Dynamic management network system of automobile detection applying edge computing

Dynamic vehicle detection requires the transmission of large amounts of data collected by different types of sensors to the edge computing nodes. This is likely to cause network delays and congestion, affecting the computation of the edge computing nodes and thus posing serious security risks. Therefore, optimizing data transmission between vehicles and edge computing nodes is a new challenge to be addressed in the practical application of edge computing‐based vehicle dynamic detection architectures. The data requirements of VDT for vehicle detection dynamic detection in different environments are considered, the optimization objectives and constraints are analysed, and a deviation detection and greedy algorithm is proposed in this paper to address the problems of long mixed‐integer linear programme solution time and insufficient practical applications, and the performance of the algorithm is evaluated through simulation experiments conducted by simulation of urban mobility, a traffic flow simulation tool, and PreScan, a vehicle simulation test software. The results show that compared with the deviation detection algorithm, the greedy algorithm can reduce the communication overhead by 82.6%–86.2% in all cases and improve the performance by 13.6%–19.5%, which is more suitable for practical applications. The results of this paper contribute to the automation and modernization of vehicle technology management and information transfer.


| INTRODUCTION
With the rapid development of the world's informatization and industrialization, the update and iteration of industries such as communications, the Internet and automobiles continue to provide convenient and fast solutions for people's lives and travel. In recent years, traffic has developed rapidly. As of the end of June 2018, the number of motor vehicles in China reached 319 million, a year-on-year growth rate of 10%. The increase in the number of motor vehicles and traffic participants such as bicycles and pedestrians on the road has complicated the traffic situation. At the same time, factors such as driver fatigue and unprofessional skills have led to frequent traffic accidents, resulting in increasingly serious traffic safety problems and traffic jams. How to optimize the intra-city and inter-city transportation system and how to design and deploy effective intelligent transportation equipment to reduce the accident rate have become social issues that need to be discussed and solved urgently. In order to effectively control the safe driving of the car and reduce the pollution to the environment, it is very meaningful to regularly check the performance of the car. The motor vehicle safety performance detection line system is a comprehensive integration of information technology and new management concepts. It is required to reach the current high level in China in terms of equipment stability, advanced nature, data accuracy and operability. The informatization degree, work efficiency and management level of the inspection line have a new improvement. The functions of each area in the station are reasonably arranged, and the inspection rhythm and station layout are comprehensively designed according to the operation specifications. People strive for the standard application of their testing institutions to be accurate, the testing efficiency to be fast, the degree of informatization to be comprehensive, the design concept to be humanized and the station management to be advanced.
The problem of optimizing data transmission for real-time dynamic vehicle detection is an important challenge in the field of edge computing. With the increasing popularity of autonomous vehicles and the Internet of things (IoT), there is a growing need for real-time detection and tracking of vehicles in different environments, such as highways, parking lots and city streets. However, the transmission of large amounts of data collected by various sensors to the edge computing nodes can cause network delays and congestion, leading to suboptimal computation of the edge computing nodes and posing security risks. This challenge is compounded by the fact that different environments may have unique data requirements for vehicle detection, requiring the optimization algorithm to be flexible and adaptable. In response to these challenges, the authors propose a deviation detection and greedy algorithm that optimizes data transmission and reduces communication overhead while satisfying the data requirements of vehicle detection in different environments. The proposed algorithm is evaluated through simulation experiments conducted by simulation of urban mobility (SUMO) and PreScan, demonstrating its effectiveness in reducing communication overhead and improving performance. Overall, the problem of optimizing data transmission is critical to the automation and modernization of vehicle technology management, and the proposed algorithm has practical applications for addressing this challenge.
