Towards a Sustainable Digital Planning Model for Traffic Management in Religious Heritage Cities: Employing Augmented Reality and Space Syntax in the Historical Center of Karbala

2.1 Augmented reality in heritage wayfinding and crowd management

A selective critical review of recent scholarship has advanced the understanding of AR in heritage and navigation contexts, yet a critical gap persists at the intersection of AR-based cognitivesupport, crowd management, and extreme-density pilgrimage settings. The four most pertinent and recent studies are reviewed below to delineate the precise contours of this gap (Table 1).

Table 1. Comparative review of the four most relevant recent studies

Man et al. [11] conducted a comprehensive systematic review and meta-analysis of 13 studies involving 400 participants, confirming that AR significantly enhances wayfinding performance (SMD = 0.992) and overall navigation effectiveness (overall effect size =0.554). Despite the robustness of this meta-analytic evidence, the authors identified no study situated within a pilgrimage context or involving extreme pedestrian density, thereby empirically substantiating the research gap that the present study addresses.

In the domain of AR-enabled crowd management, Koulouris et al. [12] proposed a gamified AR prototype integrating computer vision for visitor flow regulation in a museum environment. Although this work demonstrates the conceptual feasibility of AR-based crowd management, it is confined to a controlled, low-density, indoor setting and does not engage with the morphological complexity or spiritual sensitivities characteristic of heritage religious cities.

Complementing these navigation and management perspectives, Zhang et al. [13] examined immersive AR-based navigation systems and reported improvements in cognitive map development and wayfinding performance; however, their study was conducted in an indoor environment and did not address crowd pressure in religious heritage settings.specifically aimed at supporting mental map construction and wayfinding. Their experimental findings confirmed significant improvements in spatial cognition [15]; however, the study was conducted under normal-density urban conditions and did not incorporate crowd pressure as an experimental variable, limiting its applicability to high-density pilgrimage environments.

Within the specific context of Islamic mass gatherings, Attia and Farhan [5] discussed urban heritage revitalization challenges in the holy city of Karbala, highlighting the need for strategies that balance heritage preservation with visitor-related urban pressures. While the study provides valuable insights into crowd dynamics and recommends improvements in urban design and simulation, it remains purely observational and does not test any digital AR intervention or measure cognitive outcomes.

None of the most advanced and directly relevant studies to date has deployed and experimentally tested an AR system as a dual-function cognitive layer for wayfinding and crowd management within an Islamic heritage pilgrimage environment under authentic extreme-density conditions. The present study is specifically designed to fill this multi-dimensional gap [5, 11-13].

2.2 Theoretical framework

The study is grounded in four complementary theories that collectively explain the cognitive collapse in high-density pilgrimage environments and underpin the design of the AR model as a dual-function cognitive layer. The term “cognitive layer” is used throughout this study in a precise operational sense: a real-time digital stratum superimposed onto the physical urban environment that neither replaces nor restructures the heritage fabric, but selectively filters, reconstructs, and redistributes spatial information to compensate for the perceptual failures induced by high crowd pressure. What distinguishes a cognitive layer from a conventional navigation application is its dual function—simultaneously supporting individual wayfinding decisions and enabling collective crowd-flow management-and its theoretical grounding in the gap between available and required cognitive resources as defined by Sweller [3].

1. Cognitive Load Theory [3]: The convergence of crowd navigation, visual noise, and information scarcity inflates extraneous cognitive load, depleting the visitor's mental energy and weakening both individual wayfinding and responsiveness to crowd management guidance. The model addresses this by offloading cognitive tasks to the device interface and filtering environmental information to the essential minimum [16].

2. The Image of the City [2]: The legibility of paths, nodes, landmarks, edges, and districts collapses under crowd pressure and temporary structures. The AR layer digitally reconstructs these elements: directional arrows clarify paths, service markers highlight nodes, and annotations preserve landmark reference points, supporting both individual orientation and collective flow management [15].

3. Situation Awareness Theory [17]: The visitor's awareness of their surroundings collapses sequentially (perception → comprehension → projection), reducing them to a particle adrift in the crowd. The model restores this awareness through real-time feedback: displaying occluded elements, facilitating comprehension of spatial relationships, and visualising live density ahead to enable risk projection [18].

4. Genius Loci/Spirit of Place [4]: Wayfinding in Karbala is not a geometric transfer but an integral part of the spiritual experience. The collision of "Transient Urbanism" with the permanent heritage fabric weakens the visitor's connection to the identity of the place. The digital intervention is anchored in the principle of "Augmentation, not Replacement" to restore the legibility of the latent heritage fabric and preserve the spirit of the place [9, 10].

