simulation tool, Mantoro et al. (2011) have presented the Hajj Locator, a smartphone application, Korthaus and Dai (2012) have presented ‘HetNets’ – crowdsourcing through a heterogeneous network of devices, Nafea, Bhairy, and Zeidan (2014) have presented a Hajj tracking framework and Mitchell et al. (2013), Al-Hashedi et al. (2011) and Khan (2011) have used the RFID technology. Furthermore, Anis and Saeed (2011) presented a multilingual local positioning system for Hajj operations that uses wireless network architecture. Basalamah (2016) introduced Bluetooth low-energy tags for crowd sensing and Mohandes (2008; Mohandes et al. 2011, 2012) used wireless sensor networks and later (in 2015) the NFC technology. For educating and training pilgrims, Hameed (2010) proposed a multilingual Hajj educational model, Alt et al. (2010) added translation services to their ‘SensoRcivico’ platform, Sulaiman et al. (2009) presented HajjQAES, which gives replies to pilgrims for their queries. Fathnan et al. (2010) presented a web-based Hajj simulation software learning system, which allows an effective learning of the Hajj process, and Mulyana and Gunawan (2010) presented an intelligent Hajj crowd simulation system for pilgrims’ training before the actual pilgrimage. Mohd Yasin et al. (2010) proposed a situated learning mechanism and a virtual 3D avatar guide using virtual reality that was subsequently enhanced by Yusoff, Zulkifli, and Faisal Mohamed (2011) and by designing Virtual Hajj (V-Hajj)-a courseware as a supplementary learning material for pilgrims. Ahmad, et al. (2015a, 2015d) presented an ST-diary that serves users with past users’ multimedia guidance for their spatio-

4

A. AHMAD ET AL.

temporal activities. Bigham et al. (2010) presented Viz Wiz, a smartphone application for blind people that translates crowdsourced images to audio guidance. Anis and Saeed (2011) presented a multilingual speech translation system for Hajj operations. Zeki et al. (2012) developed a trilingual mobile dictionary for pilgrims, which uses multimedia contents to translate the services. For pilgrims’ health concerns, Nafea, Bhairy, and Zeidan (2014) presented a health-tracking framework for the Hajj crowd, which uses electronic health records in collaboration with the local health ministry and Mohandes (2015) used an NFC technology-based system to create pilgrims’ profiles which can be used in medical emergencies. Although previous researchers have proposed various models to capture the user context using a smartphone, the zoning model presented in this paper is unique since Hajj poses new dimensions of challenges. For example, the zoning model has to address the number of pilgrims per zone, making an ad-hoc social network of the pilgrims per zone, handling the context of a very large number of pilgrims moving among zones, sending a very large payload of inter-user and intra-user contextual as well as multimedia data to the big data server in realtime, mining events and other relations and provide personalised visualisation of the analysed data. In the next section, we will provide a detailed description of the proposed zoning model and its implementation.

3. System design 3.1. Sample Hajj scenario Ali and Fatima arranged a Hajj trip for their parents Abdullah and Zainab. Since it was their first time to perform Hajj they were curious to know about the Hajj rituals and understand well that failing to perform obligatory rituals will invalidate their pilgrimage. They prioritised knowing more about the rituals and understanding the environment of the holy city of Makkah. The travel agent recommended to them the installation of our App on their smartphones to be used during their Hajj journey and advised them to take full advantage of its multiple features. Abdullah selected the English language interface while Zainab was comfortable with Urdu. Abdullah and Zainab updated Ali and Fatima back in their home country about their arrival in Makkah, by sharing multimedia messages through the inbuilt Hajj Messenger service in our App. The App facilitated them to find out prayer times for Makkah and the Qibla direction as well as the nearest mosques to pray at. They went to Haram to perform Umrah and the App showed them the way to the holy mosque and a

complete ritual guide with recommended supplications. The Tawaf and Saee counter helped them to keep a record of the number of rounds. On the way back to their residence, they used the translation service to explain to the taxi driver about the previously added location through the favourite place feature. They were lucky to find a nearby restaurant for dinner and found hotels that served food of their preference through the POI interface and saved them in their favourite POI list. They also shared these favourites with co-pilgrims who were using the same App. When they needed local currency, the App showed them a path to the nearest currency exchange centre and the latest currency rates. They added some places for shopping in their favourite POIs to visit later. The App allowed Abdullah and Zainab to share the precious moments and memories with their children and other fellow pilgrims via Tweets, images and videos. Once they were in Mina on the first day of Hajj, Abdullah and Zainab were fortunate to save their tent location, which prevented them from getting lost during the remaining days. On Arafat day, remaining inside the Arafat boundary was one of their major concerns; however, thanks to the out-of-boundary service, they managed to remain inside the Arafat boundary and later, on Muzdalifah night and the remaining days in Mina. After stoning the devil on the 10th of Zulhijjah, Abdullah went to sacrifice the animal and got lost, but he took help from the App to find the shortest route to reach his pre-located tent in Mina through the favourite POI interface. They made full use of inbuilt social network services to learn about and share their experiences about crowd issues, traffic updates, accident reports, food quality, misconduct of any Hajj administratives, etc. with the concerned departments and the Makkah city council. The traffic update service really helped them a lot to choose the best time to travel in between the Hajj zones. They made full use of the weather update and Hajj news services to prepare themselves for any emergency. During their stay in Makkah, both Abdullah and Zainab used the Health Form service to remain updated about their health and in case of an emergency, they shared their location through the Emergency SMS service, which helped the health authorities to timely provide them with the desired health services. We derived the necessary requirements from the above scenario and designed the software and hardware model as presented below. 3.2. Spatio-temporal zones During the Hajj activity, pilgrims get together in the city of Makkah and perform a series of Hajj rituals restricted to certain spatial and temporal zones. Therefore, we

BEHAVIOUR & INFORMATION TECHNOLOGY

describe these rituals in terms of spatial and temporal zones separately. Spatial Zones: Figure 1 shows the spatial zones associated with the Hajj and Umrah rituals in a high-level map view within Makkah (Ahmad et al. 2014b). These spatial zones are pre-defined areas, where pilgrims move over the five-day period under certain regulations. All five spatial zones (SZ2–SZ6) are subsets of spatial Zone SZ1 (the Haram boundary). Furthermore, SZ2 (the Holy Mosque) has two sub-zones SZ21 and SZ22, and SZ6 (Jamarat) has three sub-zones SZ61–SZ63. Pilgrims’ mobility within these spatial zones is constrained due to certain ritual regulations and temporal constraints. Temporal Zones: The main event of Hajj has a span of five days (8–12th day of the 12th Arabic month). Based on different activities within these five days, we divide rituals as per their temporal constraints, mainly as TZ1 (termed as a temporal zone for Day 1), and TZ2–TZ5. Figure 1 gives an overview of these temporal zones with respect to their association with spatial zones (SZ2–SZ6), and the arrows between any two zones show the allowed movement of the pilgrims. If we deepen our study, we can further divide these temporal zones with certain regulations to fine-tune the spatio-temporal context. Users’ presence in the different temporal zones (TZ1– TZ5) and sub-zones totally depends on certain ritual regulations. TZ2 is supposed to start early in the morning (after sunrise, e.g. 7:00 am) on Day 2 of Hajj and ends by sunset (e.g. 7:00 pm on that particular day). This phenomenon can be expressed as lb (SZ4 − Day 2)7:00 am , (TZ2 ) , ub (SZ4 − Day 2)7:00 pm . Here lb and ub are the lower and upper bounds of the temporal dimension describing the starting and ending times. According to the ritual’s regulations, users’ presence in spatial zone SZ4 is mandatory for TZ2; however, this constraint could be relaxed by moderating the upper and lower bounds (e.g. difficulty for some old-aged individual to be inside SZ4 within the prescribed range due to heavy traffic). Hence, the modified constraint can be expressed as follows: 7:00 am , lb (TZ2 ) , 7:00 pm, 7:00 pm , ub (TZ2 ) , 7:00 am (sunrise of Day 3). Similarly, other temporal zones may be modified as per constraints faced by users, while keeping the SZ4 spatial constraint mandatory. 3.3. Context modelling Context modelling is an essential element to understand the behaviour of a large crowd individually and