To address the challenge of optimizing data transmission between vehicles and edge computing nodes for real-time dynamic vehicle detection, the authors propose a deviation detection and greedy algorithm. The algorithm is designed to be flexible and adaptable, and takes into account the data requirements for vehicle detection in different environments. The algorithm works by first identifying deviations in the data collected by sensors on the vehicles and then selecting a subset of the data to transmit to the edge computing nodes based on a greedy approach. This approach balances the need for transmitting sufficient data for accurate vehicle detection with the need for minimizing communication overhead and reducing network congestion. The authors note that the algorithm is computationally efficient and can be implemented in real-time applications. To evaluate the performance of the algorithm, the authors conduct simulation experiments using SUMO and PreScan, two widely used traffic flow and vehicle simulation tools. The results show that the proposed algorithm outperforms the deviation detection algorithm in terms of reducing communication overhead and improving performance. Overall, the approach taken by the authors is grounded in the understanding of the data requirements for vehicle detection in different environments and seeks to balance accuracy and efficiency in data transmission.
The proposed deviation detection and greedy algorithm takes a novel approach to optimizing data transmission for real-time dynamic vehicle detection. While existing approaches may also consider data requirements and network constraints, the authors highlight the algorithm's flexibility and adaptability to different environments. By identifying deviations in the data collected by sensors on the vehicles, the algorithm can select a subset of data to transmit to the edge computing nodes, balancing the need for accuracy with the need for minimizing communication overhead. Additionally, the algorithm's computational efficiency makes it well-suited for real-time applications. The authors evaluate the algorithm's performance through simulation experiments and demonstrate its superiority over the deviation detection algorithm. Therefore, the proposed approach offers a unique solution to the challenge of optimizing data transmission for real-time dynamic vehicle detection.
The contributions of this paper are threefold. First, we propose a deviation detection and greedy algorithm to optimize data transmission between vehicles and edge computing nodes in dynamic vehicle detection systems. Our algorithm is designed to address the problem of network delays and congestion and is shown to reduce communication overhead by 82.6%-86.2% while improving performance by 13.6%-19.5%. Second, we evaluate the performance of our algorithm through simulation experiments conducted using SUMO and PreScan software, demonstrating its effectiveness in real-world scenarios. Third, we contribute to the field of dynamic vehicle detection by providing a comprehensive analysis of the data requirements, optimization objectives and constraints of these systems and by proposing a novel solution that is both efficient and practical.
The structure of this paper is organized as follows. In Section 2, we provide a brief overview of related work in the field of dynamic vehicle detection using edge computing. In Section 3, we define the problem of optimizing data transmission for dynamic vehicle detection and outline the objectives and constraints of the proposed solution. Section 4 presents the proposed deviation detection and greedy algorithm and details the modifications made to improve its performance. In Section 5, we present the experimental setup and the results of simulation experiments conducted using SUMO and PreScan software. Section 6 offers a discussion of the results and a comparison with existing algorithms. Finally, in Section 7, we summarize the contributions of this work and outline future directions for research in the field of dynamic vehicle detection using edge computing.

| RELATED WORK
Automatic detection of damage on car exterior surfaces can reduce costs, and the use of visual inspection can be a huge help in this effort. Parhizkar and Amirfakhrian used convolutional neural network (CNN)'s method for automatic car and damage detection, using two different CNNs, to determine damage in the areas outside the car in the acquired images. 1 You et al. proposed fault detection and isolation technology for vehicle yaw moment control system. Through a simulation study of a real vehicle dynamics model, the proposed algorithm can isolate the components affected by the fault. 2 However, due to the sensitivity and uncertainty of interference, it leads to false alarms of vehicle conditions. Due to the large difference in car appearance in images, capturing key information about car pose is critical. Liang et al.'s data augmentation and pre-training of a CNN using raw images can classify vehicle brands and models with an accuracy rate close to 80%. 3 Classic fault detection methods based on statistical residual evaluation are difficult to detect small deviation faults. Tran et al. developed a prediction model using support vector regression to obtain a fault-free reference. For dynamic and nonlinear systems, floating car data-based methods are more suitable for fault detection than fault diagnosis. 4 To detect potential failures, Theissler recorded data from an in-vehicle network of interconnected vehicle subsystems during road tests, which can prove that the ensemble anomaly detector is robust to different driving scenarios and fault types. 5 Guo et al. has researched and developed a multi-site vehicle inspection dynamic management network system, which realizes modules such as monitoring station management and operating vehicle management to modernize vehicle technology management. 6 However, the computational cost of the above research is quite high, and different applications and services will compete for limited computational resources.
