Cps-based focus vehicle highway tunnel prediction cruise cloud control method

Abstract

The application relates to a CPS-based key vehicle highway tunnel prediction cruise cloud control method and belongs to the technical field of traffic network connected automobile cloud control. The method fuses ground live information between various entities of a traffic system on the basis of a cloud control system basic platform, constructs a digital information mapping layer according to a multi-source heterogeneous traffic situation of a highway section, adopts an improved multi-angle RBF neural network data-driven prediction model on a cloud control application platform to predict and perceive real-time traffic situation information, and uses an algorithm solver in the cloud under limited constraint conditions to uniformly arrange and parallelly operate and optimize the control target by using the situation information and the prediction information, so that the purpose of safe and ecological driving of the key vehicle in the highway tunnel is achieved. The application solves the problems that a traditional prediction cruise control scheme lacks organic cooperation and fusion of a cyber layer and a physical layer, is limited by a perception range and a calculation capacity, information acquisition and processing are very limited, and driving safety and energy consumption are problems.

Description

Technical Field

[0001] This invention belongs to the field of cloud control technology for connected vehicles and relates to a cloud control method for predictive cruise of key vehicles in highway tunnels based on cyber-physical systems (CPS). Background Technology

[0002] Cyber-Physical Systems (CPS) are a class of complex, heterogeneous, open, and closed-loop systems; characterized by closed-loop feedback systems, distributed network systems, and real-time systems; integrating multiple fields such as control, computing, software, and communication. Connected vehicles are a new product, model, and ecosystem that integrates autonomous and connected vehicles, representing a typical application of next-generation artificial intelligence technology. They are a product of the combination of modern high-tech technologies such as big data, artificial intelligence, electronic control technology, and mobile communication with automobiles and transportation. Based on the fusion of vehicle and roadside perception systems, and through the interaction of information and decision-making control signals with cloud control platform systems, they will be able to travel in multi-lane convoys in the future to achieve the goal of safely and efficiently reaching the destination. In the future, connected vehicles will integrate with smart mobility and smart cities, becoming a new social business model.

[0003] With the increasing mileage of highways and the continuous growth in car ownership, highway traffic safety issues are becoming increasingly serious. Highway tunnels, being relatively enclosed areas, are particularly prone to accidents. Rear-end collisions are the main type of accident in tunnel sections. Accidents involving vehicles transporting hazardous materials, in particular, can cause significant casualties and economic losses. Based on risk assessment models for hazardous materials transport vehicles, the safe driving of special vehicles or key hazardous materials transport vehicles (referred to as key vehicles), especially how to maintain stable acceleration to avoid safety issues caused by inertia and how to maintain a safe distance from the vehicle in front to avoid rear-end collisions, have become important safety issues. Furthermore, with the rapid increase in vehicles on the road, the need for environmental protection and energy conservation is becoming a consensus in the transportation sector. Against the backdrop of achieving carbon neutrality, eco-driving has become another focus of attention.

[0004] In achieving safe and eco-friendly driving, drivers of key vehicles must reasonably predict traffic and road conditions ahead and have a deep understanding of engine dynamics to avoid safety issues caused by inertia—something difficult for human drivers to do. Therefore, a control scheme capable of providing drivers with optimal driving control sequences is needed. Predictive cruise control is considered a promising approach to solving these problems, utilizing relevant heterogeneous traffic environment information and employing adaptive cruise control to optimize connected vehicles. However, traditional predictive cruise control methods typically rely on local traffic information, including upcoming traffic signals and road gradients, and the scope of this information is limited. In the cloud context, key vehicles can utilize not only local road situation information but also global / historical dynamic information from the digital information mapping layer. Furthermore, traditional predictive cruise control methods struggle to meet the computational requirements due to limitations in onboard computing. Recent emerging information and communication technologies, such as vehicular communication, cloud computing, and the Internet of Things (IoT), have enriched the real-time information on transportation networks and, leveraging the concept of cyber-physical systems to handle complex systems, offer significant potential for improving traditional predictive cruise control systems. Therefore, it is necessary to consider layered cloud computing and control methods that can fully handle the transmission and optimization of multi-source heterogeneous information, enabling vehicles to communicate with the cloud and receive cloud suggestions. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a cloud-based predictive cruise control method for key vehicles in highway tunnels based on Cyber-Physical Systems (CPS). This method addresses the problems of traditional predictive cruise control schemes lacking organic collaboration and integration between the cyber and physical layers, and being limited by sensing range and computing power, resulting in very limited information acquisition and processing. It also addresses the safety and energy consumption issues of large hazardous materials transport trucks, especially key vehicles, on highways and in tunnels. This invention proposes a control method that, within a cloud control system architecture, achieves comprehensive improvements in vehicle driving and traffic operation safety, energy efficiency, and traffic flow through the fusion of cyber-layer and physical reality perception, decision-making, and control. This method utilizes parallel and efficient cloud-based computation and vehicle-cloud hierarchical control to regulate the instantaneous optimal control input to achieve safe and eco-friendly driving for key vehicles.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A cloud-based control method for predictive cruise of key vehicles in highway tunnels based on Cyber-Physical Systems (CPS) includes the following steps:

