Application of artificial intelligence technology in the economic development of urban intelligent transportation system

Process of collecting data

This article uses artificial intelligence technology to collect, process, and analyze urban traffic data to provide accurate data support for traffic decision-making. Sensors and wireless communication technology are used to collect on-site traffic data. The specific information collection process is shown in Fig. 1.

Figure 1 demonstrates the traffic information collection process based on a randomly captured traffic map of an intersection in a certain city. The red instrument is a sensor that senses and converts the physical and chemical quantities in the surrounding environment of the intersection. Various signals or information at intersections can be transformed into electrical or other signals that can be detected, recorded, processed, and utilized. This article mainly analyzes vehicles in motion, and in response to the difficulty of collecting data information during vehicle movement, sensors are used primarily for object motion and position detection. It can provide helpful information and signals by transforming various data of moving objects into intelligent decision-making in intelligent transportation systems. In addition, the green instrument is the central server, which plays a core role in the intelligent transportation system. A large number of vehicles pass through the intersection every day. Therefore, huge data would be generated. At this time, the central server is required to store and manage data in a centralized manner, calculate and process data, manage and control the network, share and coordinate resources, manage security and permissions, and backup and restore data to provide support for the stable operation and efficient management of the intelligent transportation system. Finally, the blue instrument is the data management platform, which can effectively help the intelligent transportation system management to use the data processed by the central server and has a massive role in improving the data collection technology of the intelligent transportation system in the later stage.

Processing collected data information

Through the process of collecting data in Fig. 1, it is possible to collect various information on the vehicles driving on the road. Through data collection, it is possible to effectively grasp the traffic flow, vehicle, and road condition information of the road. Adjustments can be made in a timely manner to avoid traffic congestion, as shown in Fig. 2.

Figure 2A is the traffic flow information map of the intersection during the morning rush hour, where the horizontal axis represents the vehicle type. There are three types of car models, namely taxis, sedans, and buses. The vertical axis represents the quantity, which means the number of each type of car during the morning rush hour at an intersection. From Fig. 2A, it can be clearly seen that the number of vehicles is the highest, while the number of buses is the lowest. This indicates that residents who travel in the area in the morning tend to drive their cars. Figure 2B is the traffic flow information map of the noon peak at the intersection, where the horizontal axis represents the vehicle type, which is mainly divided into three types: taxi, sedan, and bus. The vertical axis represents the direction in which the vehicle enters the intersection. Four entrances and exits at the intersection correspond to four directions: east, south, west, and north. Figure 2B shows that the number of small cars is also the highest during the afternoon rush hour, while the number of large buses is the lowest, indicating that there are more models of small and medium-sized cars on urban roads.

Figure 2C is the traffic flow information map of the intersection during the evening rush hour, where the horizontal axis represents different directions. The vertical axis represents the number of vehicles in each direction. Figure 2C shows that the number of buses entering from the north entrance is less than that from the east entrance, while the number of taxis entering from the north entrance and west entrance is less than that from the east entrance and south entrance. Similar to taxis, cars also have fewer entrances from the north and west entrances than from the east and south entrances. Figure 2D shows the traffic flow information map of the intersection at night. It can be clearly seen that the number of three models is significantly reduced compared to Figs. 2A–2C. This is because the evening is a time for residents to rest, and it is more evident from the parallel comparison of the three models that the number of buses is the least.

In conclusion, the intelligent transportation system can monitor and analyze road traffic in real-time through data collection and processing. The data collected by devices such as traffic sensors and cameras can be used for real-time monitoring of congestion conditions, traffic flow, vehicle speed, and other indicators, and based on this data, timely traffic scheduling and management decisions can be made. Data collection and processing can provide critical traffic information, enabling traffic managers to formulate targeted traffic strategies and measures to optimize the utilization of road resources, reduce congestion, and improve traffic efficiency. In addition, through analyzing and processing traffic data, the intelligent transportation system can detect abnormal conditions and send out accident and traffic safety warnings. In a word, the intelligent transportation system’s data acquisition and processing technology can provide accurate and practical data support for traffic managers, drivers, and urban planners to improve traffic conditions and the efficiency and safety of the transportation system.