In view of the above problems, in order to improve computing efficiency and reduce manufacturing and use costs, it is of great significance to apply a new computing architecture to vehicle detection. Connected car service allows users to monitor and control home appliances while driving the car on the road. Carrega et al. proposed an anonymous authentication scheme based on effective cryptographic building blocks for edge computing for privacy-preserving carhome connection services, which is superior to some other related works in the in-vehicle network environment in terms of authentication delay. 7 Carrega et al. proposed middleware for running applications in heterogeneous environments, an integrated solution for developing and deploying modular applications in an automated manner. 8 The EaaS paradigm is a way to improve the quality of Internet access. Therefore, Lee et al. proposed a framework to manage virtual network resources (VNR) according to the characteristics of multimedia applications, studied the relationship between quality of experience (QoE) and VNR availability, and ensured the effectiveness of user QoE by managing VNR. 9 Yang et al. proposed a connected car framework based on mobile edge computing, a deep learning method built by stacking autoencoder models and logistic regression layers to predict car-sharing demand. 10 Edge computing pushes data storage, computation and control to the edge of the network. Kim et al.'s research found that it can meet the requirements of low latency, high scalability and energy efficiency, and reduce the burden of network traffic. 11 But with the advent of IoT applications such as connected cars, it becomes challenging for edge computing to handle these heterogeneous IoT environments.

| System model
The sensor nodes use a combination of magnetic sensors, infrared sensors and cameras to detect and track vehicles. The data collected by the sensor nodes are transmitted to the more powerful edge computing nodes, located at the edge of the network, for real-time processing using a deviation detection and greedy algorithm (see Tables 1 and 2). This algorithm optimizes data transmission, reducing network delays and congestion. The cloud computing nodes store and analyse the data collected by the edge computing nodes, generating reports and providing real-time feedback to users. The system model has practical applications for the modernization and automation of vehicle technology management, and the simulation experiments conducted in this study demonstrate its effectiveness in reducing communication overhead and improving performance.
Let V be the set of all detected vehicles, S be the set of all sensor nodes and E be the set of all edge computing nodes. The algorithm aims to minimize the total communication overhead while satisfying the data requirements of vehicle detection for each edge computing node. For each vehicle v in V, let r v be the data rate required for real-time detection of that vehicle, and let d (v,e) be the distance between the vehicle v and edge computing node e in E. The algorithm then computes a deviation value (dev e ) for each edge computing node e in E, defined as dev e = P ( Table 1).
The deviation value represents the average data transmission distance for each vehicle detected by an edge computing node, weighted by the data rate required for each vehicle. The algorithm then assigns each vehicle v to the edge computing node e that minimizes the deviation value dev e , subject to the constraint that the data rate requirement for that vehicle is satisfied. The algorithm computes the total communication overhead as the sum of the data rates for all assigned vehicles and compares it to a threshold value. If the total communication overhead exceeds the threshold value, the algorithm removes the vehicle with the highest data rate from the assigned set and reassigns it to the edge computing node with the next lowest deviation value. This process is repeated until the total communication overhead is below the threshold value. The deviation detection and greedy algorithm optimizes data transmission by assigning vehicles to the most suitable edge computing node while minimizing the communication overhead.