[0008] S1: Physical Layer: Traffic situation information and traffic flow information of highway tunnels are collected by the highway tunnel roadside sensing system and roadside basic unit, and transmitted to the cloud control basic platform by the roadside communication equipment.

[0009] S2: Cloud Control Basic Platform: After matching vehicles and images and filtering duplicate targets, the cloud control basic platform uses digital twins to construct a real-time digital information mapping layer for the multi-source heterogeneous traffic situation of highway sections, and transmits the digital information required by the application layer to the cloud control application platform.

[0010] S3: Data-driven model of cloud control application platform: The data-driven prediction model of improved multi-angle RBF neural network is adopted in cloud control application platform to provide accurate prediction of traffic situation information on current and potential routes for time-varying adaptive algorithm solver in cloud, aiming to improve the prediction and perception capabilities of key vehicles.

[0011] S4: Algorithm solver and cloud control decision system of cloud control application platform: The time-varying adaptive algorithm solver performs unified arrangement, parallel operation and local optimization under the constraints of physical dynamics and key vehicle-specific conditions, and with reference to a series of related indicators such as safety and energy consumption, and transmits the output data to the cloud control decision system.

[0012] S5: Cloud Control Decision System: Combining the traffic situation information from steps S1 to S4 with the calculated and optimized global data, and based on relevant indicators such as optimal traffic safety and energy consumption, the cloud control decision system summarizes the optimal control sequence and provides optimal control schemes for key vehicles and other traffic participants.

[0013] S6: Physical Layer: After the intelligent vehicle terminal platform analyzes and uniformly allocates the optimal control scheme obtained in step S5, the key vehicles and other traffic participants are optimized by the vehicle control computing platform and other systems through actuators to control the power system and braking system of the key vehicles and other traffic participants, thereby achieving the goal of safe and ecological driving of key vehicles in highway tunnels.

[0014] Furthermore, step S1 specifically includes: extracting multi-source heterogeneous traffic situation information of the highway tunnel in real time through the roadside perception system and roadside basic unit, such as traffic flow, congestion level, and traffic flow speed, and transmitting it to the cloud control system basic platform through the information security basic platform via the roadside communication equipment in a wired transmission manner.

[0015] Furthermore, step S2 specifically includes: after performing vehicle-image matching and filtering duplicate targets, constructing a digital information mapping layer in the cloud control platform using real-time traffic situation information of highway tunnels; the present invention uses digital twin and other methods to establish a twin layer model corresponding to the physical layer, which facilitates ensuring the coordination and consistency of information mapping between the cyber layer and the physical layer throughout the entire life cycle.

[0016] In transportation systems, many factors can influence traffic speed, including traffic flow, driving behavior, road conditions, road occupancy, and weather conditions; these factors typically exhibit both spatial and temporal dependencies within the transportation network. To describe this spatiotemporal traffic information for each road segment, a spatiotemporal matrix is ​​used for representation; mathematically, the traffic information matrix representation X can be expressed by a spatiotemporal matrix:

[0017]

[0018] Where t represents time-series data (timestamp), d represents spatial traffic information held by the vehicle detector, and x i,j =[v i,j m i,j e i,j ] T ,i∈[1,d],j∈[1,t],v is the average speed of traffic flow,m is the traffic volume, ande is the elevation information; the spatial traffic information representation matrix at each time period constitutes the information mapping layer.

[0019] Furthermore, in step S3, to further improve the predictive perception capability of key vehicles, a data-driven traffic prediction model based on real-time ground conditions is integrated into the cloud control application platform. This model uses real-time traffic flow information, emergency events, and terrain as reference inputs to provide accurate predictions of current and potential route traffic information for the time-varying adaptive algorithm solver in the cloud.