Hardware and infrastructure requirements

Proper implementation of artificial intelligence traffic monitoring calls for strong hardware and a base infrastructure to support a large number of systems. Some of these components include high-definition and real-time cameras for capturing video data, IoT devices for capturing real-time data on traffic and climate, and edge devices for real-time data processing to minimize on delay. Protocols like 5G are used in connecting devices to enhance the transfer of data between gadgets and some mainframes at the shortest time possible. Moreover, cloud-based solutions give technological support for storing and processing big data to integrate into one platform. Other aspects include compatibility with existing systems, including traffic signals and vehicle detectors, in order to allow the integration. These parts make up the foundational elements of AI-centred traffic systems so that accurate monitoring, analysis, and decision-making can take place throughout cities.

Traffic flow prediction model

Accurately predicting the traffic capacity of urban roads can effectively dispatch congested sections, alleviate traffic pressure on congested sections (Shengdong, Zhengxian & Yixiang, 2019), and save residents’ fuel and time costs for travel. Therefore, this article uses artificial intelligence technology to establish a traffic flow prediction model using the support vector machine (SVM) regression algorithm in machine learning algorithms to achieve prediction of urban traffic flow (Saleem et al., 2022).

First, define nonlinear mapping:

(1) $$\mathrm{\partial }(\cdot ):{\mathrm{\theta }}^{\mathrm{m}}\to {\mathrm{\theta }}^{\mathrm{m}}\mathrm{j}.$$

Next, input the training dataset:

(2) $$\left\{\left({\mathrm{C}}_{\mathrm{o}},{\mathrm{u}}_{\mathrm{o}}\right)\right\}\begin{array}{c}\mathrm{M}\\ \mathrm{o}=1\end{array}.$$

Finally, map to ${\mathrm{\theta }}^{\mathrm{m}}\mathrm{j}$.

In the plane of the dataset, there is a non-linear relationship between the input and output data, denoted as $\mathrm{g}$, and the support vector regression function is:

(3) $$\mathrm{g}\left(\mathrm{C}\right)={\mathrm{E}}^{\mathrm{Y}}\mathrm{\theta }\left(\mathrm{C}\right)+\mathrm{n}.$$

Among them, $\mathrm{g}\left(\mathrm{C}\right)$ is the predicted value, coefficient $\mathrm{\theta }\left(\mathrm{\theta }\in {\mathrm{\theta }}^{\mathrm{m}}\mathrm{j}\right)$ and coefficient $\mathrm{n}\left(\mathrm{n}\in \mathrm{\theta }\right)$ are variable values.

Support vector regression is used to minimize empirical risk minimization, as shown in the following formula:

(4) $$T_{emp}\left(g\right)={\displaystyle \frac{1}{M}\sum _{o=1}^M{\mu }_{\epsilon }\left(y_o{\delta }^E\gamma \left(c_o\right)+n\right).}$$

${\mu }_{ϵ}\left(y,\gamma \left(c_o\right)\right)$ is an insensitive Loss function, and $\mathrm{ϵ}$ is defined as follows:

(5) $${\mu }_{ϵ}\left(y,g\left(c\right)\right)=\{\begin{array}{cc}\left|g\left(c\right)-y\right|-ϵ& \left|g\left(c\right)-y\right|\ge ϵ\hfill \\ 0& \left|g\left(c\right)-y\right|<ϵ.\hfill \end{array}$$

${\mu }_{ϵ}\left(y,g\left(c\right)\right)$ searches for the optimal hyperplane in the space of high-dimensional features and divides the input training data into two sides of the hyperplane. The Hyperplane is two lines. These two lines can be denoted as $-\mathrm{ϵ}$ and $+\mathrm{\epsilon }$, respectively, to facilitate the formation of the maximum distance between subsets in the training data. Then, when training the data in the support vector machine regression model, the training error is minimized, as shown in the following formula:

(6) $$\mathrm{m}\mathrm{i}\mathrm{n}{\mathrm{T}}_{\mathrm{ϵ}}\left(\delta ,{\vartheta }^{\ast },\vartheta \right)={\displaystyle \frac{1}{2}{\delta }^{\mathrm{Y}}\delta +\mathrm{V}\sum _{\mathrm{o}=1}^{\mathrm{M}}\left({\vartheta }_{\mathrm{o}}^{\ast }+{\vartheta }_{\mathrm{o}}\right).}$$

The constraint conditions of the above formula are:

(7) $$\begin{array}{c}\begin{array}{c}{\mathrm{y}}_{\mathrm{o}}-{\mathrm{E}}^{\mathrm{Y}}\mathrm{\theta }\left({\mathrm{c}}_{\mathrm{o}}\right)-\mathrm{n}\le \mathrm{\epsilon }+{\vartheta }_{\mathrm{o}}^{\ast },\mathrm{o}=1,2,...,\mathrm{M},\\ -{\mathrm{y}}_{\mathrm{o}}+{\mathrm{E}}^{\mathrm{Y}}\mathrm{\theta }\left({\mathrm{c}}_{\mathrm{o}}\right)+\mathrm{\epsilon }+{\vartheta }_{\mathrm{o}},\mathrm{o}=1,2,...,\mathrm{M},\end{array}\\ \begin{array}{c}{\vartheta }_{\mathrm{o}}^{\ast }\ge 0,\mathrm{o}=1,2,...,\mathrm{M},\\ {\vartheta }_{\mathrm{o}},\mathrm{o}=1,2,...,\mathrm{M},.\end{array}\end{array}$$

The insensitive Loss function minimizes the data training errors of $g\left(c\right)$ and $\mathrm{y}$. The errors on both sides of the hyperplane are expressed with different parameters. The training errors above $+\mathrm{\epsilon }$ are expressed as ${\vartheta }_{\mathrm{o}}^{\ast }$, and the training errors above $-\mathrm{\epsilon }$ are expressed as ${\vartheta }_{\mathrm{o}}$.

By optimizing the parameter vector $\delta$ twice, the following formula can be obtained:

(8) $$\delta =\sum _{0=1}^M\left({\alpha }_0^{\ast }-{\alpha }_0\right)\theta \left({\mathrm{c}}_{\mathrm{o}}\right)$$

${\alpha }_0^{\ast },{\alpha }_0$ are the Lagrange multiplier obtained through secondary optimization, and finally the regression function of the support vector machine regression model is obtained in the dual space, as shown in the following formula:

(9) $$g\left(c\right)=\sum _{0=1}^M\left({\alpha }_0^{\ast }-{\alpha }_0\right)L\left({\mathrm{c}}_{\mathrm{o}},c\right)+n.$$

The traffic flow prediction in this investigation uses the SVM regression algorithm. The decision trees and neural networks were not selected instead of SVM because of the latter’s efficiency in small and medium sample size data sets for the non-linear relationships in data and generalization. On the one hand, decision trees have the benefits of being interpretable and having low computational complexity. Still, on the other hand, they lead to overfitting, especially in cases of high traffic complexity. On the other hand, ITS applications that involve neural networks need much larger datasets and computational horsepower, which does not fit ITS’s real-time needs. An efficient mathematical tool in SVM regression transforms nonlinear data into a higher-dimensional space for minimal empirical risk for accurate predictions. Due to its application of the ε-insensitive loss function, minor changes in prediction can occur without much affecting the model. Furthermore, the optimization of SVM makes it possible to define a hyperplane with maximum margin, which is the method’s main advantage when it is used to predict traffic flows in cases where nonlinear dependencies are observed.

In summary, this article uses the support vector machine regression algorithm to map the trained nonlinear data into an infinite dimensional space, which can effectively accurately predict traffic flow data with nonlinear relationships.

Prediction results of support vector machine regression model

This article establishes a traffic flow prediction model based on the support vector machine regression algorithm and applies the model to actual road vehicle prediction. The predicted results are shown in Table 1.

As shown in Table 1, the number of vehicles in the morning and evening rush hours of a certain city section is analyzed by inputting data sequence numbers for different periods to predict the number of vehicles in that section the next day at the same time. Table 1 shows that the difference between the predicted and actual values is within five vehicles, with a slight difference. This indicates that the algorithm model can accurately predict road traffic flow.

The development of a traffic flow forecast model is highly significant. It can furnish insights on forthcoming traffic conditions, aiding traffic managers in the efficient planning and allocation of traffic resources. Secondly, traffic flow prediction algorithms can forecast future traffic conditions for drivers and travellers. Precise traffic flow forecasting can assist individuals in circumventing crowded road segments, selecting optimal routes, conserving travel time, and diminishing fuel consumption, thus enhancing the travel experience and navigation services. The traffic flow prediction model can forecast traffic bottlenecks and identify traffic abnormalities. The model identifies anomalous events and provides prompt alerts through the analysis of historical and real-time data. This can help reduce accident risks, improve traffic safety, and provide an essential reference for urban planners and traffic decision-makers. Establishing a traffic flow prediction model requires many historical and real-time traffic data, data science, and machine learning techniques for modeling and prediction. This promotes the gradual transformation of traffic management to data-driven and intelligent and encourages the development of urban traffic management towards intelligent transportation systems.

