Artificial Intelligence-Based Intelligent Navigation System for Alleviating Traffic Congestion: A Case Study in Batam City, Indonesia

Urban traffic congestion is a complex issue that requires a multifaceted approach and the utilization of advanced technology [36]. This problem not only involves inefficient traffic management but also significant economic and social impacts, such as long travel times, air pollution, and increased fuel consumption. Various studies in the fields of transportation engineering, urban planning, and artificial intelligence (AI)-based systems have shown that traditional solutions are insufficient to address these challenges. Therefore, the integration of AI into traffic management systems emerges as a more efficient, adaptive, and responsive solution to the ever-changing dynamics of traffic [37].

2.1 The role of AI in traffic management

Artificial Intelligence has leveraged machine learning (ML) and deep learning algorithms to analyze large-scale transportation data, recognize traffic patterns, and provide accurate predictions. Various approaches have been applied to manage and predict traffic flow more efficiently. One significant early study was conducted by Luo et al. [38], who used support vector machines (SVM) to model urban traffic flow predictions. The results of this model showed that SVM could be used to predict traffic flow with a reasonably good level of accuracy, although it was still limited to relatively simple data [39].

However, further advancements in AI applications for traffic have shifted towards deep learning techniques, particularly the use of recurrent neural networks (RNNs) and the Long Short-Term Memory (LSTM) variant for time-series-based predictions. These techniques, developed by Shang et al. [40], are highly effective in handling temporal dependencies in traffic data, such as dynamically changing vehicle densities over time. The use of these techniques has become increasingly popular due to their ability to learn recurring patterns within the data, which is highly beneficial in the context of highly dynamic traffic conditions.

In parallel, reinforcement learning (RL) algorithms have also been applied extensively to optimize traffic signal control and vehicle routing. RL is a type of machine learning where an agent learns to make decisions by interacting with its environment and receiving feedback in the form of rewards or penalties. One well-known application of RL in traffic management is its use in optimizing traffic light control. Shang et al. [40] developed a deep reinforcement learning (DRL) framework that adjusts traffic light timings based on real-time detected vehicle density. This approach dynamically adapts the signal phases to match traffic conditions, reducing congestion and improving the flow of traffic at key intersections. By continuously learning from traffic data, the RL algorithm can determine the most efficient signal timings, which leads to reduced waiting times at intersections and a smoother overall traffic flow.

Furthermore, RL has been successfully implemented in vehicle routing. For instance, RL can be used to suggest optimal routes for drivers based on current traffic conditions, accidents, or road closures. The system continuously learns from the traffic dynamics, making real-time route recommendations that minimize travel time and reduce congestion. A popular approach involves the use of Q-learning, a reinforcement learning algorithm that helps optimize routes by evaluating the "quality" of each route option in a dynamic environment.

The benefits of implementing RL in real-world traffic systems are substantial. It offers adaptability by allowing the system to continuously improve and adjust based on changing traffic conditions. It can provide more responsive and efficient traffic management compared to traditional, fixed systems like static traffic light timings or pre-defined route maps. However, there are several challenges in implementing RL in real-world settings. Computational resources and data quality can be limiting factors, as RL algorithms require significant computational power and large datasets to train effectively. Moreover, the complexity of the traffic environment—with numerous variables such as unpredictable driver behavior, weather conditions, and varying traffic patterns—can make the application of RL algorithms challenging. Despite these challenges, the potential for RL to transform urban traffic management is immense, offering a more adaptive and efficient system for congestion reduction.

2.2 Navigation system and route optimization

Traditional GPS-based navigation systems, although effective in providing alternative routes, are often hindered by their inability to adjust routes in real-time to the continuously changing traffic conditions. These systems rely on static maps and heuristic-based routing methods that cannot adapt well to sudden traffic congestion. Therefore, various modern intelligent navigation systems integrate real-time data and adaptive algorithms to provide more effective route suggestions [41].

One of the widely used techniques for route optimization is Dijkstra’s algorithm, which remains a reliable method for finding the shortest path. However, in the context of constantly changing traffic, this algorithm needs to be combined with heuristics that account for traffic congestion levels. Bernabei and Secchi [42] suggested that for real-time navigation, Dijkstra’s algorithm can be modified to incorporate traffic-aware heuristics, allowing the system to respond to changes in traffic flow more quickly and accurately.

Furthermore, in the study [43], a hybrid model was developed that combines historical data with real-time feedback to update route recommendations in response to fluctuations in road conditions. This model not only reduces travel time but also enhances the user experience. This concept has become increasingly important as navigation systems in most large cities already use real-time data to respond to traffic changes. However, medium-sized cities like Batam often lag behind in the application of this technology.