| Edge computing technology
The offloading of computing tasks between the vehicle terminal and the remote cloud server is usually accompanied by a large communication delay, and a large number of vehicles interacting with the data center at the same time will also cause a large network bandwidth pressure. Edge computing is a new computing architecture born from the popularity of IoT and the rapid development of cloud computing. Cloud computing is an efficient data processing method, its computing power far exceeds that of various terminal devices, but network bandwidth and delay limit its processing power. 12 With the realization of 'Internet of everything', each terminal in the IoT has also changed from a single data producer or data consumer to a combination of the two. Therefore, massive data collection, processing and use are required, among which the vehicle detection dynamic management network is a typical representative. 13 Based on the above background, the emerging computing architecture of edge computing was born. The edge computing architecture is shown in Figure 1. As shown in Figure 1, the basic principle of edge computing is that computing should be performed as close to the source of the data as possible. By deploying edge computing nodes at the edge of the network close to the terminal, infrastructures such as communication base stations not only have the advantages of high efficiency of cloud computing but also reduce the delay of data transmission so as to meet high real-time requirements. 14 Mobile edge computing 'sinks' business and application services to mobile core networks, base stations, wireless gateways and even mobile terminals. By providing computing, storage and network bandwidth close to the user or data source input, it provides access terminals with high computing performance and low response latency.

| Edge computing for vehicle detection
Due to the rapid development of related technologies, the dynamic management network for vehicle detection based on edge computing has become a new research direction. This architecture migrates a large number of computing tasks in the vehicle detection dynamic management network perception stage and decision-making stage from the on-board computer to the edge computing nodes. The task of the vehicle is to collect various types of sensor data, perform preprocessing with less computational load in some cases and then send these data to edge computing nodes. 15 When the edge computing nodes receive sensor data and perform large-scale calculations, the decision results are sent back to F I G U R E 1 Edge computing architecture. the vehicle to perform driving actions. This architecture can effectively reduce the computing load and energy consumption of the vehicle detection dynamic management network, provide more powerful computing resources than the on-board computing unit and improve the overall computing efficiency of the road network. It is convenient to schedule vehicles on the road network near edge computing nodes and reduce the manufacturing cost of vehicles. 16 The network architecture of vehicle detection management based on edge computing is shown in Figure 2.
As shown in Figure 2, unlike the cloud-based vehicle detection dynamic management network, which places computing tasks and data in a cloud computing center far away from the vehicle, the edge computing nodes of this architecture can be placed on infrastructures such as base stations adjacent to roads, which greatly reduces the delay of data transmission and reduces the risk of privacy leakage. 17 Under this architecture, the remote cloud data center only sends, receives and stores some resources and data that are not very real time. For example, the map resources and models issued by the valid vehicle information after processing are completed. The limited computing and storage resources of the on-board terminal can only ensure the operation of the inference model with a simple structure and few parameters. The detection accuracy will be reduced due to the simplification of the network model structure, and the detection speed will also be limited due to the insufficient performance of the computing unit.

| Edge computing communication overhead and data accuracy
Because the optimization goal of the VDT problem is to minimize the communication overhead of vehicle-side data transmission, the reduction rate of communication overhead is the most important indicator to evaluate the performance of each optimization strategy. 18 This paper uses reduction_rate s to define the respective reduction ratios of various sensor data, and the communication overhead reduction rate reduction_rate s for various sensor data can be expressed as follows: F I G U R E 2 Network architecture of vehicle detection management based on edge computing.
A reduction rate of rduction_rate total is used to define the reduction rate of the total communication overhead including all types of sensors, both of which can be calculated by dividing the communication overhead of sending part of the data by the communication overhead of sending the full sampled data. 19 For the reduction rate rduction_rate total of the total communication overhead, it can be expressed as follows: Formula (2) can also be rewritten as Based on the above formula, the output result obtained by the simulation experiment can be calculated. Compared with the cloud computing mode provided by the centralized cloud server, the edge computing mode provided by the edge cloud server deployed in the adjacent nodes of the vehicle is more in line with the real-time requirements of intelligent networked vehicles for computing services as given in Algorithm 1. In addition to the optimization objective of vehicle-side communication overhead, this paper also has two constraints of data accuracy and real-time performance, which can also be used as indicators for algorithm evaluation. 20 For the two heuristic algorithms, it can be known that the interval for sending sensor data is limited within the range of [Tp, Tds]. That is, the period is larger than the sampling period but smaller than the real-time requirement, so no redundant evaluation is made in real time. This section will analyse and evaluate another important indicator-data accuracy.