5. Space Syntax theory [1] provides the analytical foundation for understanding the relationship between the topological structure of the street network and the natural movement patterns of pedestrians within it [19, 20]. Integration values derived from Space Syntax analysis are not merely descriptive metrics but predictive indicators of pedestrian flow, establishing a mathematically grounded criterion for identifying the corridors of maximum cognitive stress and embedding spatial evidence directly into the routing logic of the AR model. This theoretical grounding distinguishes the present intervention from conventional navigation applications by anchoring its design decisions in the morphological properties of the historic fabric itself rather than in arbitrary spatial assumptions.

The conceptual logic of the study follows a five-step causal chain, as illustrated in Figure 1 the morphological complexity of the historic urban fabric generates a structural cognitive gap in the visitor, which the AR intervention addresses through real-time spatial information, producing measurable behavioural and cognitive improvements.

This study adopted an exploratory field methodology with a quasi-experimental pre-/post-test design for a single group. Its aim was to examine preliminary indicators of the potential of a web-based AR-enhanced digital NAI in supporting spatial perception, enhancing crowd awareness, and improving access to services within the historic center of Karbala during the pilgrimage season. The methodology combined spatial analysis using Space Syntax methodology with the development of a pilot digital model, followed by field testing on a purposive sample of visitors. Given the single-group pre-/post-test design, the results are interpreted as preliminary indicators that can be further verified, and not as definitive causal evidence of the model's effectiveness. Figure 2 illustrates the overall methodological sequence of the study, from spatial diagnosis to model development, field application, statistical analysis, and expert evaluation.

The methodology began with the spatial diagnosis phase, where the street network was extracted from digital map sources, processed within a Geographic Information System (GIS) environment, and then entered into DepthmapX software for spatial integration analysis. Integration values, particularly the T1024 index, were used to identify the most clearly defined and integrated axes within the morphological structure of the historic center. The analysis showed that Al-Jumhuriya and Al-Sidra Streets were among the most integrated axes within the network, with T1024 values of 535 and 517, respectively. Figure 3 shows the spatial integration map of Karbala's historic center as derived from the Space Syntax analysis. These results were used to determine the locations of field scenarios and to establish the initial spatial foundation for the digital model (Figure 4). It is important to note that integration values are understood in this study as indicators of spatial clarity and readability, not as direct indicators of reduced or avoided congestion.

The model used in this study was developed as a web-based AR-enabled digital NAI, not a full-fledged AR system based on live visual overlays or precise spatial fixation. The model was built using HTML5, CSS3, JavaScript, and the Leaflet.js library to function as a progressive web application accessible via a mobile browser without requiring a separate app installation. It utilizes the Web Geolocation API to determine the user's approximate location and DeviceOrientationEvent to estimate device orientation. The model then displays spatial cues and orientation information to help the visitor understand their location, orientation, and relationship to surrounding corridors and services. In this study, the term "AR" refers to the addition of a lightweight spatial information layer via the browser, rather than a full-fledged AR system based on SLAM, VPS, visual markers, or fixed spatial anchors. Figure 5 shows screenshots of the AR Karbala Navigator model interface as used in the field experiment.

The routing logic in the current version of the model relies on spatial clarity derived from Space Syntax analysis, rather than a fully dynamic rerouting based on real-time crowd data. Spatial integrity values were used as an indicator of corridor readability and were then integrated into a modified routing algorithm based on Dijkstra logic. Therefore, the current model selects the most morphologically distinct routes, not necessarily the least congested. The study acknowledges that relying solely on spatial integration is insufficient for crowd management in high-traffic environments, as the most integrated axes may also be the most pedestrian-friendly and congested. Accordingly, the current routing logic is understood as a preliminary step toward aligning the model with the city's spatial structure. Future iterations will be based on a composite cost function that combines route length, observed or estimated crowd density, temporary obstruction risk, route clarity, and proximity to urban services.