[0020] The improved multi-angle RBF neural network data-driven prediction model uses real-time traffic flow information, emergency events, and terrain as reference inputs.

[0021] In traffic information prediction, multiple factors are considered. The data for each traffic factor is described as an information representation matrix and trained separately using a general RBF neural network. After training for a single factor is complete, the output is fused into the input of the training behavior for other factors for further training. The model of a single-angle RBF neural network is defined as follows:

[0022]

[0023] Among them, c ij ω ij The centers of the radial basis functions and the weights from the hidden layer to the output layer are respectively, and the K-means clustering method is used to solve them; x is the output of the i-th hidden layer node; q is the number of hidden layer nodes, x n This is the nth spatial traffic information;

[0024] A time-series matrix and a spatial traffic information matrix are used as two inputs to a multi-angle RBF neural network. The output of one angle is used as the input of the next angle to obtain the final prediction output, which is composed of spatial traffic information. Furthermore, cascaded and aggregation functions are used to fuse the predicted traffic information into a single vector, and the final prediction result is solved. It is important to note that, due to limitations of the original traffic data, the spatial traffic information x1, x2, and x3 consists of traffic flow, average speed, and elevation information; elevation information is not involved in the prediction process. Based on this, the data-driven prediction model significantly improves the ability to predict and perceive key vehicles, while the cloud-based algorithm solver also optimizes the original computing power.

[0025] Furthermore, step S4 specifically includes: the time-varying adaptive algorithm solver receives prediction information and traffic situation information from the data-driven model and information mapping layer, including the updated time-varying longitudinal vehicle model, performs unified arrangement calculation and local optimization under a series of related indicators such as physical dynamics and key vehicle-specific constraints, safety and energy consumption, and transmits the output to the cloud control decision system.

[0026] Furthermore, step S5 specifically includes: combining the traffic situation information from steps S1 to S4 with the optimized data calculated by the algorithm solver, and based on relevant indicators such as optimal traffic safety and energy consumption, the cloud control decision system introduces a quadratic programming method in the cloud according to the optimized control input to summarize the optimal control sequence and provide optimal control schemes for key vehicles;

[0027] The main objective of the optimization algorithm solver is to find the optimal control input distribution under physical dynamics and safety constraints, guided by minimizing the objective function, in order to maximize the driving safety of key vehicles and reduce energy consumption and other related indicators. Therefore, the design of control indicators is particularly critical for the predictive cruise cloud control method. The objective function is shown in the following equation:

[0028]

[0029] In this equation, the first and second terms on the right-hand side represent the deviation of the key vehicle from the given target value; the third term represents the vehicle's fuel consumption; and the fourth term represents the consumption of braking torque, which measures the specific inertial safety issues of the key vehicle. ref For reference speed information, l ref For reference vehicle distance information, l is the distance between the two vehicles, q1, q2, f1, f2, r1, r2 are weighting coefficients, and n is the distance between the two vehicles. p It is the prediction interval; v e For the vehicle's speed, T brake Where Δdis is the braking torque, and Δdis is the spatial domain interval. Let N be the fuel consumption rate, N be the current interval, n be the time interval, and (k+i|k) be the predicted value at time k+i for the future time k+i.

[0030] To ensure that key vehicles operate within the safe operating zone, physical constraints, including speed, engine energy, braking energy, and torque, were added during the optimization process; the velocity boundary of the state space is set as follows:

[0031] v min (k+i|k)≤v e (k+i|k)≤v max (k+i|k)

[0032] Among them, v min Let v be the lower boundary of the velocity boundary. max This is the upper boundary of the velocity boundary;

[0033] According to the engine's general characteristic diagram, there are limitations on engine torque and speed, as well as braking torque, within the effective operating range for key vehicles' inertial safety.

[0034] ω min (k+i|k)≤ω engine (k+i|k)≤ω max (k+i|k)

[0035]

[0036]

[0037] Where, ω min ω represents the lower boundary of the engine speed. max This represents the upper limit of the engine speed. This represents the upper boundary of the engine torque. This represents the upper boundary of the braking torque.