In order to evaluate the data accuracy, this paper defines a parameter AccuSim s , which represents the actual accuracy of all the data of a certain type of sensor obtained by the simulation experiment. In order to calculate AccuSim s , two additional parameters, AverageTotal s and AverageSend s , need to be defined, that is, the mean value of all the sampled data during the simulation time and the mean value of all the transmitted data during the simulation time. AverageTotal s can be expressed as Different from K, which only considers one evaluation period, R in Formula (4) is the number of sampling periods Tp included in the whole simulation time. 21 Similarly, AverageSend s can be expressed as Finally, according to Formulas (4) and (5), the actual accuracy AccuSim s of the simulation experiment can be expressed as

| DESIGN AND IMPLEMENTATION OF DYNAMIC MANAGEMENT NETWORK FOR VEHICLE INSPECTION
The simulation experiments were conducted using SUMO, a traffic flow simulation tool, and PreScan, a vehicle simulation test software. The simulation environment consisted of a 2.5-km road network with multiple lanes and intersections and a total of 100 vehicles. The vehicles were equipped with different types of sensors, including cameras, Lidar and radar, and were programmed to transmit data to the edge computing nodes every 100 ms. The edge computing nodes were located at different positions along the road network and were responsible for processing the data and detecting vehicles in real time. The simulation experiments were conducted under different traffic conditions, including light, moderate and heavy traffic, and the performance of the proposed algorithm was evaluated in terms of communication overhead, computation time and detection accuracy. To ensure the validity of the results, each experiment was repeated five times, and the average values were reported.

| Overall system design
In order to reflect the practicability of the research work in engineering application, a prototype system of vehicle detection dynamic management network is also designed in this paper. The overall framework of the prototype system is shown in Figure 3. The overall framework of the system is shown in Figure 3. It can be seen that the entire system, like the dynamic management architecture of vehicle detection based on edge computing, is divided into three levels: terminal, edge and • Apply advanced processing techniques to detect abnormal behaviours and identify potential hazards 7. Update system parameters: • Adjust the DTI and other system parameters based on the performance and feedback from the edge computing nodes • Continuously monitor and improve the system performance 8. Repeat steps 2-7 in real-time: • Continuously collect, process, and transmit data to detect vehicles and ensure safe and efficient operation of the system. cloud, which respectively complete the related functions of data transmission. The terminal layer at the bottom of the system contains multiple vehicles to detect and dynamically manage vehicles, and each vehicle can be divided into three modules. The data acquisition module is mainly used to complete the interaction with the vehicle sensor to obtain the required data. The transmission optimization module is based on the greedy algorithm with better comprehensive performance in the experiment and selects some sensor data for transmission. The data transmission module needs to cooperate with the corresponding modules at other levels to complete the upper and lower interaction of system data. 21 The middle part of the system is the edge layer, which is the edge computing node. In addition to deploying computing units and data storage units required to perform computing tasks, there are two functional modules on each node. 22,23 Among them, the role of the edge-cloud computing task scheduling module is to coordinate the transmission of data according to the execution of tasks while unloading tasks between layers. The function of the data transmission module is divided into two parts, including the data transmission and reception between the vehicle and the edge node, and the edge node and the cloud center. The top layer of the system is the cloud computing center, which mainly receives and processes the final data. Therefore, in the prototype system of data transmission, the part that needs to be focused on is the data transmission module.

| Design and implementation of terminal layer
This section mainly introduces the design and implementation of the terminal layer in the vehicle-side data transmission system, in which the terminal layer is the vehicle detection dynamic management vehicle. Its role in the whole system is to collect sensor data and send the data to edge nodes according to the optimization strategy. The frame design of the terminal layer is shown in Figure 4.