5

collectively, and to serve them with appropriate services (Benerecetti, Bouquet, and Bonifacio 2001). In this work, we adopted the modelling approach as introduced in (Furht and Agarwal 2013), where the context manager combines a mobile phone’s sensory data with data collected by body sensors and user-generated data during social interactions among them. Our proposed model first classifies user contexts as the fully constrained context CFC , semi-constrained context CSC and fully unconstrained context CUC , and then serves users with the best-fit set of available services. Each context type is segregated further into three categories as C = C(l, t, Dt, q)

(1)

where ‘l’ is a spatial component representing the location of the user and ‘t’ and Dt are temporal components representing time and change in time, respectively, to perform a certain activity ‘q’. The context manager (Figure 3) combines this set of contexts (spatial and temporal) to fine-tune users’ contexts. As an example, we consider the case when one pilgrim is observed with respect to the spatial zone SZ4 (Arafat zone) while performing the Hajj rituals. We consider two possibilities, either a pilgrim is inside or outside SZ4 (Figure 2(a,b)), which is explained in detail as follows: (1). Case-I: the pilgrim is outside SZ4

On the 9th day of Zulhijjah, it is obligatory for the pilgrims to the remain inside the Arafat boundary for some time. In case-I we consider the pilgrim is outside SZ4; then we define her 1st primitive spatial context [SC] as C1 = LOCATION is outside the SZ4

[SC].

Other than the 9th Day of Zulhijjah, being inside or outside the Arafat boundary has no impact on the pilgrimage. Therefore, next, we add the temporal context [TC] C2 defined as C2 = DATE is 9th Zulhijjah (Day 2 of Hajj) [TC]. The duration for which pilgrims should attend the Arafat boundary starts from sunrise (say 7:00 am) or maybe noon (12:00 pm) till sunset (say 7:00 pm); failing to do so would invalidate their pilgrimage. The system calculates the value of ‘‘Dt", which is the duration, or it can be different periods defined within the upper and lower bounds as described in Equation (1), that is Dt = ub − lb.

(2)

The value of ‘‘Dt" may vary as per season, summer or winter, and these changes are accommodated by the

6

A. AHMAD ET AL.

Figure 2. (i) A pilgrim is outside Spatial Zone SZ4 and (ii) a pilgrim is inside Spatial Zone SZ4.

relevancy manager (Figure 4). However, due to mobility issues of the large Hajj crowd, there is temporal relaxation that at least a pilgrim must be inside the Arafat boundary for some time in between these time bounds. The next primitive context is again the temporal context given by C3 = TIME is between afternoon and sunset (Day 2) [TC]. By general observation, spatial context C1 seems to be enough to understand that the pilgrim is outside the

Arafat zone; however, other than the 9th Zulhijjah and the time bounds, C1 alone has no importance. Therefore, the system integrates all C1 − C3 to deduce finally that ‘The pilgrim is outside the Spatial Zone SZ4’. In this case, since both spatial and temporal dimensions have to be strictly followed to validate the pilgrimage, the pilgrim context will be considered as ‘fully constrained’, which can be defined by CFC = {C1 , C2 , C3 }.

The system recommends a set of available services ‘S’, based on the pilgrim context given by the set S = {S1 , S2 , S3 , . . .}.

Figure 3. High-level system architecture.

(3)

Figure 4. Sensory, user-generated and social data.

(4)

BEHAVIOUR & INFORMATION TECHNOLOGY

Here, S1 = {S11 , S12 , S13 . . .},

(5a)

S2 = {S21 , S22 , S23 . . .},

(5b)

S3 = {S31 , S32, S33 . . .}.

(5c)

The subsets S11 , S12 and S13 can be fully customised. For example, S11 can be regarded as an SMS service to remind a pilgrim to move towards SZ4 before the upper bound of temporal zone TZ2. Similarly, S12 , S13 could be the second and third reminders in case a pilgrim has not started moving towards SZ4 or is still outside SZ4. S2 = {S21 , S22 , S23 . . .} includes possibly the shortest path recommended with a navigational map, an alternative path in case the route is crowded or another path in case there is an accident or a road is blocked once the user is half-way towards his/her destination. S3 = {S31 , S32, S33 . . .} includes the set of services to update a user regarding estimated time of arrival for three different paths shown to the user. Similarly, we can define S4 as guidance towards a taxi stand, available space in a taxi, fare amount, etc. In case the system acquires knowledge from GPS data that the pilgrim has started to move towards SZ4, a new primitive context adds to C (Equation (3)) given by C4 = CHANGING LOCATION (heading towards TZ4)

7

A user’s new context as per the aforementioned set of primitive contexts is expressed as CFC = {C5 , C6 , C7 }.

(7)

If the context manager declares the user context as C5 , C6 and C7 then ‘The pilgrim is inside the Spatial Zone SZ4’. In this case, the set of services to be offered would be, for example: S = {S5 , S6 , S7 , . . .}.

(8)

To illustrate examples of possible services from the pool of available services, we can define S5 = {S51 , S52 , S53 , S54 , . . .} as the following services: S51 is a reminder to start prayer after 20 minutes, S52 is the second reminder, S53 is the Qibla direction, S54 recommends supplications, etc. As the temporal zone approaches its upper bound, a new temporal primitive context will be added to a user’s context to define a new set of services, for example, S71 can be regarded as a reminder service to start preparing to move to spatial zone SZ5 (Figure 1), S72 is the navigation service to SZ5, S73 deals with traffic updates, estimated arrival time, etc. Similar to Equations (6a, 6b), we can summarise differentiated services as CFC ]i  S = {S53 , S62 , S71 , . . .},

(9a)

CFC ]i  S = {S73 , S63 , S52 , . . .}.

(9b)

Here, ‘i’ represents that the user is inside the spatial zone SZ4.

[SC]. The relevance manager could offer a different set of services to users of a similar context, which might be different by considering their history and/or personal profile. The pool of services for two different users by taking into consideration the inter-user context could be summarised as CFC ]O  S = {S12 , S22 , S23 , . . .},

(6a)

CFC ]O  S = {S13 , S21 , S24 , . . .},

(6b)

where ‘o’ represents that the user is outside SZ4. (2). Case-II: the pilgrim is inside sZ4 For the pilgrims inside the spatial zone SZ4 (Figure 2 (b)), a new set of primitive contexts could be as follows: C5 = DATE is 9th Zulhijjah (day 2 of Hajj)

[TC],

C6 = LOCATION is inside the spatial zone SZ4 [SC], C7 = TIME is any time before the next day starts [TC] .

3.4. System architecture In the following, we present our system architecture and explain the details in the data collection and service provision approaches. (1). High-level crowdsourcing environment Figure 3 shows the high-level system architecture of our proposed system. Thanks to the recent advancement and affordability in the smartphone and ubiquitous Internet availability and speed, being residents of the holy city Makkah, we have observed that majority of the pilgrims within crowds carry smartphones equipped with varieties of built-in sensors and Internet connections. The Context Manager receives sensory data captured by multiple available sensors within users’ smartphones that can record their geographical location, surrounding temperature, humidity, etc. (Furht and Agarwal 2013). It can also record data available through built-in or external sensors that can measure heartbeat, pulse rate, etc., (Ahmad, Afyouni, Murad, Rahman, Ur Rehman, et al.