[0038] To balance the safety of key vehicles with the traffic efficiency of highways, the distance between key vehicles and the vehicle in front should meet the following requirements:

[0039] l min (k+i|k)≤l(k+i|k)≤l max (k+i|k)

[0040] Specifically, considering that traffic flow speeds may vary significantly under different traffic conditions, an adaptive vehicle spacing parameter is applied, with a minimum longitudinal vehicle spacing of l. min (k+i|k) and maximum longitudinal vehicle spacing l max (k+i|k) can be represented as:

[0041]

[0042]

[0043] Among them, t headway,min t represents the distance from the front of the vehicle to the lower boundary. headway,max The distance from the front of the vehicle to the upper boundary. The traffic flow prediction speed is obtained from the data-driven prediction model, where t is the headway and C is the speed. min With C max For adaptive vehicle distance parameters, their values ​​are proportional to the speed of key vehicles. That is, when the traffic flow has high vehicle speeds, the vehicle distance parameters are set to larger values. The predictive cruise cloud control scheme adopts an online rolling information domain optimization and iterative update control method. That is, the algorithm solver uses a time-varying prediction model in each prediction range and updates this model within the interval, and then optimizes the control input solution.

[0044] Furthermore, step S6 specifically includes: after the intelligent vehicle terminal platform analyzes and uniformly allocates the optimal control sequence scheme obtained in S4 through the parser, the key vehicles and other traffic participants are optimized by the vehicle control computing platform and other systems through actuators to control the power system and braking system of the key vehicles and other traffic participants, thereby achieving the goal of safe and ecological driving of key vehicles in highway tunnels.

[0045] Cloud control does not mean that the cloud has complete control over the vehicle. Rather, the cloud provides optimized and comprehensive control solutions for key vehicles from a smarter and more macroscopic perspective. However, since the on-board control unit of key vehicles has limited adjustment capabilities, the key vehicle's decision-making system, within the safety and efficiency limits, refers to the control solutions provided by the cloud and ultimately makes decisions to control the vehicle based on its actual situation. This is known as vehicle-cloud layered control. Through vehicle-cloud layered control, the cloud control decision-making system provides optimized control solutions for key vehicles and other traffic participants, and transmits the output to the intelligent on-board terminal platform for unified allocation. Ultimately, key vehicles and other traffic participants use the vehicle control computing platform and other systems to optimize the control of their power and braking systems, thereby achieving the goal of safe and eco-friendly driving for key vehicles in highway tunnels.

[0046] The beneficial effects of this invention are as follows:

[0047] 1) This invention incorporates the highway tunnel perception system into the physical layer, expanding the perception range and data capacity of intelligent connected vehicles, and providing richer and more accurate data values ​​and information flow for cloud control system calculation and analysis decision-making.

[0048] 2) The present invention, based on the cloud control system, significantly improves the safety and energy efficiency of key vehicles compared with existing key vehicle driving methods.

[0049] 3) This invention uses a cloud control system as its architecture. Through the macro-control and mutual integration among various platforms of the cloud control system, and the data-driven model and algorithm solver, it provides optimal control strategies for key vehicles. It is an important application of cloud control systems in the transportation field and a typical embodiment of the new generation of intelligent transportation systems.

[0050] 4) Compared with traditional vehicle-mounted computing terminals, this invention deploys tasks with higher computing requirements in high-performance computing nodes in the cloud, thereby improving computing efficiency and accuracy.

[0051] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0052] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0053] Figure 1 This is a simplified flowchart of the structural framework of the present invention;

[0054] Figure 2 This is a diagram of the cloud control system architecture, in which the cyber layer and the physical layer are organically coordinated and integrated.

[0055] Figure 3 This is a diagram of the real-time monitoring interface for highway tunnels.

[0056] Figure 4 This provides the main support for the structure of the present invention, which is based on a layered cloud control architecture for key vehicles using cyber-physical systems.

[0057] Figure 5 To improve the multi-angle RBF neural network for traffic information prediction;

[0058] Figure 6 State space and computational logic diagram of the solver for the time-varying adaptive algorithm;

[0059] Figure 7 A digital mapping layer for constructing highway tunnels and their vehicles in simulation software;

[0060] Figure 8 Speed ​​data of some traffic flow in the tunnel recorded by the tunnel's internal sensing system after being controlled by the cloud control application platform;

[0061] Figure 9To compare the simulation results of the predictive cruise cloud control method with various other driving strategies or algorithms after accidents occur in tunnels and on ramps. Detailed Implementation