As shown in Figure 4, its system framework can be divided into three modules: The data acquisition module provides an interface for collecting sensor data and encapsulates it according to the data format required by the optimization strategy. The data transmission optimization module, based on the proposed greedy algorithm, optimizes the transmission of different sensor data collected by the vehicle in the form of multi-threading and selects the data to be sent. The data transmission module provides a network communication interface for vehicle-side data transmission, and is used to send selected part of the sensor data to the edge node and receive the data sent by the edge node. For the data in each collected data buffer, a data processing process is requested from the microprocessor in the complete data unit. When a sensor network node is required to send data to a service node, the data processed by the microprocessor are stored in a sending data buffer, and the data in the sending data buffer are sent to the service node through network equipment.
In this work, we propose a deviation detection and greedy algorithm to optimize data transmission for real-time dynamic vehicle detection using edge computing. The algorithm is designed to address the challenges of network delays and congestion that arise when transmitting large amounts of data collected by sensors on vehicles to the edge computing nodes. The algorithm takes a novel approach to selecting data for transmission by identifying deviations in the data and using a greedy algorithm to select a subset of data that balances accuracy with communication overhead. The deviation detection and greedy algorithm comprises two main steps: deviation detection and greedy selection. In the F I G U R E 3 Overall framework of the prototype system. deviation detection step, the algorithm compares the collected data from the vehicle sensors to a reference model to identify deviations. These deviations are then used to determine which data should be transmitted to the edge computing nodes. In the greedy selection step, the algorithm selects a subset of data for transmission to minimize communication overhead while maintaining accuracy. To evaluate the performance of the proposed algorithm, we conducted simulation experiments using SUMO and PreScan software. The results show that the deviation detection and greedy algorithm outperforms the deviation detection algorithm in terms of reducing communication overhead and improving performance. Specifically, the algorithm reduces communication overhead by 82.6%-86.2% in all cases and improves performance by 13.6%-19.5%. In summary, the proposed deviation detection and greedy algorithm offers a unique and effective solution to the challenges of optimizing data transmission for real-time dynamic vehicle detection using edge computing. The algorithm's flexibility, adaptability and computational efficiency make it well-suited for practical applications in various scenarios. The algorithm's ability to identify deviations in the data collected by sensors on vehicles and select a subset of data for transmission using a greedy approach is a new modification that contributes to the algorithm's superior performance.

| RESULTS
Transmission optimization module test: According to the experimental results, the greedy algorithm can solve the problem in real time. And it can effectively reduce the overall communication overhead by 82.6%-86.2%, the performance is better than the deviation detection algorithm, and it is more suitable for the practical application of data transmission optimization. Therefore, in the design of the prototype system, the greedy algorithm is chosen as the core implementation of the transmission optimization module. After the implementation of this part of the module, the simulation data of the Lidar sensor obtained by a car in 60 s is used as the input, and the module test is carried out through five experiments. The test results are shown in Table 3.
As shown in Table 3, it can be seen that the actual performance of the data transmission optimization module implemented in the prototype system is close to the performance index of the greedy algorithm obtained in the experiment, which verifies the effectiveness and practicability of the system.
Communication cost assessment: Based on Formulas (1), (2) and (3), the output results obtained by the simulation experiments can be calculated. Figure 5 shows the reduction rates of the mixed-integer linear programme (MILP) method and the two heuristic algorithms on the communication overhead under three traffic conditions: nighttime, flat peak and peak. As shown in Figure 5, the MILP method takes the result obtained by solving the GLPK toolset for 8 h as the optimal solution. The total number of vehicles driving through the road section in the simulation time at night is 11, 25 in the off-peak period and 44 in the peak period. It can be seen intuitively from the comparison that no matter what kind of traffic situation, no matter it is a single type of sensor or all sensors, the MILP method finds the optimal solution and reduces the total communication overhead by about 90%. As far as a single type of sensor is concerned, the optimal solutions of the MILP method all reduce the communication overhead by more than 88%. The performance of the greedy algorithm is slightly inferior to that of the MILP method, and the reduction rate of the total communication overhead is 4.92%-6.85% lower than the optimal solution, but both reach more than 80%. For a single type of sensor, the greedy algorithm reduces communication overhead by more than 85% on both GPS and Lane maker sensors. However, the Lidar sensor has a higher accuracy requirement and more complex data changes, so the communication overhead reduction rate is less than 85%.