8

A. AHMAD ET AL.

2015a; Ahmad et al. 2014a). Moreover, the Context Manager documents user-generated data during their social interactions, a survey conducted by managing authorities and users’ health information shared especially on request for monitoring certain epidemic threats. Hence, the Context Manager provides us with both intra-user and inter-user contextual information containing user and event locations, event details, user interactions (historical data), user surroundings, user health and other rich data types of all the pilgrims. The Context Manager forwards the collected data to the Quality and Context-Aware Manager for first filtering data through multiple stages and then refining user contexts. The Quality and Context-Aware Manager is added to maintain data quality through a series of filtering processes based on different contextual dimensions (Ahmad, Afyouni, Murad, Rahman, Ur Rehman, et al. 2015a). The sub-components of this manager are detailed next. The Environmental Manager takes care of the environmental effects (temperature, humidity, etc.) on the collected data that in general can affect data collection from the built-in sensors within smartphones. The Quality Manager executes data validation according to certain prescribed quality levels (e.g. noisy data) within certain threshold limits, which can be changed based on the need. The Relevancy Manager reads data and determines the relevance as per users’ rating for any service used by them and selects authentic data sources. The system is updated by the Rating Manager about integrity and relevance of users as ‘crowdsources’. This is to increase the trust level of the collected data. This component also reconsiders data rejected in case there is an alert to redefine the data limits. The User Profile Manager is the component that decides whether both collected sensory data and user-generated data come from a particular user or from others as well. For example, in the case of a user’s account validation process, a security code is sent during the registration or authentication process, which is managed by the User Profile Manager. The Spatio-Temporal Manager performs spatial and temporal checking over captured data and restricts data sources to allow only relevant data from users’ physical presence within one’s area and time of interest. With the help of the User Profile Manager, it can offer different sets of services for the users with similar spatio-temporal contexts. The Task Manager is related to administration when they need certain feedback or opinions from the crowd. It can be something related to health, emergency and or user survey. The COI (Community of Interest) Manager reads the relationships among members of each COI for data validation as per an assigned task. We expect that different users in a particular COI have similar feedback and it

is accepted or rejected once it is within an acceptable spatial range or not, respectively. Once crowdsourced data are scrutinised by passing through different components of the Quality and Context-Aware Manager, our architecture defines the ultimate user(s) context(s) as fully constrained, semiconstrained or fully unconstrained. These three types of context(s) depend on the type of rituals being performed by the pilgrim(s) in the respective spatio-temporal zone. Accordingly, our architecture offers suitable service(s) selected from a pool of services (rituals guide, navigation to a selected or recommended POI, weather update, news and out of boundary service) in the right place and time. (2). Crowdsourced data classification Depending upon the nature of crowdsourced data, the system classifies them into two main categories: sensory data (generated by smartphone built-in sensors like GPS, accelerometer, proximity sensor, camera, temperature sensor, etc., and/or externally paired sensors) and usergenerated contents (Figure 4), which individually or collectively provide a user’s particular context (M. A. Burke et al. 2006; Ganti, Ye, and Lei 2011; Rahman et al. 2012). Abrupt changes in the accelerometer and GPS data might define the context of a user as ‘running’. Likewise, above a certain threshold value, temperature data collected by temperature sensors may be translated as a ‘certain area is under fire’. If GPS, accelerometer and temperature data are combined, we can deduce that the ‘user is running away from an area under fire’. To validate this situation, similar data are measured for other nearby users and in case of similarity in data values, that spatial zone can be declared as a ‘fire zone’ and can be shared with relevant authorities who can validate and make necessary arrangements to handle this emergency situation. User-Generated Data can be either unidirectional or bidirectional. If the data are generated and shared by users themselves (such as traffic update for any accident or road under construction, complaints about available services for their low standards, social data contents, etc.), they are unidirectional by nature (Rehman et al. 2014). Bidirectional user-generated data originate in scenarios such as a certain request by managing authorities which are followed by users’ replies. For example, if a certain area is under observation for a certain epidemic spread or terrorist attack, users are continuously asked for updates regarding their surroundings and to report any serious threat. Accordingly, alerts can be generated for public awareness to keep them prepared for any

BEHAVIOUR & INFORMATION TECHNOLOGY

emergency. Unidirectional data generated over the social networks can also be observed and users can be contacted to keep updating them in a serious situation. For example, if a certain number of users report a traffic jam, then upon necessary action by the authorities, users can be asked for continuous update/feedback for ongoing events. In the next section, we explain how spatio-temporal zones can be utilised to deduce users’ contexts to offer the right services at the right time.

4. System Implementation In the following, we detail our implementation of the front-end, back-end and service provisioning. 4.1. Front-end application To validate our proposed system for context-aware crowdsourcing services, we have deployed an Android/ iOS-based smartphone application, specifically designed for pilgrims to address their possible needs during their Hajj activities. This application was initially designed for six languages. Accordingly, it was very popular and attracted multiple nationalities (more than 90), which indicates its attractive user interface and high usability rating. Figure 5 gives an overview of the usage of this application during the Hajj 2014 event. Among different services offered [(Authors)], the most popular was adding friends in the Hajj Messenger service, which allowed pilgrims to develop a social network among them (Akhlaq et al. 2014b). 4.2. Pool of services Our implemented system offers a set of important services which will help pilgrims to perform their rituals more efficiently [(Authors)]. Besides the Adding Friends

Figure 5. Hajj and Umrah app, an overview.

9

Service through the Hajj Messenger, the Out of Boundary Service was mostly used as it provided pilgrims with their real-time location when they were enquiring about their positions in certain spatial zones such as Haram, Mina, Muzdalifah and Arafat (SZ1–SZ6) under temporal constraints (TZ1–TZ5) [(Authors)]. Similarly, POI searching service was also very popular as most of the pilgrims come to the pilgrimage at the end of their life and the environment is completely new to them. Its implementation and pilgrims’ behaviour will be explained in the next section. The Lost and Found Service was a service to help pilgrims in finding one another when one is lost. Pilgrims were connected to one another or to their family members through the Hajj Messenger Service (a multimedia chatting environment), locally or back at their homes to share their real-time user experience. The multimedia environment helped pilgrims in lost and found cases, especially when they added multimedia information about their physical locations. The Vehicle and/or Pilgrim Tracking Service was mostly used by group officials coordinating their fellow pilgrims. This service supports tracking vehicles carrying items for pilgrims’ daily needs. Addressing a multilingual crowd was challenging and pilgrims were given a Translation Service, to help them translate their words with the audio-playback facility and ask for necessary assistance from pilgrims of other nationalities or local organisers. For registered users, a Free SMS Service was provided to contact local administrators, fellow pilgrims and/or family members in any emergency cases. The recipients were able to see the location of the sender to proceed for necessary preparation. Emergency Services were for coordinating with the police, fire stations, civil defence and hospitals for any necessary help. To help pilgrims prepare for the next spatio-temporal rituals, a Weather Service was provided with the weather forecast for the three cities most relevant to pilgrims, i.e. Makkah, Madinah and Jeddah. The E-Health Service platform was provided to crowdsource the health information for any possible epidemic threat. It also helped group admins to remain updated about their fellow pilgrims’ health. News about Hajj and Umrah were provided through the News Service. Since navigation is a key feature used by many services, the offline map was introduced, which is automatically stored in the cache to facilitate offline views. It supports navigational services offline, thereby overcoming the Internet connectivity issues. The Shortest Path Service was also part of many other services (e.g. lost and found, finding POI, etc.) where the system returns the shortest path on a navigational map. Learning from users’ experience is always worthwhile. The Twitter service helps pilgrims to share their experiences