[0062] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0063] Please see Figures 1-9 This invention provides a cloud-based predictive cruise control method for key vehicles in highway tunnels based on Cyber-Physical Systems (CPS). Leveraging the high-efficiency computing and optimization control capabilities of cloud control systems, this method designs a cloud-based predictive cruise control method for key vehicles in highway tunnels based on cyber-physical systems (CPS) to improve the predictive perception range and optimization computing capabilities, and enhance the coordination and integration of various levels. Specifically, the cloud control system's basic platform integrates ground-based real-time information from various entities in the transportation system and constructs a digital information mapping layer based on the multi-source heterogeneous traffic situation of highway sections. An improved multi-angle RBF neural network data-driven prediction model is proposed in the cloud control application platform to predict and perceive real-time traffic situation information. Using situational and predictive information under limited constraints, an algorithm solver in the cloud is introduced to uniformly orchestrate and perform parallel computation and optimization of the control objective, thereby achieving the goal of safe and ecological driving for key vehicles in highway tunnels. The specific implementation process of this method includes the following steps:

[0064] S1: Through the roadside basic unit, the highway tunnel perception system extracts real-time traffic situation information of the highway tunnel, such as traffic flow, road occupancy, congestion level, and traffic flow speed. It then transmits this information to the cloud control system basic platform via wired transmission through the roadside communication equipment, awaiting the establishment of the information monitoring and mapping layer.

[0065] S2: After performing vehicle-image matching and filtering for duplicate targets, a digital twin and mapping method is used to establish a twin mapping layer model corresponding to the physical layer. Real-time multi-source heterogeneous traffic situation information from highway tunnels is used to construct a digital information mapping layer within the cloud control platform, such as... Figure 3As shown, the information monitoring and mapping module consists of a multi-source sensing subsystem comprising LiDAR, high-definition cameras, an edge computing server, an industrial Ethernet switch (Ethernet optical transceiver), and related data fusion software. Based on base stations providing full coverage within the tunnel, it can match multi-dimensional information such as the location and characteristics of key vehicles, enabling vehicle-to-base station tracking across the entire tunnel and accurately determining the location of a specified vehicle within the tunnel at any given time. The vehicle-to-cloud information mapping relationship ensures the coordinated consistency of cyber-layer and physical-layer information mapping throughout the entire lifecycle.

[0066] In transportation systems, many factors can influence traffic speed, such as traffic flow, road conditions, road occupancy, and weather conditions. These factors typically exhibit both spatial and temporal dependencies within the transportation network. To describe this spatiotemporal traffic information for each road segment, a spatiotemporal matrix is ​​used for representation. Mathematically, the traffic information matrix representation X can be expressed by a spatiotemporal matrix:

[0067]

[0068] Where t represents time series data (timestamp), d represents spatial traffic information held by highway vehicle detectors, and x i,j =[v i,j m i,j e i,j ] T Let i ∈ [1, d], j ∈ [1, t], v be the average speed of traffic flow, m be the traffic volume, and e be the elevation information. The traffic information representation matrices in each time and spatial domain constitute part of the information mapping layer.

[0069] S3: To further improve the beyond-line-of-sight prediction and perception capabilities of key vehicles, a data-driven traffic prediction model based on real-time ground conditions is integrated into the cloud control application platform. This model uses real-time traffic information, emergency events, and terrain information as reference inputs to provide accurate predictions of current and potential route traffic information for the time-varying adaptive algorithm solver in the cloud.

[0070] In order to consider multiple factors in traffic information prediction, such as Figure 5 As shown, this invention proposes an improved multi-angle RBF neural network: the data for each traffic factor is described as an information representation matrix and trained separately using a general RBF neural network. After training a single factor, the output is fused to the input of the training behavior of other factors for further training. The model of the single-angle RBF neural network in this paper is defined as follows:

[0071]

[0072] In the above formula, c ij ω ijThese are the centers of the radial basis functions and the weights from the hidden layer to the output layer, respectively. In this paper, the K-means clustering method is used to solve for them. q represents the output of the i-th hidden layer node; q represents the number of hidden layer nodes.

[0073] In this invention, a time series matrix and a spatial traffic information matrix are used as two inputs to a multi-angle RBF neural network. The output of one angle is used as the input of the next angle to obtain the final prediction result output 2, which is composed of spatial traffic information. Furthermore, this invention employs cascaded functions and aggregation functions to fuse the predicted traffic information into a single vector and solve for the final prediction result. It should be noted that, limited by the original traffic data, the spatial traffic information x1, x2, and x3 consists of traffic flow, average speed, and elevation information. Elevation information is not involved in the prediction process. Based on this, the data-driven prediction model significantly improves the ability to predict and perceive key vehicles, while the cloud-based algorithm solver also optimizes the original computing power.

