Improving Cyclists’ Safety Using Intelligent Situational Awareness System

Abstract

According to the World Health Organization (WHO), over 1.35 million people died in road traffic-related accidents worldwide in 2020 of which 41,000 are related to the cyclists. Bike safety is one of the most serious issues facing urban riders. According to Statistics Canada, this number represents 1654 cyclist deaths in Canada, an average of 74 deaths each year from 2006 to 2017. Cyclists are a critical component in traffic collisions, where they face a greater risk of serious injury or death. As a result, they are classified as vulnerable road users. To avoid this, the need for intelligent transportation systems (ITSs) that increase susceptible cyclists’ awareness of their surroundings is becoming apparent. As a result, we proposed a situational awareness system as part of ITS to enhance bike safety through the employment of three layers of applications: (1) the users tier (mobile application), (2) the virtual private server (VPS) and processing system, and (3) the database management system (DMS). These decision support systems (DSSs) improve vulnerable road users’ situational awareness by identifying high-risk regions for cyclists or motorcyclists using static and dynamic data and then notifying vulnerable road users. The suggested situational awareness system collects and integrates incoming data, prioritizes criteria, and notifies users based on a static hot-spot map produced from accident locations and dynamic data, such as traffic flow, weather conditions, and the user’s speed. The developed work made use of both single threading (for requests from less than 1000 users) and multi-threading (for requests from more than 1000 users), resulting in a highly scalable system based on an open source platform for higher numbers of requests.

1. Introduction

According to a research by the World Health Organization (WHO) [1], around 1.35 million people died in vehicle accidents worldwide in 2020; in other words, one person dies in a car accident every 25 s. In Toronto, the proportion of cyclists who ride for recreation or fitness has remained stable over the past decade (26% in 2019, 25% in 2009). Nevertheless, the percentage of cyclists who ride to work, school, go shopping, or visit friends has increased from 29% in 2009 to 44% in 2019 [2]. Bike and motorcycle safety is one of the most serious issues confronting bikes in urban transportation [3]. Cyclists who are engaged in traffic accidents face a significantly increased risk of serious injury or death [4]. As a result, they are classified as vulnerable road users. Bjornskau et al. [5] established that cyclists have a five fold greater accident rate than other road users. As a result, the actual number of motorcycle accidents is likely to be higher than the published report. Taking into account unreported bike incidents, it is estimated that bikers have a collision rate nearly 20 times that of car users. Although bikers have a larger risk of injury in traffic than automobile occupants [6], there are around 800 million bikes on the planet, more than twice the number of cars. As a result, increased attention to these populations’ transportation needs is required. Additionally, an increase in accidents has a detrimental effect on people’s lives and social development.

In both developing and developed countries, bike accidents occur primarily on urban roads with a high cycling rate and large traffic volumes [7]. Additionally, it is critical to identify bike accident hot zones in order to better understand the causes of bike accidents and to improve cycle safety [8]. Indeed, the majority of applications and research today are connected in some way to the smart city [9,10]. The smart city encompasses a variety of study areas, including smart transportation [11], smart buildings [12], and smart health [13]. On the other hand, the systems utilized in the development of intelligent transportation are referred to as intelligent transportation systems (ITSs) [14]. ITS has two primary purposes: to promote user safety and convenience [15]. In the ITS, the term “users” refers to passengers, drivers, pedestrians, cyclists, and motorcycle riders. Biking is gaining popularity in many places in Canada as more individuals become aware of the possibilities of cycling to alleviate traffic congestion, enhance air quality, and improve personal health. The growing demand for safe and convenient bike facilities to support cyclists is resulting in an increase in biking for transportation and recreation. The need is extremely high in places with a high level of engagement and conflict with motor vehicles. Nonetheless, little research has been conducted to determine whether and how ITS and geographic information systems (GISs) design can lessen confrontations between cars and bikes, hence increasing safety [16].