10

A. AHMAD ET AL.

with their pilgrims’ social network. The most important service to provide users with a complete guidance of performing Hajj and Umrah rituals is the spatio-temporal and context-aware Rituals Guide. This virtual guide acts as a companion and hence engaged many pilgrims for all spatial-temporal activities both in terms of reminders and in terms of multimedia guidance. 4.3. Server-side architecture Figure 6 shows the server-side architecture details of the proposed cloud-based context-aware crowdsourcing system, which is implemented using the Amazon Web Services (AWS) Framework. Front-end connects with the serve-side framework through the HTTP REST API in order to get context-aware services. A pilgrim’s smartphone App is connected to the server through this API and the format of input data is JSON. Front-end includes the users equipped with smartphones or tablets with an iOS or Android-based system connected to our network as Pilgrims, Hajj Agencies or organisers and accessing a pool of services explained in next section. The server integrates and validates the information by pilgrims and related Hajj agencies. These services could include complaints related to their residence, traffic updates about an accident or road blockage, and sharing a location with the COI. Moreover, camera sensors share related images, audios and videos. The set of in-built

Figure 6. Server-side architecture.

sensors within the mobile phone provides location information, light, sound, ambient temperature, etc. For vehicle tracking services, their spatio-temporal information is collected by OBD-II sensors attached to them. The main components of our API server are as follows. Simple Queue System (SQS): To not lose any message from the front-end during peak hours, we resort to the Amazon Simple Queue System (SQS), which can be configured for an unlimited pipeline of message queues. SQS is a fast, reliable, scalable, fully managed message queuing service and cost-effective for a cloud application. Like standard queues, it follows FIFO (first-in, firstout) and is very effective in transmitting the huge volume of data without losing messages or requiring other services to be always available. If the response is an error due to time-out because of low connection, invalid location or any other invalid request, the SQS error queue will deal with it. The server side reads from the SQS, processes the message content and stores the result in either a relational database or in the Dynamo DB. Auto-scaling: Due to the dynamic nature of the crowd, it is difficult to identify the usage/hits on the server without any prior experience. Using a low-end machine or a fixed number of machines can slow down the server or crash the server during heavy load. In large crowds like Hajj and Umrah, there are multiple instances (e.g. day of Arafat) that can generate heavy server-side load. To

BEHAVIOUR & INFORMATION TECHNOLOGY

support this spike of load, the AWS framework helps the framework in easily scaling up both vertically and horizontally. We have defined a threshold for the API response time. Cloudwatch is a monitoring service to monitor the performance of EC2 including log files, alarms and metrics. If cloudwatch found that the response time of EC2 is more than the defined threshold response time, it is configured to create new instances dynamically to handle the load using the load balancing elastic beanstalk. Furthermore, if the number of the instances is more than the requirement, it will also delete extra instances, which is very cost-effective for a dynamic scenario such as Hajj. During peak hours, multiple autoscaled servers read from the unlimited pipeline of message queues, process the content based on the requests and store the results in either a relational database PostgreSQL or in the Dynamo DB. PostgreSQL/PostGIS: We used open source software to implement a mashup of services such as users’ current location, old trajectory and POIs and stored them in the PostgreSQL database. To implement a navigation service or to find the nearest POIs, we extracted Open street map data of Saudi Arabia and stored them in the PostGIS database. The PostGIS database supports spatial queries on top of the PostgreSQL. Specifically, we used pgrouting to support the open source navigation service. For any enquiry by the user, the application server reads one’s location and executes a spatial query in the PostgreSQL database by providing the output at the user’s front-end interface. DynamoDB: For OLTP (online transactional processes) and OLAP (online analytical processing), users’ input data are stored in the cloud through the Dynamo DB AWS framework. Dynamo DB is an example of a NoSQL database that can retrieve and store real-time data. Similar to other big data architecture such as Hadoop and Cloudera, Dynamo DB also supports the map reduce methodology to provide scalability and to improve the efficiency. We use all the location-aware data related to pools of services used by users in dynamoDB. Moreover, we used the MapReduce framework for OLAP, which allows to process data independently and do parallel processing efficiently. Amazon S3: For big data storage, we use Amazon S3. It provides an interface to store and retrieve any amount of data from anywhere at any moment of time efficiently. We used S3 for logs files that are generated by EC2 and backup of users’ archived data. 4.4. POI collection While collecting the overall usability of our deployed system, we observed users’ diligence in collecting POIs for

11

personal or for public use. In this section, we explain in detail what incentive mechanism was followed in collecting different POIs by users and web administrators, which were later approved to be available for instant use. Appendix E (a, b) is the description of an instance of a user interface that was used by pilgrims to add POI information with related location and multimedia content to describe the point visually and submit. Users were given access to this screen by dual channels, one directly from adding a POI and second from a favourite point. The favourite channel was dedicated to POIs added for personal use (tent in Mina, residence in Makkah, favourite restaurant to visit later, laundry, etc.), whereas from another channel, POIs added were open to be viewed by all users. Once the POIs were added to our server, they were available to view only after approval from the App administrators. Administrators were given rights to edit/modify the added POI, and or rejected due to incompleteness or irrelevancy in the data, especially POI location. Once the POIs are added and approved by the administrators, they were available to be viewed. Appendix E (c) is user interface offered, where users can filter the required POI type and can visualise. Appendix E (d) shows an interface where users can view POIs on a map to select the best fit as per requirement. A special zooming mechanism will allow visualising more details about each POI at any spatial location. A user is given multiple options after visualising a POI. The Review option allows reviewing the name and or any other missing information, Add Fav allows adding the selected POI to one’s favourite list, the Navigate option allows travelling from the current position to the selected POI and Show Reviews reveals other users’ experiences about the POI. Figure 7 gives an overview of adding the POI activity by users during a Hajj event. Holy places being central parts of all spatial zones were added more as POIs. Salient POIs by their types are presented in Figure 7.

5. System evaluation To validate our deployed system, we conducted a field survey after the Hajj event. By considering mainly users who opted for English and Arabic Languages, we circulated a set of questionnaires among users in these two languages through personal contacts and online survey services. Among Perform Hajj and Umrah App users, 212 participated and provided their feedback, which was analysed to know overall rating, system usability and complete user experiences by using different offered services. User feedback and reviews were quite satisfactory, motivating and encouraging, especially

12

A. AHMAD ET AL.

Figure 7. POI(s) collected from the pilgrims’ crowdsourced data shown with %values.

from users who performed Hajj and Umrah for the first time in their lives. Users’ responses have been categorised into multiple evaluation standards and are presented in this section.

5.1. Overall rating This section explains user satisfaction by rating the application over five rating scales: best imaginable, excellent, good, fair and worst imaginable (Figure 8). We have taken feedback from both types of users: Those who have performed Hajj before with a knowledge about such a context-aware system and users with first time experience. Overall rating showed that the top three scales were more than 80%.

5.2. System usability To know the overall system usability, the subset of the survey questionnaire consisting of questions about the importance of the system was analysed. These questions

were designed to know whether it is easy or difficult to use the system, technical background needed, the performance of distinct functions, HCI (Human-Computer Interaction) aspect, attraction to learn quickly and user confidence level. We have subdivided the set of questions to measure different outcomes about system usability: 1Tech. Know. (Technical Knowledge), which explains the users’ prior knowledge to use the App or instant guidance required. 2-Difficult to Use addresses the complexity of distinct functions to use and inconsistency in different screens related to different functions. 3- Easy to Use explains confidence level of users while they used the App for multiple functions and friendly user interfaces which enhance users’ learning ability. 4Importance of the App to be used as a companion during the Hajj event. These four categories of questionnaire items were rated on a five-item rating scale, namely Strongly Agree (STAG), Agree (AGRE), Neutral (NEUT), Disagree (DIAG) and Strongly Disagree (STDA). Figure 9 summarises the discussion about these categories of system usability on these five rating scales. Majority of the users (among STDA and DIAG), which is about 50%, gave their opinion that not much technical knowledge is needed to get the benefit from the application. About 78% users (among NEUT, AGRE and STAG categories) reported that it was easy to use the application (Kwon, Choi, and Kim 2007). Moreover, a higher percentage of users voted in favour of an application that it is an essential companion for real-time guidance. 5.3. Service details

Figure 8. Overall rating.