[0074] S4: The optimization algorithm solver receives prediction information and traffic conditions from the data-driven model and information mapping layer, including the updated time-varying longitudinal vehicle model. Under the constraints of physical dynamics and key vehicle-specific conditions, it performs unified programming calculations and optimizations on a series of related indicators such as safety and energy consumption, and transmits the output data to the cloud control decision system.

[0075] Based on Newton's second law and the transmission system model, the longitudinal dynamic equilibrium equation for the key vehicle can be expressed as follows:

[0076]

[0077]

[0078]

[0079] Where A is the frontal area of ​​the vehicle, g is the acceleration due to gravity, and v e Let C be the vehicle's speed, ρ be the air density, and C be the air density. D F is the air resistance coefficient, θ is the angle between the road slope and the horizontal plane, and F is the air resistance coefficient. engine For traction force, T engine For engine torque, F brake For braking force, T brake For braking torque, ψ t (n) represents the transmission ratio ψ with a cyclic set. t (n)∈{ψ t (1),ψ t (2),},i0 is the final transmission ratio,η t R is the transmission efficiency, μ is the tire rolling resistance coefficient, and R is the transmission efficiency. eWhere V is the effective radius of the tire, and the wind speed is V. w The value is negative when the rear is facing the wind and positive when the front is facing the wind. M is the vehicle's equivalent mass, including the vehicle's mass and the inertia of the rotating components of the powertrain. The final drive ratio model is constant. The relative longitudinal dynamics of the vehicle in question with respect to the vehicle in front can be described as:

[0080]

[0081]

[0082]

[0083] Among them, v f Let x be the speed of the vehicle in front, l be the distance between the two vehicles, and x = [l, v]. e ] T u = [T engine ,T brake ] T Let D be the disturbance. We integrate the state equation and the control input over [t, t + Δt]. By the mean value theorem for integration, we have:

[0084]

[0085] Where ξ∈[t,t+Δt]. Then, applying the forward Euler method to the control input and the midpoint Euler method to the state variables respectively, the expression of x(t+Δt) in the time domain is obtained as follows:

[0086]

[0087] The travel distance Dis = n·Δdis, and the time interval Δt is the average time required for a key vehicle to travel the distance Δdis. Traffic situation information and vehicle-related parameters are not only time-varying in the time domain but also change with spatial movement in the spatial domain. Therefore, to better ensure the safety and energy efficiency of key vehicles, we use the spatial interval Δdis to transfer the time-domain differential to the spatial domain, obtaining a discrete longitudinal dynamic model of the system:

[0088]

[0089] at this time E is the identity matrix.

[0090] S5: Combining the traffic situation information from steps S1 to S4 with the optimized data calculated by the algorithm solver, and based on relevant indicators such as optimal traffic safety and energy consumption, the cloud-based control decision system summarizes the best control sequence by optimizing the control input through the quadratic programming method in the cloud, and provides the optimal control scheme for key vehicles.

[0091] The main objective of the optimization algorithm solver is to find the optimal control input distribution under physical dynamics and constraints, with the goal of minimizing the cost function, in order to maximize the optimization and improvement of the driving safety of key vehicles and reduce related indicators such as energy consumption. Figure 6 As shown. Therefore, the design of control indicators is particularly crucial for predictive cruise cloud control methods, and the cost / objective function is shown in the following equation:

[0092]

[0093] In the above formula, the first and second terms represent the deviation of the key vehicle from the given target value, the third term represents the vehicle's fuel consumption, and the fourth term represents the consumption of braking torque, which measures the inertial safety issue. q1,q2,f1,f2,r1,r2 are weighting coefficients. p This is the prediction interval. ref For reference speed information, l ref Vehicle distance information for reference.

[0094] To ensure that key vehicles operate within the safe operating zone, physical constraints, including speed, engine energy, braking energy, and torque, were added during optimization. The velocity boundary settings for the state space are as follows:

[0095] v min (k+i|k)≤v e (k+i|k)≤v max (k+i|k)

[0096] According to the engine's general characteristic diagram, there are limitations on engine torque and speed, as well as braking torque, within the effective operating range for key vehicles' inertial safety.