2.3 Concept of smart cities and urban mobility

The development of smart cities is a broader concept than just transportation technology. Smart cities focus on the use of information and communication technology (ICT) to improve the efficiency of urban services, one of which is the transportation system. Smart mobility is one of the main pillars of this concept, involving the use of analytics, the Internet of Things (IoT), and AI to enhance the efficiency, safety, and sustainability of urban transportation systems [44].

In recent years, several major cities worldwide have successfully implemented AI-based traffic systems as part of their smart city initiatives. Singapore, for example, has developed an AI-driven traffic management system that integrates real-time data from sensors and cameras with predictive analytics. This system optimizes traffic light timings, adjusts the flow of vehicles, and helps alleviate congestion. One notable example is the Smart Traffic Management System used in Singapore’s Central Business District (CBD), which has significantly reduced traffic delays and improved the flow of vehicles during peak hours. Moreover, it has been shown to reduce fuel consumption and air pollution by minimizing idling times at traffic signals [45].

Similarly, Barcelona has adopted a smart mobility approach where AI-based systems are integrated with urban infrastructure to monitor traffic flow, provide real-time information to drivers, and suggest the best routes to avoid congestion. The city uses AI-powered sensors to detect congestion patterns and adjust traffic light timings accordingly. These innovations have led to a 20% reduction in travel time in highly congested areas, and Barcelona’s smart city initiative has also resulted in a 15% decrease in carbon emissions by optimizing vehicle movement and reducing idle times [45]. Furthermore, these systems have enhanced safety by reducing the likelihood of traffic accidents through real-time alerts and adaptive signal adjustments.

These examples demonstrate the power of AI in optimizing traffic flow, improving safety, and contributing to sustainability goals by reducing traffic congestion and lowering emissions. The integration of AI into urban transportation systems not only makes cities more efficient but also significantly improves the quality of life by providing smoother commutes, cleaner air, and safer roads.

However, the implementation of these smart systems in developing cities, including those in Indonesia, faces distinct challenges. These include incomplete data availability, inadequate infrastructure, and limitations in collaboration among stakeholders. In Indonesia, particularly in Batam, the research and development of AI-based navigation systems for traffic management is still relatively limited, with most studies being focused on large cities or capital cities with more advanced infrastructure and resources. Therefore, the development of an intelligent navigation system tailored to local conditions, such as the one developed in this study, holds high strategic value. This system could help Batam and other medium-sized cities improve their traffic management and urban mobility, aligning them with the global trend toward smart cities [43, 46].

2.4 Research gaps

Existing research has made significant contributions to the development of AI-based traffic prediction systems and route optimization. However, the majority of these studies focus on large cities with sufficient infrastructure and budgets. Meanwhile, medium-sized cities like Batam, which are experiencing rapid urbanization, often do not receive enough attention in similar research. The existing AI-based traffic management systems are not adaptive enough for cities with unique characteristics such as Batam, which has more limited infrastructure and data that is not always complete, as illustrated in Figure 1.

1.png

Figure 1. Research methodology flow

Moreover, although many AI-based navigation systems have been successfully implemented in large cities, there is still a lack of research developing systems that can integrate real-time traffic predictions with navigation interfaces that respond to changing traffic conditions. This study aims to fill this gap by designing an AI-based intelligent navigation system that provides solutions for traffic problems in Batam. This system not only utilizes machine learning for traffic prediction but also integrates real-time data to respond to continuously changing traffic conditions.

This section presents the results of data collection and processing obtained from primary and secondary data collected from the Batam City Transportation Department, as well as surveys and interviews with road users. The data provides a clearer picture of the causes of traffic congestion, patterns of congestion occurring at various locations, and the solutions that have been implemented to reduce congestion.

4.1 Data collection

1. Population in Bata

The graph above depicts the population growth of Batam City from 2019 to 2024. The data demonstrates a consistent annual increase in the population, reflecting a stable upward trend in the city's demographic expansion.

In 2019, the population of Batam was recorded at 1,107,551 individuals. Figure 3 rose to 1,196,396 in 2020, indicating a notable increase. The upward trajectory continued in subsequent years, with the population reaching 1,207,082 in 2021, 1,240,792 in 2022, and 1,260,785 in 2023. Projections suggest that by 2024, the population will reach 1,294,548, further emphasizing the sustained growth trend.

image005.png

Figure 3. Population in Batam 2019-2024 (sources BPS)

This population increase can be attributed to various factors, including urbanization, migration, and the region's economic opportunities, which likely contribute to the attraction of new residents. The steady population growth presents both opportunities and challenges for the city, particularly in terms of the need for enhanced infrastructure and the provision of public services to accommodate the expanding population.