ITS can affect the safety of road users in three modes. The first mode is concerned with systems and platforms that are designed to prevent accidents [17]. The second mode is concerned with the actions taken during the accident, while the third mode is concerned with the systems that begin immediately following the event [18]. The suggested system is connected to smart transportation services that can act in front of incidents. Our study is focused on enhancing cyclist safety. In general, we hope to promote cycling safety by analyzing dynamic GIS data and then alerting cyclists to unsafe circumstances. Four primary parameters are evaluated when sending a warning to the users: (1) the user’s velocity, (2) historical information, (3) weather conditions, (4) and traffic congestion. This research makes use of three dynamic data sets and one static data set. Static data are historical information that can be utilized to identify hot-spots or black-spots. As with other types of traffic incidents, bike accidents have location (geocoded) data. As a result, numerous research works have been conducted on the topic of traffic accident analysis utilizing GIS. Such methods enable the visual representation of traffic incidents and the spatial analysis of them in order to detect hot-spots.

Numerous methodologies can be utilized to identify traffic accident hot-spots, utilizing spatial point pattern analysis (PPA) techniques based on GIS [19]. It is critical to identify accident hot-spots in order to properly deploy resources and improve road safety. However, once established, hot zones can vary with time and environmental conditions. Determining bike accident hot-spots on road segments by location and time allows for a more accurate assessment of cyclist safety since it enables the road network’s problematic portions to be prioritized. Once we have the hot-spot map, we can connect the static data (user speed, traffic congestion, and weather condition) to the dynamic data (user speed, traffic congestion, and weather condition) (Figure 1).

2. Problem Definition

As previously stated, the accident rate for bikers and motorcyclists is increasingly high. These accidents result in irreversible economic and societal devastation. Accidents do not occur randomly in transportation networks, and certain places dubbed hot-spots or black-spots are becoming more prevalent. After analyzing the black-spots, it was determined that human error, road construction, and the driver’s field of view are the primary causes of accidents. Occasionally, bikes and motorcyclists are involved in accidents as a result of negligence. This essay has two primary objectives. To begin, spatio-temporal data should be analyzed to determine Toronto’s hot-spots. The analytical hierarchy processing (AHP) technique is being utilized to identify these locations. Second, once the hot-spot maps are constructed, an intelligent system that integrates static and dynamic data must be developed to recognize harmful circumstances. Numerous variables contribute to the occurrence of accidents. In this paper, the probability of accidents in a variety of circumstances is calculated using four variables. The first factor is the historical data of accidents in a specific area which affect the potential of occurring accidents. The user’s speed is the second factor that can directly influence the occurrence of an accident. The third factor is traffic congestion. Traffic congestion can affect the risk of bike accidents. The main reason is because of the high density of traffic, not only because of traffic congestion. The fourth factor is the weather, including rain, snow, ice, and strong winds. Wet roads can make it more difficult for bicycles to maintain traction, increasing the likelihood of falls. Snow and ice can also reduce visibility and make roads slippery. Strong winds make it more difficult for a cyclist to maintain control of his or her bicycle. Additionally, reduced visibility due to dense fog or darkness can make it more difficult for motorists to see cyclists. The dataset may be regarded as active data, such as the speed of the user, or static data, such as previous accident datasets. Numerous concerns exist about hot-spot detection. However, the primary objective of this project is to increase cyclists’ awareness of the need to be more cautious on the road by merging Web GIS with ITS.

For intelligent risk identification and accurate dynamic analysis, an intelligent assistance system consists of front-end and back-end components. The device is intended to alert bikers when they are at risk of an accident. To begin, we conducted an analysis in ArcGIS to determine the accident risk associated with each road. Following that, a geodatabase was established. Additionally, we changed the ArcGIS server to dynamically assess the hazard risk based on the user’s location.

The mobile application is designed to dynamically determine the user’s position. This application communicates with the server in order to determine the probability of an accident based on past data for that route, as well as the speed and other aspects that may affect this model, such as weather conditions and traffic congestion.

3. Related Works

The state-of-the-art car-safety system has been described in many previous works [20,21,22,23]. To be specific, ITS safety systems have received much attention over the last years [18,24,25,26].