Optimization scheduling algorithm

Comparison of results before and after scheduling

This article uses optimization scheduling algorithms to optimize the scheduling of congested roads, and compares the number of stops on frequently congested road sections, average waiting time of vehicles, and average speed of vehicles before and after the application of scheduling algorithms, as shown in Fig. 3.

As shown in Fig. 3, this article analyzes the number of stops, average waiting time, and average speed of vehicles on frequently congested road sections in a certain city before and after the application of scheduling algorithms at the same time. Figure 3A shows the number of stops on frequently congested sections before and after using the scheduling algorithm. The abscissa is the time, in seconds, and the ordinate is the number of stops. It can be clearly seen from Fig. 3A that the number of stops on this road section fluctuated around 60 before using the scheduling algorithm, while the number of stops after using the scheduling algorithm fluctuated around 30. It indicates that effective scheduling of frequently congested road sections through scheduling algorithms greatly reduces the number of stops for vehicles on that road section.

Figure 3B shows the average waiting time of vehicles on frequently congested road sections before and after the use of scheduling algorithms. This statistics shows the waiting time for the red light on the road section. From Fig. 3B, it can be clearly seen that the waiting time for vehicles at the red light before applying the scheduling algorithm fluctuated around 75 s. After applying the scheduling algorithm, the time for vehicles to wait for red lights fluctuated around 65 s, indicating that the application of the scheduling algorithm greatly shortens the time for vehicles to wait for red lights on congested road sections.

Figure 3C shows the average speed of vehicles on frequently congested road sections before and after the use of scheduling algorithms, which measures the speed at which vehicles are traveling on that road section. The slower the congested vehicles travel, the faster the less congested vehicles travel. From Fig. 3C, it can be seen that the driving speed of vehicles after applying the scheduling algorithm is faster than that before applying the scheduling algorithm, indicating that the effective scheduling algorithm ensures the smoothness of the road, thereby improving the driving speed of vehicles on this section. By optimizing route planning, reducing congestion, improving transportation efficiency, saving resources and energy, and providing more efficient and sustainable solutions for logistics transportation and traffic management, vehicle scheduling algorithms have the potential to create solutions that are more efficient and sustainable (Saleem et al., 2022). It is of great significance to improve the efficiency of the transportation system, reduce logistics costs, and improve environmental quality.

Mathematical model of urban traffic signal control

The current urban traffic signal control time has a certain delay (Jinsong & Mingjun, 2020). After long-term research and summary by many experts, the average delay time $\mathrm{f}$ calculation formula is proposed as follows:

(18) $$\mathrm{f}={\displaystyle \frac{\mathrm{V}{\left(1-\mathrm{\epsilon }\right)}^2}{2\left(1-\mathrm{\epsilon }\mathrm{c}\right)}+{\displaystyle \frac{{\mathrm{c}}^2}{2\mathrm{w}\left(1-\mathrm{c}\right)}-0.65\left({\displaystyle \frac{\mathrm{V}}{{\mathrm{w}}^2}}\right){\displaystyle \frac{1}{3}{\mathrm{c}}^{\left(2+5\mathrm{\epsilon }\right)}.}}}$$

Among them, $\mathrm{c}$ is the saturation of traffic flow; $\mathrm{V}$ is the time period of the signal light; $\mathrm{\epsilon }$ is the time of the green signal light; $\mathrm{w}$ is the traffic volume (pch/h).

In practical applications, it was found that the value of $0.65\left({\displaystyle \frac{\mathrm{V}}{{\mathrm{w}}^2}}\right){\displaystyle \frac{1}{3}{\mathrm{c}}^{\left(2+5\mathrm{\epsilon }\right)}}$ in the formula is very small, so it can be ignored in actual calculations to obtain the calculation formula for the actual average delay time $\mathrm{f}$:

(19) $$\mathrm{f}={\displaystyle \frac{\mathrm{V}{\left(1-\mathrm{\epsilon }\right)}^2}{2\left(1-\mathrm{\epsilon }\mathrm{c}\right)}+{\displaystyle \frac{{\mathrm{c}}^2}{2\mathrm{w}\left(1-\mathrm{c}\right)}.}}$$

Taking the intersection of a real city as an example, the total delay time at that intersection can be calculated using the following formula:

(20) $$\mathrm{F}=\sum _{\mathrm{o}=1}^4\sum _{\mathrm{k}=1}^2\left\{{\mathrm{w}}_{\mathrm{k}\mathrm{o}}\left[{\displaystyle \frac{\mathrm{V}{\left(1-{\mathrm{\epsilon }}_{\mathrm{o}}\right)}^2}{2\left(1-{\mathrm{\epsilon }}_{\mathrm{o}}{\mathrm{c}}_{\mathrm{k}\mathrm{o}}\right)}+{\displaystyle \frac{{{\mathrm{c}}_{\mathrm{k}\mathrm{o}}}^2}{2{\mathrm{w}}_{\mathrm{k}\mathrm{o}}\left(1-{\mathrm{c}}_{\mathrm{k}\mathrm{o}}\right)}}}\right]\right\}.$$

Among them, ${\mathrm{w}}_{\mathrm{k}\mathrm{o}}$ is the traffic flow at the $\mathrm{k}$ entrance and exit in the direction allowed by Class $\mathrm{o}$ signal lights. ${\mathrm{c}}_{\mathrm{k}\mathrm{o}}$ is the saturation of traffic flow at the entrance and exit of the $\mathrm{o}$ class signal lamp and the $\mathrm{k}$ class signal lamp. ${\mathrm{\epsilon }}_{\mathrm{o}}$ is the time allowed for the green signal light in the direction of travel for Class $\mathrm{o}$ signal lights, ${\mathrm{\epsilon }}_{\mathrm{o}}={\mathrm{y}}_{\mathrm{o}}/\mathrm{V}$.

As shown in Formula (21), assuming the time of the green signal light is $r$, the time of the green signal light needs to meet the following constraints:

(21) $$y_o\le r\le V-A-30.$$

Among them, $A$ is the time lost due to delay at the intersection.

Furthermore, assuming that the maximum saturation value of the traffic flow at the intersection does not exceed 0.8, then $y_o$ needs to meet the following constraint conditions:

(22) $$y_o=h_{ro}\ge V{y}_{o,max}/\mathrm{0.8.}$$

Among them, $y_{o,max}$ is the maximum flow ratio at the intersection, and $h_{ro}$ is the effective time.

In summary, the set of constraints for the urban traffic signal control model is:

(23) $$\{\begin{array}{c}\sum _o^4{y}_o=V-A\\ y_o\le r\le V-A-30\\ y_o=h_{ro}\ge V{y}_{o,max}/0.8.\end{array}$$

Signal lamp controls four routes of vehicle driving

This article uses artificial intelligence technology to establish a mathematical model for urban traffic signal control, and analyzes the four driving routes of vehicles controlled by urban traffic signals based on the model, as shown in Fig. 4.

As shown in Fig. 4, the left side is a schematic diagram of an intersection, and the lines with arrows in different colors at the intersection are the schematic diagrams of four routes where vehicles are controlled by signal lights. These four routes are represented by A, B, C, and D, as shown in the upper right corner of Fig. 4. The directions of the entrance and exit of this intersection are up north, down south, left west, and right east, with specific markings shown in the bottom right corner of Fig. 4. Route A has four directions of travel. The first type is from west to east, the second type is from west to south, the third type is from east to west, and the fourth type is from east to north. Route B has two directions of travel, the first from west to north, and the second from east to south. Route C has four directions of travel. The first type is from north to south; the second type is from north to east; The third type is from south to north; the fourth type is from south to west. Route D has two directions of travel; the first is from north to west; the second type is from south to east. By controlling signals and allocating vehicles appropriately, it is possible to ensure residents’ safety while effectively alleviating traffic pressure and avoiding traffic accidents that can lead to road congestion.

Using models to calculate flow ratio

The urban traffic signal control model established above is used to calculate the traffic flow at intersections in the direction consistent with Fig. 4, as shown in Table 2.

As shown in Table 2, assuming the time period of the signal light is 130 s, the lost time due to delay is 10 s, and the minimum green signal time of the signal light at the intersection is 10 s. Assuming that the time of the yellow signal light is the same as the time lost due to delay, then the actual time of the green signal light shall prevail.

Taking the route allowed by Class A signal lights as an example, taking a relatively large $y$ value and substituting it into Formula (22) can obtain $y_A\ge 34$. Similarly, it can be concluded that Class B signal lights allow a flow ratio of $y_B\ge 29$ on the driving route; Class C signal lights allow a flow ratio of $y_C\ge 34$ on the driving route; Class D signal lights allow a traffic ratio of $y_D\ge 22$ on the driving route.