[0097] ω min (k+i|k)≤ω engine (k+i|k)≤ω max (k+i|k)

[0098]

[0099]

[0100] To balance the safety of key vehicles with the traffic efficiency of highways, the distance between key vehicles and the vehicle in front should meet the following requirements:

[0101] l min (k+i|k)≤l(k+i|k)≤l max (k+i|k)

[0102] Specifically, considering that traffic flow speeds may vary significantly under different traffic conditions, this invention applies an adaptive vehicle spacing parameter, with a minimum longitudinal vehicle spacing of l. min (k+i|k) and maximum longitudinal vehicle spacing l max (k+i|k) can be represented as:

[0103]

[0104]

[0105] in, The traffic flow prediction speed is obtained from the data-driven prediction model, where t is the headway and C is the speed. min With C max For adaptive vehicle distance parameters, their values ​​are proportional to the speed of priority vehicles; that is, when traffic flow has high vehicle speeds, the vehicle distance parameters are set to larger values. The predictive cruise cloud control scheme employs an online rolling information domain optimization and iterative update control method. Specifically, the algorithm solver uses a time-varying predictive model within each prediction range, updates this model within the interval, and then optimizes the control input solution.

[0106] S5: The intelligent vehicle terminal platform analyzes and uniformly allocates the control sequence scheme through the parser, and then optimizes the transmission system, power system, braking system, etc. of key vehicles and other traffic participants through the execution controller, thereby achieving the goal of safe and ecological driving of key vehicles in highway tunnels.

[0107] Cloud control is a more macroscopic vehicle control solution. The cloud has stronger processing and computing capabilities, providing a broader perspective and offering optimized and comprehensive control solutions for key vehicles. However, due to the limited control capabilities of the onboard control units of key vehicles, within the safety and efficiency limits, the key vehicle's decision-making system, referencing the cloud-provided control solutions, ultimately makes decisions based on the actual situation of the vehicle to control it—this is known as vehicle-cloud layered control. Based on the concept of vehicle-cloud layered control, this invention provides optimized control solutions for key vehicles and other traffic participants through the cloud-based decision-making system, transmitting the output to an intelligent onboard terminal platform for unified allocation. Ultimately, key vehicles and other traffic participants utilize the vehicle control computing platform and other systems to optimize the control of their power, braking, and transmission systems, thereby achieving the goal of safe and eco-friendly driving for key vehicles in highway tunnels.

[0108] Finally, the cloud control method proposed in this invention was tested and verified. The simulation process and results are as follows: Figures 7-9 As shown, the evaluation results demonstrate the effectiveness and feasibility of this method.

[0109] Any process or method described in the flowcharts or otherwise herein in this invention can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order according to the functions involved, as should be understood by those skilled in the art to which the embodiments of this application pertain.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A cloud-based control method for predictive cruise of key vehicles in highway tunnels based on Cyber-Physical Systems (CPS), characterized in that, The method specifically includes the following steps: S1: Physical layer: Traffic situation information and traffic flow information of highway tunnels are collected by the highway tunnel roadside sensing system and roadside basic unit, and transmitted to the cloud control basic platform by the roadside communication equipment. S2: Cloud Control Basic Platform: After matching vehicles and images and filtering duplicate targets, the digital twin method is used in the cloud control basic platform to construct a real-time digital information mapping layer for the multi-source heterogeneous traffic situation of the highway section, and transmit the digital information required by the application layer to the cloud control application platform. S3: Data-driven model of cloud control application platform: The data-driven prediction model of improved multi-angle RBF neural network is adopted in the cloud control application platform to provide accurate prediction of traffic situation information on current and potential routes for the time-varying adaptive algorithm solver in the cloud. S4: Algorithm solver and cloud control decision system of cloud control application platform: The time-varying adaptive algorithm solver performs unified arrangement of parallel operation and local optimization under the constraints of physical dynamics and key vehicle-specific conditions, with reference to safety and energy consumption indicators, and transmits the output data to the cloud control decision system; S5: Cloud Control Decision System: Combining the traffic situation information from steps S1 to S4 with the global data calculated and optimized by the algorithm solver, based on the optimal traffic safety and energy consumption, the cloud control decision system introduces a quadratic programming method in the cloud according to the optimized control input to summarize the optimal control sequence and provide optimal control schemes for key vehicles and other traffic participants. The goal of the optimization algorithm solver is to find the optimal control input distribution under physical dynamics and safety constraints, guided by minimizing the objective function, in order to maximize the driving safety of key vehicles and reduce energy consumption; the objective function is shown in the following equation: In this equation, the first and second terms on the right side represent the deviation of the key vehicle from the given target value; the third term represents the vehicle's fuel consumption; and the fourth term represents the consumption of braking torque, which measures the specific inertial safety issues of the key vehicle. v ref Speed ​​information for reference, l ref For reference vehicle distance information, The distance between the two vehicles. These are the weighting coefficients. It is the prediction interval; For the vehicle's speed, T brake For braking torque, For spatial domain interval, For fuel consumption rate, N For the current interval, n For time intervals, In order to be in k Always looking towards the future k + i The predicted value at any given time; To ensure that key vehicles operate within the safe operating zone, physical constraints including speed, engine energy, braking energy, and torque were added during the optimization; the speed boundary of the state space is set as follows: in, The lower boundary of the velocity boundary. This is the upper boundary of the velocity boundary; According to the engine's general characteristic diagram, there are limitations on engine torque and speed, as well as braking torque, within the effective operating range for key vehicles' inertial safety. in, This represents the lower boundary of the engine speed. This represents the upper limit of the engine speed. This represents the upper boundary of the engine torque. This represents the upper boundary of the braking torque. To balance the safety of key vehicles with the traffic efficiency of highways, the distance between key vehicles and the vehicle in front should meet the following requirements: Apply adaptive vehicle distance parameters to achieve the minimum longitudinal vehicle distance. and maximum longitudinal distance Represented as: in, The distance from the front of the vehicle to the lower boundary. The distance from the front of the vehicle to the upper boundary. The traffic flow prediction speed obtained from the data-driven prediction model. t For the headway, and For adaptive vehicle distance parameters, their values ​​are proportional to the speed of the key vehicle; the predictive cruise cloud control scheme adopts an online rolling information domain optimization and iterative update control method, that is, the algorithm solver uses a time-varying predictive model in each prediction range and updates this model in the interval, and then performs optimization control input solution; S6: Physical Layer: After the intelligent vehicle terminal platform analyzes and uniformly allocates the optimal control scheme obtained in step S5, the power system and braking system of key vehicles and other traffic participants are optimized by the vehicle control computing platform through actuators, thereby achieving the goal of safe and ecological driving of key vehicles in highway tunnels.