When we discuss ITS components, one critical topic is their communication modes. These networks’ communication performance has been evaluated and compared to that of other wireless technologies, such as IEEE 802.11p-based DSRC. Refs. [27,28], which demonstrates the great potential of this technology in safety applications.

Most previous works considered position, direction [29,30], and speed [31] to predict potential collision between cars and vulnerable users. However, in [32,33], the limitations of this information for collision detection have been described.

Numerous factors contribute to the occurrence of accidents, including texting [34], watching a video, conversing on the phone, stopping, weather conditions [35], the time of day, walking, waiting, jogging, or crossing a curb. It is a little difficult to assert that a single system can account for all possible causes and prevent accidents [36,37].

More precisely, the authors proposed in [33] a method for utilizing the pedestrian context of crossing a curb and stepping onto a road in order to increase both positioning accuracy and collision detection likelihood. Numerous techniques have been proposed for the collision detection algorithm based on the exchange of context information between automobiles and vulnerable humans [38,39].

One of the main reasons people hesitate to use bikes is related to safety matters. Department of Transportation in Charlotte [40] conducted interesting research to analyze users’ behavior regarding their interest in using bikes. They found that 51% of the people living in Charlotte are interested in bikes. However, they are afraid of riding a bike in the city because of safety. Preparing a safe environment for cycling can reduce traffic congestion [41,42] and help with energy consumption [43]. Moreover, it delivers a healthier and greener Earth [44]. Hence, improving cyclists’ safety must be addressed.

4. Methodology

The suggested system for bike safety is based on the (centralized) architecture represented in Figure 2. Throughout the system, communication is established via REST-Full APIs (application programming interfaces). The next sections describe our system architecture, which includes some components that are processed locally and others that are offloaded to the cloud using the Mean-Stack (MongoDB, Express, Angular, Node.js) platform. As illustrated in Figure 2, the system is composed of three primary entities: (1) the user (mobile application), (2) virtual private server (VPS), (3) and the information system. The fundamental concept is to design a decision-making system capable of detecting harmful conditions for cyclists or motorcyclists through the use of static and dynamic data and then notifying susceptible users. The decision-maker notifies users based on four primary data points. One of these data points is static, while the remaining three are dynamic. The static data are the hot-spot map, which is created by analyzing previous Toronto accidents. Traffic congestion, weather conditions, and the user’s speed are three dynamic data points. To be more explicit, the next section describes the process of creating a hot-spot map and how the decision maker operates.

To begin, the database is populated with a map of black-spot (hot-spot) locations. These data are maintained throughout the decision-making system’s procedures. The following stage is to develop a mobile application that can be loaded on the user’s phone, and when the user is riding, we can obtain their location (latitude, longitude) using GPS. The server is notified of the users’ current location. When the server receives the user’s data, it begins integrating them with the hot-spot areas to determine if the user is in a dangerous area or not. The preceding step’s data are then combined with real-time weather conditions, traffic flows, and the user’s speed. The AHP method is used to weight the layers in order to find hot-spots (car accident layer, bike accident layer, and motorcycle accident layer). Additionally, fuzzy reasoning and the Mamdani technique [45] are used to integrate diverse types of data.