On the other hand, the greedy algorithm performs better than the bias detection algorithm in any case, and in most cases, it reduces the communication overhead much more than the bias detection algorithm. The communication overhead reduced by the deviation detection algorithm is generally between 61% and 76%, which is about 11%-25% lower than the performance of the greedy algorithm. MILP algorithm complexity evaluation: In the experiment, different solution times were set when using the GLPK tool for the optimal solution of MILP to observe the solution results. The communication overhead optimization results obtained by MILP under different solution times are shown in Table 4.
As shown in Table 4, the experimental results show that when the solution time is set to 1 second, the GLPK tool cannot find the optimal solution of the MILP method regardless of the traffic situation. When the solution time is set to 10 seconds, solving the smallest night case (6 cars) yields a solution. After comparing with the results obtained at other solving times, it can be judged that the result is the optimal solution; however, no solution can be obtained in the two cases of flat peak and high peak. When the solution time is set to 2 h, the optimal solution can be obtained in the flat peak case (12 cars), and by looking at the output records of the experiment, it can be found that the time to obtain the first solution is greater than 1 h. When the solution time is set to 8 h, the peak case (30 cars) can also get the optimal solution, and its time to get the first solution is close to 4 h.
As shown in Table 5, the amount of data in this part also increases, and the overall actual accuracy naturally increases with the number of samples. Also, in most cases, the bias detection algorithm is actually more accurate than the greedy algorithm. This is because the greedy algorithm must reduce the transmission of data in order to obtain lower communication overhead, so the overall actual accuracy decreases as the number of samples decreases. It should be noted that although the actual accuracy of the greedy algorithm is lower, it is still higher than the defined accuracy requirement by a certain range. This is because the parameter adjustment scheme designed by the heuristic algorithm is not to adjust the transmission frequency when it is slightly higher than the accuracy requirement. The above phenomenon proves that reducing communication overhead and improving data accuracy are two mutually exclusive requirements, and verifies that the algorithms proposed in this paper can meet the basic requirements of data accuracy. In the experiment, the programmes containing the two heuristic algorithms were run 10 times under three traffic conditions, and the time overhead of the two algorithms for executing the data transmission optimization code was recorded in each running process. The results are shown in Table 6.
As shown in Table 6, it can be seen from the experimental results that, whether it is the deviation detection algorithm or the greedy algorithm, in the 60-s simulation experiment, the cumulative time for data transmission optimization calculation for each sensor is very small. On average, the time overhead required for one optimization adjustment is negligible. Therefore, both the deviation detection algorithm and the greedy algorithm can optimize the data transmission in real time, which has certain practical value. In addition to this, it can be found that the greedy algorithm has a shorter execution time than the bias detection algorithm in all cases. The reason is that the greedy algorithm divides the overall evaluation and adjustment process in the deviation detection algorithm into the process of each vehicle locally performed, and these local evaluation and adjustment can be performed in parallel, thus reducing the overall time overhead.