Time cost

This article applies artificial intelligence technology to the urban intelligent transportation system in three congested city sections at different levels. It carries out time cost statistics from four aspects: waiting time for red lights, time for traffic congestion, time for accident resolution, and time ultimately required to pass the sections. It is assumed that the economic cost of urban intelligent transportation systems before and after the application of AI technology is calculated at 10 yuan per minute, as shown in Table 3.

As shown in Table 3, the total time before the application of artificial intelligence technology in three congested sections of the city was between 100–200 min, and the calculated economic cost was 1,000–2,000 yuan. After applying artificial intelligence technology to three congested sections of the city with varying degrees of congestion, it ranged from 50 to 120 min. The calculated economic cost ranged from 500 to 1,200 yuan, indicating a significant improvement in time cost for three congested road sections. The time cost is relatively low, which can significantly reduce car fuel consumption and travel costs.

Labor costs

This article applies artificial intelligence technology to three different levels of congested road sections in the city and calculates labor costs. Labor cost refers to the use of a large number of manual means to command and allocate congested road sections. Labor costs are mainly calculated based on the number of people commanding traffic, the number of people handling traffic congestion, and the number of people solving accidents. This article assumes that 50 yuan per person is used to calculate the artificial economic cost before and after the application of AI technology in the urban intelligent transportation system, as shown in Table 4. Although application of traffic systems greatly lowers the labor expenses through deployment of intelligent traffic systems, this has an aspect of foe within acquaintance general traffic systems since the development itself bears an adverse impact within the worker genetically occupied within general traffic systems. Staff performing manual traffic control and congestion may find themselves displaced from their positions. However, the shift to new, AI-driven systems is also a chance for workforce reskilling for upskilling. Sectors seeking to reduce threats of job loss as a result of AI technology adoption can involve the development of systematic training sessions to enable workers displaced to master specific tasks involving AI system management, maintenance, and data analysis. Further, new employment opportunities may be developed and connected with new technologies in development, integration and newly planned cities’ systems, so there will always be needed personnel to work with intelligent transport systems.

As shown in Table 4, the economic cost of applying artificial intelligence technology to three different levels of congested road sections in the city was above 300 yuan. In contrast, the economic cost of applying artificial intelligence technology was reduced to below 300 yuan. Therefore, it can be effectively evaluated that the application of AI technology in urban intelligent transportation systems can greatly reduce labor costs, thus reducing road finance expenditure. The saved funds can be used to promote the development of other aspects of the city.

The new AI-driven systems of traffic maintenance were then compared with traditional traffic management methods to evaluate the number of economic gains in terms of efficiency and cost. The time costs of conventional methods were 112.5–194.4 min overall, and the financial costs were 1,125–1,944 yuan (see Table 3). When integrated with AI, the time costs were 59–110 min, and the cost significantly decreased to 590–1,100 yuan. Labour costs were also reduced; in the conventional system, human interfaces deployed 350–600 yuan, whereas AI-integrated interfaces optimized the amount to 150–250 yuan with less human intervention (Table 4). In aggregate, the cost-cutting model of employing AI was 6,051 yuan under the traditional system, whereas its AI counterparts were 3,090 yuan, thereby showing an overall saving of 2,961 yuan. These results support the argument that AI plays a significant part in enhancing cities’ traffic flow and other resource efficiency.

Comparison of economic costs

This article comprehensively compares the time economic and artificial economic costs in Tables 3 and 4 to better evaluate the economic benefits of urban intelligent transportation systems after the application of artificial intelligence technology, as shown in Fig. 6.

As shown in Fig. 6, it is evident that both the time and labor economic costs before the application of AI technology in urban intelligent transportation systems are much higher than those after the application of AI technology in urban intelligent transportation systems. The total economic cost before applying artificial intelligence technology was 6,051 yuan; the total economic cost after applying artificial intelligence technology was 3,090 yuan; the total economic cost after applying artificial intelligence technology was 2,961 yuan less than the total economic cost before applying artificial intelligence technology, which significantly reduces the maintenance cost of roads. AI outperforms traditional methods in processing real-time data, enabling quicker and more accurate decision-making in dynamic environments such as urban transportation systems. Its ability to continuously adapt and learn from vast datasets makes it more scalable for large and complex transportation networks, unlike traditional methods that struggle with such systems. Additionally, machine learning models used in AI systems offer higher precision in predicting and optimizing traffic flows compared to static algorithms used in traditional methods.