Figure 10 shows users’ activeness while using different services. About 100 users were active during the whole

BEHAVIOUR & INFORMATION TECHNOLOGY

13

Figure 9. Users’ experience of system usability.

event and were seen between 100 and 200 times. A statistical analysis about different services used during their pilgrimage is summarised in Table 1. Based on data shown in Figure 10 and Table 1, users were asked different questions about all offered services. Users’ responses were scaled over a five-item rating scale that included STAG, AGRE, NEUT, DIAG and STDA. The main objective of this discussion was to know the overall user experience about the system’s effectiveness. Based on the ISO 9241 definition, we evaluate our system on three measures, Effectiveness, Efficiency and Satisfaction. The effectiveness of the system is validated by the number of times each service was used (Table 1). Our system’s log files showed that some services such as Hajj Rituals, Out of Boundary and Finding Favourite Friend services were frequently used. Users’ satisfaction about these three services are also very promising

(Figure 11). Around 89% of our users thought the Hajj and Umrah mobile application was very efficient in helping them to find the prayer time in any location, remember the sequence of particular rituals and check if they were within the boundary of certain spatial zones (Ahmad, Afyouni, Murad, Rahman, Ur Rehman, et al. 2015b). Figure 11 shows some of the most popular services’ usability analysis. The overall results are quite satisfactory and encouraging. They support the satisfaction metric of users where a large number of users showed their interest in reusing the services multiple time. Figure 11 also shows the STAG measure of satisfaction that tops the five measures. Figure 12 shows spatio-temporal intrauser activities of a single user and different services used over the whole Hajj event. The higher numbers of many services show user satisfaction about the system and are a convincing factor for overall effectiveness of the system.

Figure 10. Number of users seen on 21 September 2014 to 8 October 2014 during the Hajj 2014 event.

14

A. AHMAD ET AL.

Table 1. Service usage chart during Hajj. Services Hajj rituals Update traffic Health form Find POI Hajj&Umrah news Twitter search map Currency exchange Umrah rituals Prayer times

%

Services

%

13.5% 1.2% 0.7% 4.0% 6.1% 0.4% 3.2% 5.6% 6.7%

Out of boundary Find favourite POI Find favourite friend Add POI Complaints Weather Emergency SMS Add to favourite list Translation

12.6% 1.1% 30.4% 4.1% 0.3% 4.2% 0.6% 0.4% 4.9%

5.4. Discussion The main objective of this research was to first capture pilgrims’ contexts and then using our proposed model to provide them with the needed services. We successfully deployed the proposed system presented in the above section, Systems’ evaluation, in detail. Users’ satisfaction (Figure 11) includes usage of the zoning model in modelling pilgrims’ contexts. However, connectivity and battery issues were grave concerns. The summary of the ‘number of users seen’, presented in Figure 12, illustrates that to stay online for longer periods was a significant issue. This was due to their devices’ batteries and/or complete loss of wireless connectivity, thereby deterring them from using many of the Internet-dependent services. To address the issue, our quick response was allowing about 90% of the services to work offline. Offline maps were introduced for browsing POIs with some other services that need the shortest path. Battery issue was mostly reported from Day 2 onwards because pilgrims were away from their hotels, and their transition

in tents, buses or trains offered them limited access to power outlets. One way to increase the battery life is by reducing the need for Internet operations and GPS usage that depletes the battery faster than others. Providing offline maps is also expected to increase the battery life since it eliminates the need to communicate with the server. In addition, educating the users to keep their connectivity off while they were using offline services can also improve the battery life. Moreover, we found a considerable number of users who reported about technical knowledge required and the difficulty level to use the deployed application. Their dissatisfaction educated us to further work on and improve the user interfaces considering services’ usage statistics (Figure 12) where numbers were less which could be due to these two raised concerns.

6. Conclusion and future work We presented a crowd-sensing system that is specifically designed for providing context-aware services to a very large heterogeneous crowd, helping users to perform certain spatio-temporal obligatory rituals. Our proposed system takes advantage of a mobile phone’s sensory data to determine user contexts. Moreover, the spatiotemporal zoning model was favourable for deducing users’ contexts for their spatio-temporal activities, to offer suitable services at the right time. As a use case, we have included implementation details of our deployed testbed that was used by pilgrims during Hajj 2014 and 2015 events. Further, we presented users’ feedback that

Figure 11. Users’ satisfaction about services offered over the five evaluation measures.

BEHAVIOUR & INFORMATION TECHNOLOGY

15

Figure 12. A single user’s activities showing usage of different services during the Hajj 2014 event.

helped us to analyse system evaluation based on the overall rating, system usability and users’ interest in using the offered services. We have discussed users’ diligence about POIs for different Hajj scenarios that can be generalised to any similar large crowd. Since we can collect a very large number of pilgrims’ data, we plan to extend our research to analyse the captured data to deduce more dimensions of semantic knowledge, such as efficient handling of the emergency situation, information diffusion process during accidents, isolating events based on zone and notify users of respective zones. Information diffusion processes are considered by finding pilgrims with higher connectivity within their social network and are resourceful in bridging two disjoint components of the network based on different languages spoken by pilgrims. Moreover, we plan to analyse users’ health data to possibly model the Middle East Respiratory Syndrome Coronavirus (MERS-CoV) spreading, which is one of the serious concerns for future pilgrimages.

Disclosure statement No potential conflict of interest was reported by the authors.

Funding This work was supported by the King Abdulaziz City for Science and Technology: [Grant Number 11-INF1683-10, 11-INF1700-10 and 13-INF2455-10.].

ORCID Akhlaq Ahmad http://orcid.org/0000-0001-7449-7724 Md. Abdur Rahman http://orcid.org/0000-0002-4105-0368 http://orcid.org/0000-0002Mohamed Ridza Wahiddin 0244-2115 Faizan Ur Rehman http://orcid.org/0000-0001-6230-3056 Abdelmajid Khelil http://orcid.org/0000-0002-4536-8058 Ahmed Lbath http://orcid.org/0000-0002-8457-6941

References Abowd, Gregory D., Christopher G. Atkeson, Jason Hong, Sue Long, Rob Kooper, and Mike Pinkerton. 1996. “Cyberguide : A Mobile Context-aware Tour Guide.” Baltzer Journals 3: 1–21. doi:10.1023/A:1019194325861. Ahmad, Akhlaq, Md. Abdur Rahman, Imad Afyouni, Faizan Ur Rehman, Bilal Sadiq, and Mohamed Ridza Wahiddin. 2014a. “Towards a Mobile and Context-aware Framework from Crowdsourced Data.” In The 5th International Conference on Information and Communication Technology for the Muslim World (ICT4M), 1–6. Kuching: IEEE. doi:10.1109/ICT4M.2014.7020672. Ahmad, Akhlaq, Md. Abdur Rahman, Faizan Ur Rehman, Ahmed Lbath, Imad Afyouni, Abdelmajid Khelil, Syed Osama Hussain, Bilal Sadiq, and Mohamed Ridza Wahiddin. 2014b. “A Framework for Crowd-sourced Data Collection and Context-aware Services in Hajj and Umrah.” In 2014 IEEE/ACS 11th International Conference on Computer Systems and Applications (AICCSA), 405– 412. IEEE. doi:10.1109/AICCSA.2014.7073227. Ahmad, Akhlaq, Imad Afyouni, Abdullah Murad, Md. Abdur Rahman, Faizan Ur Rehman, Bilal Sadiq, Saleh Basalamah, and Mohamed Ridza Wahiddin. 2015a. “ST-diary: A Multimedia Authoring Environment for Crowdsourced Spatio-temporal Events.” In Proceedings of the 8th ACM