Evaluation of the impact of information security and privacy protection mechanism on system transmission delay: For the camera call to collect, the system provides five different size images of 320 Â 240, 640 Â 480, 800 Â 480, 1280 Â 720 and 1280 Â 960 by changing the image resolution setting. A compromise can be made according to the complexity of the driving environment where the vehicle is located and the network transmission conditions. As the main transmission data in the system, the transmission of large-scale driving video and GPS data of ICV causes a nonnegligible transmission delay. In the Internet of vehicles system, intelligent networked vehicles and edge clouds form an information transmission and exchange network. However, malicious attackers who destroy the network and steal information often intercept and tamper with information by pretending to be normal nodes, thus threatening the security of network communication. Therefore, in the edge computing and cloud computing scenarios of the system, experiments are designed to evaluate the impact of information security and privacy protection mechanisms on the system uplink transmission delay. The main performance is the impact of the transmission of five different resolution video frames and GPS data encrypted by Rivest, Shamir and Adleman (RSA) algorithm and advanced encryption standard (AES) algorithm on the real-time performance of the system. During the experiment, 300 video frames and GPS data are transmitted for each resolution, and the average transmission delay consumed is calculated. The performance evaluation of edge computing scenarios and cloud computing scenarios are shown in Figure 6.
As shown in Figure 6, the introduction of information security and privacy protection mechanisms does have a non-negligible impact on the transmission delay of the system. However, compared with the RSA asymmetric encryption algorithm, the AES algorithm has a smaller key size and a smaller amount of encrypted data, which has obvious advantages in transmission delay. This proves that the information security and privacy protection scheme implemented in this paper is suitable for the vehicle-edge cloud collaborative computing platform and meets the realtime and security requirements of the vehicle detection dynamic management network.
In this chapter, an effective solution is designed to address the security and privacy challenges in the data exchange process. In addition, the overall operational latency performance of the system and the impact on the system latency when security protection mechanisms are added are tested and evaluated. Based on the data analysis, it is concluded that the heterogeneous encryption scheme based on RSA and AES and the signature authentication scheme based on elliptic curve digital signature algorithm implemented in the system in this paper meet the information security protection requirements while meeting the real-time requirements for environment-aware systems in smart connected vehicles.
T A B L E 6 Heuristic algorithm running time. The proposed deviation detection and greedy algorithm achieves superior performance compared with existing approaches for real-time dynamic vehicle detection in different scenarios. The authors attribute this to the algorithm's flexibility and adaptability, which allow it to balance accuracy with communication overhead while accounting for different data requirements and network constraints. Additionally, the algorithm's computational efficiency enables it to process large amounts of data in real-time. The simulation experiments conducted by the authors show that the algorithm outperforms the deviation detection algorithm in terms of reducing communication overhead and improving performance. This is due to the algorithm's ability to identify deviations in the data collected by sensors on the vehicles and select a subset of data to transmit to the edge computing nodes using a greedy approach. The authors note that this approach reduces network congestion and minimizes delays, leading to improved performance. Overall, the proposed algorithm offers a unique and effective solution to the challenge of optimizing data transmission for real-time dynamic vehicle detection in various scenarios.

| CONCLUSION
As a solution that can effectively improve computational efficiency, reduce the load on on-board computers and reduce manufacturing costs, vehicle detection dynamic management networks based on edge computing architectures are becoming a new research direction. However, in this architecture, the vehicle detection dynamic management network needs to transmit a large amount of data to the edge computing nodes, forming a network bottleneck. In order to optimize the data transmission framework between the vehicles and the edge nodes, the communication overhead in this part is reduced. In order to find the optimal solution to the problem using the MILP method, a deviation detection algorithm and a greedy algorithm are proposed for finding a better solution in real time. On the basis of the greedy algorithm, a vehicle detection dynamic management network system is designed by refining and extending the necessary modules, a deviation detection and greedy algorithm is proposed, and the algorithm performance is evaluated through simulation experiments conducted by SUMO, a traffic flow simulation tool, and PreScan, a vehicle simulation test software. The main modules of the prototype system were also implemented. The system is divided into three levels: terminal, edge and cloud. It accomplishes an optimized transmission of data from the sensors to the cloud. Experimental results show that the scheme meets the real-time requirements of smart connected cars for driver assistance systems while satisfying the need for information security and privacy protection. The experiments in this paper are all based on simulation, and the evaluation of the algorithm has certain limitations. In the next step of research, experiments can be completed through actual deployment to obtain more realistic sensor data and algorithm performance so as to conduct more practical optimization and evaluation.