As illustrated in Figure 3, the hot-spot map is created by combining three different historical accident layers. Due to the fact that this project’s objective is more closely tied to bikes and motorcycles, the bike and motorcycle accident layers are more relevant than the automobile accident layer. Additionally, the motorcycle accident layer is more critical than the bike accident layer. With this information in mind, the ratios for initializing values in the AHP approach are as follows: bike/car ratio is 4, motorcycle/car ratio is 4/3, and bike/motorcycle ratio is 3. The final hot-spot area map is generated using the ArcGIS model builder. Once we have the map, we may begin fusing datasets. The mobile application can calculate the user’s speed. Multiple rules govern the mobile application used in this study, including tracking users, determining their speed, connecting to the server to submit data, and receiving messages from the server. To obtain current weather conditions, a web service API connecting to the open weather map website is built. However, for each user, the system must obtain the current weather conditions in the area in which the user is located. As a result, an API is developed that takes the user’s latitude and longitude into account and then returns the weather conditions for that location, identical to how real-time traffic congestion was obtained from the (https://developer.here.com/, accessed on 10 October 2022) website. This way, we now have access to the user’s speed, the weather conditions in the user’s location, traffic congestion in the user’s location, and the number of accidents in that area at any given time. The following step is to combine the data using fuzzy logic (Mamdani method). Since a decision should be made based on the state of each criterion, and the states of the criteria are vague concepts, fuzzy logic is used. The inputs of fuzzy reasoning should be fuzzy set values. For each criterion, two linguistic values are defined to create fuzzy membership functions as follows: (low, high) for collision, (light, heavy) for traffic flow, and (low, fast) for user speed (Figure 4). Different types of functions are available to be used as membership functions, such as singleton, triangular, trapezoid, sigmoid, and Gaussian. Each of these functions has various characteristics, and one or more of them should be used based on the application. In this paper, the trapezoid function is selected because it is fast in terms of computational complexity and is suitable for our real-time model [46]. Each member of a set X (for instance, the user’s speed) is mapped to the Y axis, which is a value between 0 and 1. Y is called truth, degree of membership, or membership value. In other words, Y quantifies the grade of members in X to a fuzzy set A (Equation (1)).

$${\mu }_A:X\to Y=[0,1]$$

(1)

where ${\mu }_A$ is the trapezoid membership function.

Four parameters define the trapezoidal membership function: $x1$, $x2$, $x3$, and $x4$ (Figure 5-left). The range between $x2$ to $x3$ represents the maximum membership value that can be assigned to a member. If x is within ($x1$, $x2$) or ($x3$, $x4$), its membership value will be between 0 and 1. Equation (2) defines the membership function of a trapezoidal (Figure 5-left).

$${\mu }_{trapeoidal}(x;x1,x2,x3,x4)=\left\{\begin{array}{cc}0,\hfill & x<x1\hfill \\ \frac{x-x1}{x2-x1},\hfill & x1\le x<x2\hfill \\ 1,\hfill & x2\le x<x3\hfill \\ \frac{x4-x}{x4-x3},\hfill & x3\le x<x4\hfill \\ 0,\hfill & x4\le x\hfill \end{array}\right.$$

(2)

There are two different types of trapezoidal functions based on function openness. They are designated by the terms right-function or open-right and left-function or left-open. Figure 5-right depicts these two types. The blue one is a left-function, and the red one is a right-function. As it is mentioned before, for each criterion, two linguistic values are defined to create membership functions. One linguistic value is defined as a left-function and the other is defined as a right-function. Equations (3) and (4) define the membership functions for left-function and right-function (Figure 5-right).

$${\mu }_{left}(x;xr1,xr2)=\left\{\begin{array}{cc}1,\hfill & x<xr1\hfill \\ \frac{x-xr2}{xr1-xr2},\hfill & xr1\le x<xr2\hfill \\ 0,\hfill & xr2\le x\hfill \end{array}\right.$$

(3)

$${\mu }_{right}(x;xb1,xb2)=\left\{\begin{array}{cc}0,\hfill & x<xb1\hfill \\ \frac{x-xb1}{xb2-xb1},\hfill & xb1\le x<xb2\hfill \\ 1,\hfill & xb2\le x\hfill \end{array}\right.$$

(4)

For each criterion, these two equations (Equations (3) and (4)) should be defined.

To determine the parameters of trapezoidal membership functions two potential approaches are available: (1) expert knowledge and (2) real dataset. However, acquiring expert knowledge is a challenging task. In this paper, the parameters are defined based on the statistics of the real dataset, in which for each membership function, a clustering algorithm (k-means) is defined, and the upper bound and lower bound values of the clusters are considered the values of the membership functions [47]. It is also worth mentioning that parameter tuning is required to find the best values, especially for decision-maker systems [48,49].