Regarding positioning, mobile and web browser location data were treated as approximate data at the urban corridor level, rather than precise data at the meter or sidewalk level. This is due to the nature of Karbala's historical fabric, which includes narrow streets and alleyways where positioning signals may be affected by reflections, multiple paths, and the inherent inaccuracy of dense urban environments. Therefore, the model was not used in this study as a critical safety tool or a precise emergency routing system, but rather as an aid layer for general spatial awareness, providing approximate guidance on directions, routes, and service locations. A low-pass filter was also used to reduce fluctuations in device orientation data and improve the stability of signals displayed to the user. Table 2 shows the data sources used in the study, including the street network, spatial integrity values, approximate positioning data, device orientation data, field observation records, pre- and post-tests, and a user acceptance questionnaire.

Table 2. Inventories the complete dataset

The study employed a purposive sample of 50 participants, selected from visitor categories most likely to experience spatial awareness issues during their visit, particularly first-time visitors, repeat visitors with varying degrees of spatial familiarity, and visitors experiencing high crowd pressure. Since the study did not employ a random probability sampling method, the adequacy of the sample was not justified by random selection equations, but rather by the nature of the quasi-experimental exploratory design based on pre-/post measurement, and by the suitability of the purposive sample for the objectives of the field experiment. In this design, each participant acts as a self-comparison before and after using the model, which helps to minimize the effect of individual differences among participants. Therefore, the sample here is understood as suitable for testing the indicators of suitability and usability in a preliminary field study, not for the purpose of broad statistical generalization to all visitors to Karbala.

The field experiment was further supported by an independent survey of 60 academic experts in architecture, urban design, AR technologies, heritage preservation, and crowd management. This survey was used to support the apparent validity, content suitability, professional acceptability, and future scalability of the model, not to prove its effectiveness in actual user performance. Therefore, the expert evaluation complements the field experiment and is not a substitute for a controlled pilot test involving a control group or a comparison with traditional navigation tools.

The field experiment was structured around three spatial scenarios derived from actual movement patterns within the historic center of Karbala during the pilgrimage season. The first scenario represented a visitor struggling to form a clear mental map within a relatively crowded area. The second scenario represented a repeat visitor with prior knowledge of the location but experiencing greater crowd pressure and requiring support in spatial decision-making. The third scenario represented confusion or hesitation at visually similar intersections and passageways. These scenarios were designed to test four key dimensions: spatial perception and mind mapping, navigational efficiency, crowd awareness and rerouting, and access to urban services. Table 3 details the field scenarios in terms of location, participant type, observed crowd pressure, main perceptual challenge, and the tested model function. Figure 6 illustrates a field application of the model on Republic Street during peak visiting season.

Table 3. Field scenario specifications for the exploratory pre-test/post-test assessment

(a)

(b)

(c)

Data were collected using a 14-item Likert scale questionnaire administered to participants in two phases: before and immediately after using the model. This allowed for measuring the change in participants' responses between the pre- and post-tests for each dimension of the study. User acceptance data was also collected through a post-test questionnaire focusing on ease of use, clarity of information, speed of response, support for mind mapping, the model's impact on the spiritual and spatial experience, and the willingness to recommend it. These results should not be interpreted as definitive causal evidence due to the absence of a control group, but rather as preliminary indicators of the model's ability to support spatial perception and crowd awareness in a crowded field setting.

The statistical analysis began by examining the nature of the data distribution using the Shapiro-Wilk test to determine the appropriate test for each study axis. Based on the distribution analysis results, the Wilcoxon signed-rank test for paired samples was used for axes where the normality assumption was not met, with the effect size calculated using r = Z/√N. For axes that met the normality criteria, the paired samples t-test was analyzed, and the effect size was calculated using Cohen's d. The study adopted a significance level of 0.05, with results considered highly significant at p < 0.001. The statistical procedures used were also explained, including the type of test, the nature of the data, the significance level, the method for calculating the effect size, and the approach to handling incomplete data. No missing data were present in the final dataset.

To enhance the reliability of the research procedure, a methodological triangle was adopted, linking the outputs of spatial analysis, field experience data, participant responses, and expert evaluation. However, the study clearly adheres to its methodological limitations; it does not claim to provide definitive causal proof of the model's effectiveness, nor does it claim complete operational accuracy for the positioning system, nor does it present a complete model for dynamic crowd rerouting. Rather, it presents an exploratory field model of a lightweight digital knowledge layer that can support spatial clarity and the visitor experience in religious heritage cities, with the need for further development that includes a control group, more accurate crowd density measurement, integration of real-time crowd data, and more precise positioning techniques.