16

A. AHMAD ET AL.

SIGSPATIAL International Workshop on Location-based Social Networks – LBSN’15, 1–7. New York: ACM Press. doi:10.1145/2830657.2830664. Ahmad, Akhlaq, Imad Afyouni, Abdullah Murad, Abdur Rahman, Faizan Ur Rehman, Bilal Sadiq, and Mohamed Ridza Wahiddin. 2015b. “Quality and Context-aware Data Collection Architecture from Crowd-sourced Data.” In International Multi-topic Conference, 73–83. Jamshoro, Pakistan. Ahmad, Akhlaq, Md. Abdur Rahman, Bilal Sadiq, Shady Mohammed, Saleh Basalamah, and Mohamed Ridza Wahiddin. 2015c. “Visualization of a Scale Free Network in a Smartphone-based Multimedia Big Data Environment.” In IEEE International Conference on Multimedia Big Data, 286–287. IEEE. doi:10.1109/BigMM.2015.48. Ahmad, Akhlaq, Faizan Ur Rehman, Md. Abdur Rahman, Abdullah Murad, Ahmad Qamar, Bilal Sadiq, Salah Basalamah, and Mohamed Ridza Wahiddin. 2015d. “Idiary.” In Proceedings of the 23rd ACM International Conference on Multimedia – MM ‘15, 809–810. New York: ACM Press. doi:10.1145/2733373.2807983. Al-Hashedi, Abdullah Hussein, Muhammad Rafie Mohd Arshad, Hasimah Hj Mohamed, and Ahmad Suhaimi Baharuddin. 2011. “Identifying the Determinants of RFID Adoption Intention in Hajj Organizations.” In 2011 International Conference on Research and Innovation in Information Systems, 1–6. IEEE. doi:10.1109/ICRIIS.2011. 6125714. Alt, Florian, Alireza Sahami Shirazi, Albrecht Schmidt, Urs Kramer, and Zahid Nawaz. 2010. “Location-based Crowdsourcing: Extending Crowdsourcing to the Real World.” In Proceedings of the 6th Nordic Conference on Human-computer Interaction: Extending Boundaries, 13– 22. New York: ACM Press. doi:10.1145/1868914.1868921. Anis, M., and Z. Saeed. 2011. “Wireless Network Architecture and RF Transceiver Front-end to Provide Multilingual Speech Translation and Local Positioning System for Hajj Operations.” In 2011 Saudi International Electronics, Communications and Photonics Conference (SIECPC), 1–3. IEEE. doi:10.1109/SIECPC.2011.5876929. Basalamah, Anas. 2016. “Sensing the Crowds Using Bluetooth Low Energy Tags.” IEEE Access 6 (1): 1–1. doi:10.1109/ ACCESS.2016.2594210. Benerecetti, Massimo, Paolo Bouquet, and Matteo Bonifacio. 2001. “Distributed Context-aware Systems.” HumanComputer Interaction 16 (2): 213–228. doi:10.1207/ S15327051HCI16234_06. Bigham, Jeffrey P., Samual White, Tom Yeh, Chandrika Jayant, Hanjie Ji, Greg Little, Andrew Miller, et al. 2010. “VizWiz: Nearly Real-time Answers to Visual Questions Jeffrey.” In Proceedings of the 23nd Annual ACM Symposium on User Interface Software and Technology – UIST ‘10, 333. New York: ACM Press. doi:10.1145/1866029.1866080. Burke, J., D. Estrin, M. Hansen, A. Parker, N. Ramanathan, S. Reddy, and M. Srivastava. 2006. “Participatory Sensing.” In Centre for Embedded Network Sensing. Campbell, Andrew T., Shane B. Eisenman, Nicholas D. Lane, Emiliano Miluzzo, and Ronald A. Peterson. 2006. “Peoplecentric Urban Sensing.” Proceedings of the 2nd Annual International Workshop on Wireless Internet – WICON ‘06, 18–31. Boston, Massachusetts, USA. doi:10.1145/ 1234161.1234179.

Chatzimilioudis, Georgios, Andreas Konstantinidis, Christos Laoudias, and Demetrios Zeinalipour-Yazti. 2012. “Crowdsourcing with Smartphones.” IEEE Internet Computing 16 (5): 36–44. doi:10.1109/MIC.2012.70. Cheverst, Keith, Nigel Davies, Keith Mitchell, and Paul Smith. 2000. “Providing Tailored (Context-aware) Information to City Visitors.” In: Adaptive Hypermedia and Adaptive WebBased Systems: AH 2000, LNCS, edited by P. Brusilovsky, O. Stock, and C. Strapparava, Vol. 1892, 73–85. Berlin: Springer. Dey, Anind, Gregory Abowd, and Daniel Salber. 2001. “A Conceptual Framework and a Toolkit for Supporting the Rapid Prototyping of Context-aware Applications.” Human-Computer Interaction 16 (2): 97–166. doi:10.1207/ S15327051HCI16234_02. Estrin, D. 2010. “Participatory Sensing: Applications and Architecture [Internet Predictions].” IEEE Internet Computing 14 (1): 12–42. doi:10.1109/MIC.2010.12. Fathnan, Ashif Aminulloh, Canggih Puspo Wibowo, Novrian Fajar Hidayat, Dimas Agil Marenda, and Ridi Ferdiana. 2010. “Web-based Hajj Simulation Software.” In Information and Communication Technology for the Muslim World ICT4M 2010 International Conference. Jakarta, Indonesia. doi:10.1109/ICT4M.2010.5971881. Furht, Borko, and Ankur Agarwal. 2013. Handbook of Medical and Healthcare Technologies, edited by Borko Furht and Ankur Agarwal. New York: Springer. doi:10.1007/978-14614-8495-0. Ganti, Raghu, Fan Ye, and Hui Lei. 2011. “Mobile Crowdsensing: Current State and Future Challenges.” IEEE Communications Magazine 49 (11): 32–39. doi:10. 1109/MCOM.2011.6069707. Hameed, Shihab A. 2010. “ICT to Serve Hajj: Analytical Study.” In International Conference on Computer and Communication Engineering (ICCCE’10), 1–7. IEEE. doi:10.1109/ICCCE.2010.5556793. Hara, Tenshi, Thomas Springer, Klemens Muthmann, and Alexander Schill. 2014. “Towards a Reusable Infrastructure for Crowdsourcing.” In 2014 IEEE/ACM 7th International Conference on Utility and Cloud Computing, 618–623. IEEE. doi:10.1109/UCC.2014.98. Höller, Jan, Vlasios Tsiatsis, Catherine Mulligan, Stamatis Karnouskos, Stefan Avesand, and David Boyle. 2014. “Participatory Sensing.” In From Machine-to-Machine to the Internet of Things, 295–305. Boulder, CO: Elsevier. doi:10.1016/B978-0-12-407684-6.00015-2. Khan, Esam A. 2011. “An RFID-based System for Pilgrim Management in King Abdul Aziz International Airport.” In 2011 International Conference on Information Management, Innovation Management and Industrial Engineering, 1, 124–129. IEEE. doi:10.1109/ICIII.2011.36. Korthaus, Axel, and Wei Dai. 2012. “Crowdsourcing in Heterogeneous Networked Environments – Opportunities and Challenges.” In 2012 15th International Conference on Network-based Information Systems, September. 483–488. IEEE. doi:10.1109/NBiS.2012.130. Kwon, O., K. Choi, and M. Kim. 2007. “User Acceptance of Context-aware Services: Self-efficacy, User Innovativeness and Perceived Sensitivity on Contextual Pressure.” Behaviour & Information Technology 26 (2015): 483–498. doi:10.1080/01449290600709111. Mantoro, Teddy, Abi Dzar Jaafar, Mohd Fadhli Md Aris, and Media A. Ayu. 2011. “HajjLocator: A Hajj Pilgrimage