Equations (5) and (6) are membership functions of the traffic flow. Equations (7) and (8) are membership functions of the accidents/area. Equations (9) and (10) are the membership functions of the user’s speed. The ranges in the membership functions are defined based on the statistics in the dataset. For instance, the minimum value in the user speed to be considered as high is equal to 30 km/h; however, for collisions, it is equal to 240 collisions/area. The membership function parameters for the user’s speed and traffic flow indicator can be generalized and is always the same for different case studies. However, the values for collision per area should be updated based on the location, as in other cities, the average of total collisions might be different.

Additionally, the web service API is included (available from the result area), which can be tested using the inputs desired. The Mamdani method’s rules for determining the result are listed in

Table 1. The rules that are considered for fuzzy reasoning, Mamdani.

Rule 1	if speed is high and collision is high and traffic flow is light => send alarm
Rule 2	if speed is high and collision is low and traffic flow is heavy => send alarm
Rule 3	if speed is low and collision is low => don’t send
Rule 4	if speed is low and collision is high and traffic flow is light => send alarm
Rule 5	if speed is low and collision is high and traffic flow is heavy => don’t send alarm

.

Another essential aspect of this research is the manner in which the server communicates with the mobile application. After the decision maker determines to send the warning to the users, the server will transmit the alert using Google Cloud Messaging (GCM) or, more recently, Firebase Cloud Messaging (FCM). FCM is a messaging platform that enables us to send free messages. We can use FCM to send a new message or other data to a client application. In some circumstances, such as instant alert, data can be transferred to a client app with a payload of up to 4KB (Figure 6). When an alarm is given to the user, as you could assume, if the alarm appears on the screen, it may cause the user to become agitated. To avoid any disruption, we created a text-to-speech converter. This converter for Android was created using Google’s TTS (text to speech) API (Figure 7).

5. Implementation and Results

The parameters for the membership functions are calculated based on Equations (3) and (4) and the results are represented in Equations (5)–(10).

$${\mu }_{left\_traffic}(x;xr1,xr2)=\left\{\begin{array}{cc}1,\hfill & x<4\hfill \\ \frac{8-x}{4},\hfill & 4\le x<8\hfill \\ 0,\hfill & 8\le x\hfill \end{array}\right.$$

(5)

$${\mu }_{right\_traffic}(x;xb1,xb2)=\left\{\begin{array}{cc}0,\hfill & x<5\hfill \\ \frac{x-5}{4},\hfill & 5\le x<9\hfill \\ 1,\hfill & 9\le x\hfill \end{array}\right.$$

(6)

$${\mu }_{left\_collision}(x;xr1,xr2)=\left\{\begin{array}{cc}1,\hfill & x<40\hfill \\ \frac{140-x}{100},\hfill & 40\le x<140\hfill \\ 0,\hfill & 140\le x\hfill \end{array}\right.$$

(7)

$${\mu }_{right\_collision}(x;xb1,xb2)=\left\{\begin{array}{cc}0,\hfill & x<40\hfill \\ \frac{x-40}{200},\hfill & 40\le x<240\hfill \\ 1,\hfill & 40\le x\hfill \end{array}\right.$$

(8)

$${\mu }_{left\_speed}(x;xr1,xr2)=\left\{\begin{array}{cc}1,\hfill & x<10\hfill \\ \frac{x-20}{-10},\hfill & 10\le x<20\hfill \\ 0,\hfill & 20\le x\hfill \end{array}\right.$$

(9)

$${\mu }_{right\_speed}(x;xb1,xb2)=\left\{\begin{array}{cc}0,\hfill & x<10\hfill \\ \frac{x-10}{20},\hfill & 10\le x<30\hfill \\ 1,\hfill & 30\le x\hfill \end{array}\right.$$

(10)

Table 2. Data sources that are used to develop this project.