The results of this study provide preliminary field evidence regarding the usability of a web-based AR-enhanced digital NAI in supporting spatial perception and crowd awareness within the historic center of Karbala during the pilgrimage season. Pre- and post-test results showed a clear improvement in participants' responses after using the model, particularly in the area of crowd awareness and rerouting, which recorded the highest improvement among the study's axes. However, these results should be interpreted with caution, as the design is a quasi-experimental, single-group design and does not include a parallel control group or direct comparison with a traditional navigation tool. Therefore, the results do not establish a definitive causal relationship between model use and cognitive improvement, but rather suggest a potential effect that warrants further investigation in more controlled studies.

From a cognitive load theory perspective, the improvement in participants' responses can be interpreted as a possible consequence of the model's role in reducing the external cognitive load on visitors. In crowded urban religious environments, crowd movement, temporary elements, signage, processions, and services all contribute to making spatial awareness and the formation of a clear mental map more challenging. In this context, the digital layer can play a supporting role by simplifying and presenting spatial information directly, such as traffic directions, service locations, and suggested routes. However, this interpretation remains theoretical and supported by preliminary field indicators. It cannot be considered definitive empirical confirmation without controlled testing that isolates the model's effect from other factors, such as the novelty of the experience, participant expectations, or increased familiarity with the task after the pre-test.

City image theory and situational awareness theory also help explain why the crowd awareness and rerouting axis recorded the highest improvement. In congested conditions, visitors not only struggle to find their way but may also have difficulty understanding their surroundings and anticipating potential blockages or hazards. Therefore, providing simplified spatial indicators via smartphone can help support situational awareness, especially at intersections, similar passageways, or locations where directions are difficult to discern. However, this result should be understood as a perceived and measured improvement based on participant responses, not as conclusive operational evidence of the model's ability to manage crowds in real time.

The results also suggest that subtle digital interventions may be more suitable for religious heritage cities than direct construction interventions. In the case of Karbala, the historical fabric is part of the symbolic and spiritual value of the place, and any extensive physical intervention, such as widening streets or removing architectural elements, could affect this value. Therefore, the concept of a “digital cognitive layer” offers a more sensitive planning approach, as it seeks to enhance the reading of the place without altering its physical structure. User evaluations showed that the model was not perceived as a distraction from the spiritual experience, but rather as an aid. However, user acceptance alone is insufficient to prove effectiveness, as it measures perception and usability more than actual operational performance.

The results of the expert survey should be understood within their proper methodological context. The results demonstrated good professional acceptance of the model concept and supported its apparent validity, content relevance, and scalability. However, expert evaluation does not replace controlled field testing, nor does it provide direct evidence of the model's effectiveness with users. Therefore, the expert survey should be treated as a supporting and complementary level of evaluation, not as independent proof of the model's success. In this sense, the expert survey supports professional acceptance, apparent validity, and content relevance, but it does not compensate for the lack of empirical verification with users under controlled comparative conditions.

The study also reveals important technical limitations that must be considered. The model used is a web-based AR-enhanced digital NAI, not a full-fledged AR system based on precise spatial fixation, SLAM or VPS-based visual overlays, or fixed spatial anchors. Furthermore, reliance on browser-based positioning results in approximate positioning accuracy, especially in dense urban environments where positioning signals may be affected. Since positioning errors can be significant compared to the width of historical corridors, the study's findings should be understood in terms of supporting general perception and broad trends, not precise metric navigation or critical emergency management.

From a planning perspective, the current routing model should not be interpreted as a complete system for dynamic rerouting and crowd management. The current version relies on spatial integration values derived from Space Syntax analysis to favor the most morphologically distinct axes. While this is a useful step for enhancing spatial readability, it is insufficient on its own to avoid congestion. The most integrated axes may simultaneously be the most attractive to pedestrians and the most congested. Therefore, future versions should aim to develop a composite routing function that combines route length, observed or estimated crowd density, temporary obstructions, proximity of services, and route morphological clarity. This would achieve a better balance between improving spatial readability and mitigating crowd pressure, rather than relying solely on the T1024 value.

Accordingly, the significance of this study lies not in proving the ultimate effectiveness of an intelligent navigation system, but rather in presenting a scalable exploratory model that integrates morphological analysis of the urban fabric, subtle digital intervention, and user experience in a crowded religious and heritage environment. The results provide a preliminary basis for designing more rigorous subsequent studies, which could include a control group, comparisons with traditional maps or signage, direct measurement of crowd density, and testing of more precise positioning techniques. In this sense, the study offers a preliminary methodological and applied contribution, not a definitive judgment on the model's effectiveness in all visit conditions.