BEHAVIOUR & INFORMATION TECHNOLOGY

Tracking Framework in Crowded Ubiquitous Environment.” International Conference on Multimedia Computing and Systems –Proceedings, doi:10.1109/ICMCS. 2011.5945629. Miluzzo, Emiliano, N. D. Lane, Kristóf Fodor, Ronald Peterson, Hong Lu, Mirco Musolesi, S. B. Eisenman, Xiao Zheng, and A. T. Campbell. 2008a. “Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the Cenceme Application.” Proceedings of the 6th ACM Conference on Embedded Network Sensor Systems, 337–350. doi:10.1145/1460412.1460445. Miluzzo, Emiliano, Nicholas D. Lane, Kristóf Fodor, Ronald Peterson, Hong Lu, Mirco Musolesi, Shane B Eisenman, Xiao Zheng, and Andrew T. Campbell. 2008b. “Sensing Meets Mobile Social Networks.” In ACM Conference on Embedded Network Sensor Systems, 337–350. New York: ACM Press. doi:10.1145/1460412.1460445. Mitchell, Ricardo O., Hammad Rashid, Fakir Dawood, and A. AlKhalidi. 2013. “Hajj Crowd Management and Navigation System: People Tracking and Location Based Services via Integrated Mobile and RFID Systems.” In 2013 International Conference on Computer Applications Technology (ICCAT), 1–7. IEEE. doi:10.1109/ICCAT.2013. 6522008. Mohandes, Mohamed. 2008. “An RFID-based Pilgrim Identification System (a Pilot Study).” In 2008 11th International Conference on Optimization of Electrical and Electronic Equipment, 107–112. IEEE. doi:10.1109/ OPTIM.2008.4602508. Mohandes, Mohamed Ahmed. 2015. “Mobile Technology for Socio-religious Events: A Case Study of NFC Technology.” IEEE Technology and Society Magazine 34 (1): 73–79. doi:10.1109/MTS.2015.2396122. Mohandes, M., M. A. Haleem, A. Abul-Hussain, and K. Balakrishnan. 2011. “Pilgrims Tracking Using Wireless Sensor Network.” In 2011 IEEE Workshops of International Conference on Advanced Information Networking and Applications, 325–328. IEEE. doi:10.1109/ WAINA.2011.123. Mohandes, Mohamed, Mohamed Haleem, Mohamed Deriche, and Kaviarasu Balakrishnan. 2012. “Wireless Sensor Networks for Pilgrims Tracking.” IEEE Embedded Systems Letters 4 (4): 106–109. doi:10.1109/LES.2012.2215913. Mohd Yasin, Anita, Fakhrul Hazman Yusoff, Mohd Ali Mohd Isa, and Nurul Hidayah Mat Zain. 2010. “Avatar Implementation in Virtual Reality Environment Using Situated Learning for “sa’i” (Muslim Hajj Ritual).” 2010 International Conference on Educational and Information Technology 2: 286–290. IEEE. doi:10.1109/ICEIT.2010.5607574. Muller, Michael, Susanne Hupfer, Stephen Levy, Daniel Gruen, Reid Priedhorsky, and Andrew Sempere. 2011. “MoCoMapps – An Experiment in Crowdsourcing Both Data and Applications for Mobile Services.” In CHI 2011 Workshop, Data Collection: By the People, for the People, 1–4. Vancouver, BC, Canada, May 2011. Mulyana, Willy Wahyu, and Teddy Surya Gunawan. 2010. “Hajj Crowd Simulation Based on Intelligent Agent.” In International Conference on Computer and Communication Engineering (ICCCE’10), 1–4. IEEE. doi:10.1109/ICCCE.2010.5556818. Murturi, Arber, Burak Kantarci, and Sema F. Oktug. 2015. “A Reference Model for Crowdsourcing as a Service.” In 2015

17

IEEE 4th International Conference on Cloud Networking (CloudNet), 64–66. IEEE. doi:10.1109/CloudNet.2015. 7335281. Nafea, I., M. N. Bhairy, and Z. A. Zeidan. 2014. “Health Tracking Framework for Hajj Pilgrims Using Electronic Health Records for Hajj.” In 2014 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 605–607. IEEE. doi:10.1109/BIBM.2014.6999230. Pan, Yue, and Eli Blevis. 2011. “A Survey of Crowdsourcing as a Means of Collaboration and the Implications of Crowdsourcing for Interaction Design.” In 2011 International Conference on Collaboration Technologies and Systems (CTS). May, 397–403. IEEE. doi:10.1109/CTS. 2011.5928716. Rahman, Abdur, Nor Asilah Wati Abdul Hamid, Amir Rizaan Rahiman, and Basim Zafar. 2015. “Towards Accelerated Agent-based Crowd Simulation for Hajj and Umrah.” In 2015 International Symposium on Agents, Multi-agent Systems and Robotics (ISAMSR), 65–70. IEEE. doi:10.1109/ ISAMSR.2015.7379132. Rahman, Md, Heung-Nam Kim Abdur, Abdulmotaleb El Saddik, and Wail Gueaieb. 2012. “A Context-aware Multimedia Framework Toward Personal Social Network Services.” Multimedia Tools and Applications 71 (3): 1717–1747. doi:10.1007/s11042-012-1302-y. Rehman, Faizan Ur, Md. Ahmed Lbath, Abdur Rahman, Saleh Basalamah, Imad Afyouni, Akhlaq Ahmad, and Syed Osama Hussain. 2014. “Toward Dynamic Path Recommender System Based on Social Network Data.” In Proceedings of the 7th ACM SIGSPATIAL International Workshop on Computational Transportation Science – IWCTS ‘14, 64–69. New York: ACM Press. doi:10.1145/ 2674918.2674927. Rehman, Faizan Ur, Ahmed Lbath, Bilal Sadiq, Akhlaq Ahmad, Abdullah Murad, Faizan Ur Rehman, Ahmed Lbath, et al. 2015. “A Constraint-aware Optimized Path Recommender in a Crowdsourced Environment.” In 2015 IEEE/ACS 12th International Conference of Computer Systems and Applications (AICCSA), Morocco: IEEE, 1–8. doi:10.1109/AICCSA.2015.7507185. Shah, Sumit, Fenye Bao, Chang-Tien Lu, and Ing-Ray Chen. 2011. “CROWDSAFE: Crowd Sourcing of Crime Incidents and Safe Routing on Mobile Devices (Demo Paper).” In Proceedings of the 19th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems – GIS ‘11, 521. New York: ACM Press. doi:10.1145/2093973.2094064. Sulaiman, Shahida, Hasimah Mohamed, Muhammad Rafie Mohd Arshad, Nur’ Aini Abdul Rashid, and Umi Kalsom Yusof. 2009. “Hajj-QAES: A Knowledge-based Expert System to Support Hajj Pilgrims in Decision Making.” 2009 International Conference on Computer Technology and Development 1: 442–446. IEEE. doi:10.1109/ICCTD.2009.190. Xu, Zheng, Yunhuai Liu, Neil Yen, Lin Mei, Xiangfeng Luo, Xiao Wei, and Chuanping Hu. 2016. “Crowdsourcing Based Description of Urban Emergency Events Using Social Media Big Data.” IEEE Transactions on Cloud Computing, 1–1. doi:10.1109/TCC.2016.2517638. Yan, Zhixian, and Dipanjan Chakraborty. 2014. “Semantics in Mobile Sensing.” Synthesis Lectures on the Semantic Web: Theory and Technology 4 (1): 1–143. doi:10.2200/ S00577ED1V01Y201404WBE008.