Source	Data	Name
Data.torontopolice.on.ca	2018	Cyclist Accident
Data.torontopolice.on.ca	2018	Car Accident
Toronto Public Safety Data	2018	Motorcycle Accident
Traffic Developer Website	Real-Time	Traffic Data
Open Weather Map	Real-Time	Weather Data
Mobile Application	Real-Time	User Speed

summarizes the data used in this study. As you can see, a hot-spot map is generated using three layers: cycling accidents, motorcycle accidents, and car accidents in 2018. All of the static data on this page were taken from the Toronto Police Department’s website. Traffic Developer’s website, Open Weather Map, and Android application are used to obtain real-time traffic, weather, and speed data. Mean-Stack is used to construct the back-end side. Mean-Stack is a free and open-source JavaScript software stack that enables the development of dynamic web services and web applications. Due to the fact that all components of the MEAN stack enable JavaScript apps, MEAN applications can be developed in a single language for both back-end and front-end execution contexts. While the MEAN stack is sometimes compared directly to other popular web development stacks, such as the LAMP stack, the MEAN stack’s components are more abstract, including a web application display layer, but not an operating system layer. Due to the adaptability, scalability, and extensibility of MEAN stack applications, they are an ideal fit for cloud hosting (Figure 8).

The front-end is developed using Java and Kotlin in the Android Studio IDE. Additionally, Google Cloud Messaging is used to communicate between our server and the mobile application. The mobile application’s base map is a Google map that displays the user’s location. Additionally, the TTS Google API is used to translate the alarm text to speech. Additionally, a Python environment is used to evaluate the system.

6. Evaluation and Results

Figure 9a–c illustrate the three layers utilized to detect hot-spot map locations. Prior to combining these layers, it is necessary to classify the border surrounding each point. Figure 10a–c depict the categorized maps. Figure 11 depicts the final map of discovered hot-spots. This map depicts a dangerous area in Toronto based on historical automobile, bike, and motorbike accidents that were merged using AHP in the model builder ArcGIS. More vivid spots indicate a greater likelihood of an accident, and the majority of them have occurred at intersections. The weights estimated for these layers are shown in

Table 3. The attributes and values that are considered for weighting three historical accident layers to generate the hot-spot map.

Criterion				Normalized			Average
-	Bike	Motorcycle	Vehicle	Bike	Motorcycle	Vehicle	-
Bike	1	3	4	0.632	0.631	0.631	0.63133
Motorcycle	1/3	1	4/3	0.21	0.229	0.21	0.2163
Vehicle	1/4	3/4	1	0.158	0.157	0.157	0.157
Sum	1.58	4.75	6.33	≈1	≈1	≈1	≈1

.

To have a better grasp of the system, you can use the following link to download the Android application: https://github.com/Amirhossein-Nourbakhsh/gis_bike/tree/main (accessed on 10 October 2022).

The application’s initial page is an introductory page; by selecting the proceed button, the application’s main page will appear. When you start traveling on the main page, the coordinates are updated and the perspective view is turned behind you by 30 degrees. On the map, there are two buttons (Figure 12). The first button is used to evaluate the system, which is discussed in detail in the assessment section. The second button reveals the Firebase identification number. You can have access to the unique ID by clicking on this button. This unique identifier is required in order to experiment with the web services API built for this project. To send an alarm to your phone, for instance, you must have a Firebase ID as an input (

Table 4. The web service APIs that are implemented in this project by using Mean-Stack.

IP and Port	Name	Input	Output
http://212.90.102.16:4242	getWeathermap	(lat, long)	Json Object
http://212.90.102.16:4242	getTrafficLs	(lattop, lontop, latbtm, lonbtm)	Json Object
http://212.90.102.16:4242	fuzzyReasoning	(speed, tr, col, weather)	String
http://212.90.102.16:4242	sendNotification	(firebaseId, txtAlarm)	String

). The following section describes how to use the APIs. Table 4 lists the web services and their associated inputs and outputs. You can obtain the results by pasting the API IP address, port number, and name into a web browser and adjusting your input. For instance, if a user is riding a bike in Toronto, the user’s current latitude and longitude are transmitted to the server, and the server may now determine the weather conditions in the area where the user is now located by using the following link: http://212.90.102.16:4242/amir/getweathermap?lat={your_latitude}&long={your_longitude} (accessed on 10 October 2022).