2. The method for predictive cruise cloud control of key vehicles in highway tunnels according to claim 1, characterized in that, Step S2 specifically includes: after performing vehicle-image matching and filtering duplicate targets, constructing a digital information mapping layer in the cloud control platform using real-time traffic situation information of highway tunnels; and using the digital twin method to establish a twin layer model corresponding to the physical layer, so as to ensure the coordination and consistency of information mapping between the cyber layer and the physical layer throughout the entire life cycle. In transportation systems, factors affecting traffic speed include traffic flow, driving behavior, road conditions, road occupancy, and weather conditions. To describe the spatiotemporal traffic information of each road segment, a spatiotemporal matrix is ​​used for representation. Mathematically, the traffic information matrix represents... X Represented by the spacetime matrix: in, t For time series data, d Spatial traffic information held by vehicle detectors, , v The average speed of the traffic flow. m For traffic flow, e The information mapping layer consists of elevation information and spatial traffic information representation matrices for each time period.

3. The method for predictive cruise cloud control of key vehicles in highway tunnels according to claim 1, characterized in that, In step S3, the data-driven prediction model of the improved multi-angle RBF neural network uses real-time traffic flow information, emergency events, and terrain as reference inputs; In traffic information prediction, multiple factors are considered. The data for each traffic factor is described as an information representation matrix and trained separately using a general RBF neural network. After training for a single factor is complete, the output is fused into the input of the training behavior for other factors for further training. The model of a single-angle RBF neural network is defined as follows: in, , The values ​​of the radial basis function center and the weights from the hidden layer to the output layer are respectively used to solve the problem using clustering methods; For the first i The output of each hidden layer node; q This represents the number of hidden layer nodes. For the first n Individual spatial traffic information; A time series matrix and a spatial traffic information matrix are used as two inputs to a multi-angle RBF neural network. The output of one angle is used as the input of the next angle to obtain the final prediction output, which is composed of spatial traffic information. Furthermore, cascade functions and aggregation functions are used to fuse the predicted traffic information into a single vector, and the final prediction result is then solved. It should be noted that, due to limitations of the original traffic data, the spatial traffic information... It consists of traffic flow, average speed, and elevation information; elevation information is not involved in the prediction process.

4. The method for predictive cruise cloud control of key vehicles in highway tunnels according to claim 1, characterized in that, Step S4 specifically includes: The time-varying adaptive algorithm solver receives prediction information and traffic situation information from the data-driven model and information mapping layer, including the updated time-varying longitudinal vehicle model, performs unified arrangement calculation and local optimization under physical dynamics and key vehicle-specific constraints, safety and energy consumption indicators, and transmits the output to the cloud control decision system.

Images