18

A. AHMAD ET AL.

Yan, Tingxin, Matt Marzilli, Ryan Holmes, Deepak Ganesan, and Mark Corner. 2009. “mCrowd – A Platform for Mobile Crowdsourcing.” In Proceedings of the 7th ACM Conference on Embedded Networked Sensor Systems – SenSys ‘09. New York: ACM Press, 347. doi:10.1145/ 1644038.1644094. Yurur, Ozgur, Chi Harold Liu, and Wilfrido Moreno. 2014. “A Survey of Context-aware Middleware Designs for Human Activity Recognition.” IEEE Communications Magazine. doi:10.1109/MCOM.2014.6829941. Yusoff, Mohd Fitri, Abdul Nasir Zulkifli, and Nor Fadziana Faisal Mohamed. 2011. “Virtual Hajj (V-Hajj) — Adaptation of Persuasive Design in Virtual Environment (VE) and Multimedia Integrated Approach Learning Courseware Methodology.” In 2011 IEEE Conference on

Open Systems, 250–255. IEEE. doi:10.1109/ICOS.2011. 6079280. Zambonelli, Franco. 2011. “Pervasive Urban Crowdsourcing: Visions and Challenges.” In 2011 IEEE International Conference on Pervasive Computing and Communications Workshops (PERCOM Workshops), 578–583. Seattle, WA: IEEE. doi:10.1109/PERCOMW.2011.5766956. Zeki, Akram M., Hassen Alsafi, Rasheed M. Nassr, and Teddy Mantoro. 2012. “A Mobile Dictionary for Pilgrims.” In 2012 International Conference on Information Technology and Eservices, 1–5. IEEE. doi:10.1109/ICITeS.2012.6216620. Ziefle, M., and S. Bay. 2005. “How Older Adults Meet Complexity: Aging Effects on the Usability of Different Mobile Phones.” Behaviour & Information Technology 24 (5): 375–389. doi:10.1080/0144929042000320009.

Appendix Perform the Hajj and Umrah application A. Salient features

The Hajj and Umrah application is designed to provide pilgrims with all the necessary steps required for performing Hajj based on a user’s day, time and location in a personalised way. It has the following key features: . . . . .

A cloud-based architecture by combining a set of web services specially tailored for the Hajj and Umrah pilgrimage. Novel spatio-temporal ritual services where we model each ritual as a set of primitive rituals. Each primitive ritual is then mapped to a set of spatio-temporal activities. Each spatio-temporal activity can then be tracked by our novel algorithms running in a smartphone. Using our proposed applications, any number of existing or new web services can be offered to the pilgrims. The proposed framework is multilingual.

BEHAVIOUR & INFORMATION TECHNOLOGY .

. . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

. .

19

We have added context-aware capability to identify a user context such as location, time and significant events around the ambient environment of each pilgrim and provide necessary services based on the user context. For this, we use context-awareness in which a pilgrim having a smartphone with Internet connectivity can consume a subset of services in any given situation. We proposed a social network specifically tailored for the pilgrims coming for Hajj and Umrah, which connects all the pilgrims with each other, with their community of interest and different social network services. The proposed framework helps pilgrims in enjoying Hajj- and Umrah-related services in a personalised way at the right place and time. Cloud-based server-side architecture can support a very large crowd’s multimedia data. SMS-based registration and mobile authentication with verification code Passport authentication Lost and found services Out-of-boundary services Location-based dynamic Hajj guide Dynamic users’ interactive step-by-step guide to perform Hajj Ifrad Dynamic users’ interactive step-by-step guide to perform Hajj Tamattu Dynamic users’ interactive step-by-step guide to perform Hajj Qiran Dynamic users’ interactive step-by-step guide to perform Umrah Dynamic uses’ interactive step-by-step guide to perform Tawaaf Places of interest (POI) shows nearby hotels, mosques, hospitals, pharmacies, barbers, laundries, money exchanges, zamzam points, toilets, parking lots, restaurants, holy places, fish markets, vegetable markets, fruit markets, shops, malls, parks, ATMs, slaughter houses, petrol pumps, hotels with parking, and hotels without parking, to name a few. Shortest path to the point of interest from the current location. Current waiting time for a money exchange through crowdsourcing. Availability of parking and hotel through crowdsourcing. Prayer times with azan feature at the time of prayer. Time left for the next prayer and next prayer time reminder. Also, the qibla direction is available from any place, which is location-aware. Weather about Jeddah, Makkah and Madinah News feed mashup related to Hajj and Umrah. Twitter service related to Hajj and Umrah ○ Map-based tweet view ○ Text view based on search ○ Tweets ○ Heat map Currency converter and location of a nearby money exchange. Emergency service that shows a police station, fire station and ambulance and their shortest routes in real-time. Translation and text-to-speech service Crowdsourced traffic updates service, where a user updates about the traffic and earns credit. Complaint service Health service Lost and found service ○ Friends: Shows nearby friends with the path to the friend in real-time ○ Places: Nearby places and places in favourites Hajj Messenger with audio, video and text message exchange features. Reviews about POI ○ Show reviews based on current location ○ Save to POI as favourite ○ Search POIs and view favourite POIs ○ Add new point of interest to earn credit

20

A. AHMAD ET AL.

B. Main menu Two Screens offering multiple services as shown below.

C. Rituals service A user is given the option to select Umrah, Hajj and/or Tawaaf only. If he selects Tawaaf, then multiple screens guide him till he finishes by offering supplications for each round and helps in counting the number of rounds (1–7).

BEHAVIOUR & INFORMATION TECHNOLOGY

21

D. Types of Hajj service A pilgrim is offered three types of Hajj to opt from. The following screens will guide him during each step till the end.

E. POIs adding and navigation (a) A user taps on the screen to add a POI. (b) The user needs to fill in the data about the POI and submit. (c) The user can filter the POIs to choose their desired one. (d) The user visualises the filtered POI (e.g. nearby hotels).

22

A. AHMAD ET AL.

F. News, weather and currency converter services News about Hajj and Umrah; weather for three main cities, Makkah, Madinah and Jeddah; and the Currency converter helps in knowing the current exchange rates and offers to navigate to nearby currency exchange centres.

BEHAVIOUR & INFORMATION TECHNOLOGY

23

G. Out-of-boundary service Out-of-boundary service is the most popular service. The pilgrims can check whether they are inside or outside of the Mina, Muzdalifah, Arafat and Haram boundaries. They can visualise their current position with the selected boundary over map and can get messages if they are inside or outside the selected boundary. This service is important because most of the Hajj and Umrah rituals are constrained to these spatial zones.

24

A. AHMAD ET AL.

H. Traffic update, translation and emergency services Users can update traffic by seeing if there is any accident, road block and or if there is a temporary diversion. It helps the subsequent visitors to find alternative paths to avoid congestion. The translation service with audio facility help users to communicate with one another and overcome the language barrier. The emergency service helps users to communicate to the police, fire station or other related services when required. They can contact the Hajj agency officials and share SMS for some emergency situation.

I. Registering a complaint, Qibla and chat services Pilgrims can register complaints against quality of services in their residence. Qibla Compass helps them to find the right direction to pray, especially in heavily crowded areas. Through the Hajj Messenger service, pilgrims can share their experiences through audio and video, which will be helpful, especially in lost and found-related issues.

BEHAVIOUR & INFORMATION TECHNOLOGY

25

J. Favourite places and friends service and navigation Pilgrims can browse their previously added favourite places. They can browse their friends’ location with visualisation over the map. They can get their route to reach one another.