2. Vihicles on Batam

As of April 2025, the total number of motor vehicles in Batam, Riau Archipelago, has reached 1.09 million units. This data, sourced from the Electronic Registration and Identification (ERI) system of the Traffic Corps of the Indonesian National Police (Korlantas Polri), was published on April 16, 2025. Of these, motorcycles accounted for 880,710 units.

In addition to motorcycles, passenger cars were recorded at 179,410 units, while cargo vehicles (freight trucks) totaled 27,930 units. Furthermore, the data indicates that Batam has 1,786 buses and 207 special vehicles, which include utility and specialized vehicles.

When compared to other regions in the Riau Archipelago, Tanjung Pinang, the provincial capital, reported the second-largest number of motor vehicles, with 205,150 units. In contrast, the Kabupaten of Kepulauan Anambas reported the lowest vehicle count, with only 4,383 units.

The following table summarizes the distribution of motor vehicles in Batam as of April 16, 2025, based on the ERI Korlantas Polri data:

Motorcycles: 880,710 units

Passenger Cars: 179,410 units

Cargo Vehicles: 27,930 units

Buses: 1,786 units

Special Vehicles: 207 units

A. Historical Traffic Data

Based on data obtained from the Batam City Transportation Department (January-December 2024), as well as results from surveys and interviews with road users, several congestion hotspots have been identified as key areas of concern in Batam City.

Table 1 shows historical traffic data obtained from the Batam City Transportation Department, which includes major congestion points, peak hours, vehicle speeds, and the causes of congestion at several strategic locations. Based on this data, it can be observed that congestion primarily occurs during peak hours, namely in the morning and evening, coinciding with the times people commute to and from work. The recorded vehicle speeds at congestion points are generally low, averaging between 20 and 45 km/h, indicating that the limited road capacity is unable to accommodate the increasing vehicle volume.

Additionally, Figure 4 provides a deeper insight into the fluctuations in vehicle density at various major locations in Batam City over time. This graph shows peak congestion occurring between 17:00 and 18:00, especially at points like Batam Center and Sekupang, which are the areas with the highest vehicle density. This data reinforces the urgent need for more efficient traffic management solutions that can respond to traffic dynamics in real-time, such as those offered by AI-based intelligent navigation systems.

image006.png

Figure 4. Vehicles in Batam 2025 April (sources data box)

Several areas in Batam, particularly in the city center and border regions, have roads that are not wide enough to accommodate the increasing number of vehicles. Additionally, the poor maintenance of roads, as well as road narrowing at certain points, also contribute to congestion. Inadequate road infrastructure plays a significant role in the traffic congestion that occurs.

Based on Figure 5, it can be seen that at several locations such as Sekupang, the vehicle volume (90 vehicles/km) exceeds the available road capacity (81.82 vehicles/km), indicating significant congestion. In other locations, such as Batam Center and Piayu, the vehicle volume is close to the road capacity, suggesting that the road capacity is nearly surpassed by the existing traffic demand. This reinforces the need for more effective traffic management solutions to address the mismatch between road capacity and vehicle volume at various congestion points.

image007.png

Figure 5. Vehicle volume chart based on time in Batam City

image008.png

Figure 6. Comparison chart of road capacity vs vehicle volume

As shown in Figure 6, the comparison between road capacity and vehicle volume indicates a significant mismatch, especially in several urban zones of Batam.

B. Real-Time Data

Data is collected through the integration of GPS technology, IoT, and APIs from navigation platforms, as shown in Table 2.

Table 2. Real-time data for the intelligent navigation system in Batam City

C. Topographic and Road Network Data

Traffic and Road Network Data in Batam City The topographic and road network data in Batam City was obtained through GIS-based map digitization (Geographic Information System), strengthened with direct field observations, as shown in Table 3.

Table 3. Topographic and road network data in Batam City

D. User Behavior and Preference Data (Survey)

As part of the effort to develop an AI-based intelligent navigation system for Batam City, a survey was conducted with 500 private vehicle users and online motorcycle taxi drivers. Data collection was carried out through two methods: online forms and direct interviews, in order to gain a more comprehensive understanding of navigation app usage patterns and user expectations for a more adaptive system.

image009.png

Figure 7. Survey result: Navigation app usage and perception in Batam City

The survey respondents ranged in age from 17 to 55 years old, with a professional background primarily consisting of private-sector employees, students, and online motorcycle taxi drivers. The survey explored their experiences using digital navigation apps, their perceptions of system accuracy, and their willingness to switch to local AI-based technology if proven more efficient, as shown in Figure 7.