Assuming that the latitude = 43.77 and longitude = −79.492749, the weather condition can be accessible by using below link: http://212.90.102.16:4242/amir/getweathermap?lat=43.77&long=-79.492749 (accessed on 10 October 2022).

You can modify the latitude and longitude of the above link to obtain information on the weather in a specific area. At each point in time, the server will repeat this operation in order to obtain all factors (weather, traffic, historical data, speed). When the decision maker wishes to make a decision, the web service (described in Table 4) will be invoked. This web service is based on the fuzzy reasoning (Mamdani) technique. Additionally, you can attempt to obtain the outcome of this online service by providing your own specifications. For instance: http://212.90.102.16:4242/amir/fuzzyreasoning?speed=13&tr=2&col=20&weather=rainy (accessed on 10 October 2022).

If the decision maker chooses to transmit an alarm to the user, the web service (described in Table 4) is engaged. As previously stated, each user has a unique Firebase ID that is used to receive notifications from the server. Once the program is installed, you can copy and paste your Firebase ID into the below link, along with your alarm wording, such as “be cautious”. If you use this API, you will receive an alarm, which the mobile device will translate to voice and play. http://195.248.241.80:4242/amir/sendNotification?firebaseId={your_firebase_id}&txtAlarm={your_text_alarm} (accessed on 10 October 2022).

When analyzing the proposed system, the primary concern should be the system’s ability to react to all requests at any given time. Due to the difficulty of venturing out in this situation, we create a simulated environment using Python. Ten thousand moving bikes are manufactured in this scenario. To obtain traffic and weather data, we will require the latitude and longitude; the coordinate system will be translated from x,y to latitude, longitude. Various amounts of requests at any given time are tested in this scenario. Four distinct timings are logged for each request. The first record indicates when the user sends the request to the server, the second indicates when the server sends the request to Firebase, the third indicates when Firebase sends the alert to the user, and the fourth indicates when the user receives the warning. The final evaluation demonstrates that if the system only uses the server’s main thread to react to requests, it can handle up to 1000 requests. When more than 1000 requests are received, it is necessary to distribute the threads by using multi-threading programming to reply to the requests. For instance, if the system utilizes the main thread to answer to 1000 requests for 10,000, the system will be delayed by 6.781 s (Figure 13).

Many researchers have been working on improving the safety of cyclists [50,51,52,53]. In order to compare the proposed system in this paper with other publications, they can be analyzed from different perspectives. Some of them improved the safety of cyclists by working on the physical aspects of entities such as road networks and roundabouts [54,55,56,57,58]. On the other hand, the contribution of some of them is mostly related to intelligent transportation systems and road infrastructure [59,60]. Most available models focus on the approach’s theory; however, in this paper, we attempted to bridge the gap between theory and real-world application to improve cyclist safety by integrating geospatial data layers with different criteria.

7. Conclusions

We propose a situational awareness system that employs an integrated method to improve cyclic safety. The created system has the potential to significantly improve cyclists’ situational awareness when they use their mobile phones as communication tools that include real-time weather, environment, and traffic information. It is worth mentioning that the interaction between users and their phones is not needed during the riding. Finally, by integrating real-time and static historical data, a fuzzy decision model offers a series of suggestions to assist cyclists in making a more informed and safe decision during their route. The system was built on the Android platform with extremely lightweight to minimize battery drain and network consumption. The developed system warns the end user via speech to avoid destroying the interface. The app’s functionality can be modified and scaled to support a large number of users and, potentially, other app platforms. We are aware of the limitation of the current version of the system. Some parts of the system are designed by hand-crafted weighting and rules that may cause a hard and heuristic constraint. Many different improvements and experiments remain for future work due to lack of time. As the safety and healthiness of the cyclists is the primary goal, we are planning to integrate the heat-map of COVID-19 cases with our system in order to decrease the passing of an area with a high risk of exposure to the virus. Moreover, to run the system on a large road network, we plan to bring our system to a distributed architecture based on geospatial data.