4.2 Data calculation and analysis

A. Traffic density analysis

Traffic density analysis is conducted to identify major congestion points and measure their impact on travel efficiency in Batam City. Vehicle density data is collected from the Transportation Department, traffic sensors, and GPS, and is then calculated based on vehicle volume and road capacity. The results of this analysis provide an overview of the vehicle distribution on main roads and alternative routes during peak hours, which will be used to plan a more efficient intelligent navigation system. To measure traffic density, the following formula is used:

K=Q/V

where,

K=Density

Q=Flow rate (number of vehicles per hour)

V=Average speed (km/hour)

Table 4. Kepadatan lalu lintas pada pukul 07:00-08:00

Table 5. Traffic density between 07:00 and 08:00

Table 6. Traffic density between 16:00 and 17:00

Table 7. Traffic density between 17:00 and 18:00

Table 4, Table 5, Table 6, Table 7 present a summary of traffic density calculations at several key locations at different times.

B. Speed and travel time analysis

In this section, an analysis is conducted on vehicle speed and travel time at several major congestion points in Batam City during peak hours. Travel time is calculated using the following formula:

T=S/V

where,

T=Travel time (in hours)

V=Vehicle speed (in km/h)

S=Distance traveled (in km)

Table 8. Speed and travel time analysis between 07:00 and 08:00

Table 9. Speed and travel time analysis between 08:00 and 09:00

Table 10. Speed and travel time analysis between 16:00 and 17:00

This analysis aims to provide an overview of how traffic conditions affect travel duration at various congestion points, both in the morning, afternoon, and evening. By examining the recorded vehicle speed and travel time, we can identify locations with high congestion and assess the impact on transportation efficiency in Batam. The travel time calculations at each congestion point will be displayed in Table 8, Table 9, Table 10, Table 11, which shows how vehicle density and speed affect travel duration during peak hours.

Table 11. Speed and travel time analysis between 17:00 and 18:00

C. Traffic prediction model using LSTM

In analyzing the performance of the traffic prediction model based on Long Short-Term Memory (LSTM), we test the model's ability to predict two key parameters in traffic analysis: travel time and vehicle density.

Figure 8 compares actual travel time with the predicted travel time forecasted by the LSTM model at several locations in Batam City. In this graph, the blue line represents the actual recorded travel time on the ground, while the red line illustrates the predicted travel time estimated by the system. In general, it can be seen that the prediction system shows fairly accurate results, with some small differences between actual and predicted times, particularly in Batam Center and Tanjung Uncang, where the differences are minimal.

Locations such as Punggur and Nongsa show predicted travel times that closely align with actual times, while at other locations like Sekupang, there is a slightly larger difference. This indicates that, although the LSTM model is capable of providing relatively accurate predictions, several factors affect the accuracy of predictions, such as dynamic traffic conditions and other external factors.

Figure 8 illustrates the importance of implementing an LSTM-based navigation system that can predict travel times in real-time, providing more accurate information for drivers to select optimal routes and reduce travel time.

image010.png

Figure 8. Travel time vs prediction graph

In Figure 9, the blue line represents the actual traffic density recorded during the observation, while the red line illustrates the predicted traffic density estimated by the AI system. Overall, the system's prediction results tend to be quite accurate, with minimal differences between actual and predicted density values, especially at locations such as Batam Center, Piayu, and Sekupang.

image011.png

Figure 9. Comparison graph of actual vs predicted density

However, at some points like Punggur and Nongsa, there is a slight difference between the prediction and actual data, indicating that although the traffic density predictions are fairly good, external factors affecting traffic (such as accidents, weather, and sudden changes in vehicle volume) still influence the accuracy of the predictions.

Figure 9 illustrates the importance of implementing an AI-based navigation system that can predict traffic density in real-time. This not only helps drivers choose more optimal routes but also provides better insights for infrastructure planning and traffic management in large cities like Batam.

D. AI-Based intelligent navigation system simulation

The simulation results show that the AI-based navigation system performs better compared to Google Maps. The average travel time with the AI system is faster, at 14.2 minutes, compared to 18.4 minutes with Google Maps, resulting in a 22.8% reduction in average travel time. Additionally, the AI system is more adaptive in rerouting, with a rerouting frequency of 3.6 times compared to 1.2 times on Google Maps. The rerouting time is also faster, with the AI system requiring less than 5 seconds, while Google Maps takes about 14 seconds to reroute. In terms of user satisfaction, the AI system received a score of 4.5 (on a scale of 1-5), significantly higher than Google Maps, which received a score of 3.7. These results indicate that the AI-based navigation system is not only more efficient in terms of travel time but also more responsive and preferred by users, as shown in Table 12.

Table 12. Comparison of ai-based intelligent navigation system and google maps performance in Batam