Traffic flow fundamental diagram construction method, system and device of cruise vehicle and storage medium

Abstract

The invention discloses a traffic flow fundamental diagram construction method, system and device for a cruise vehicle and a storage medium, and the method comprises the steps: obtaining the following data of the cruise vehicle, constructing a neural network architecture containing a sub-network, training the neural network through a training data set, and obtaining the traffic flow fundamental diagram of the cruise vehicle; the sub-network is extracted from the trained neural network by minimizing the error between the accelerated speed predicted value and the actually measured accelerated speed, a symbol regression training data set is constructed based on the input and output relation of the sub-network, and the accelerated speed is obtained based on the symbol regression training data set. And establishing a mapping relation among the speed difference, the spacing and the optimal speed. Based on the mapping relation, the traffic flow basic diagram of the cruising vehicle is constructed, and the problems that a pure data driving method based on a neural network depends on a large amount of data and the model interpretability is poor, and a single symbol regression method is low in search efficiency and difficult to process a high-dimensional nonlinear relation can be solved.

Description

Technical Field

[0001] This application relates to the field of intelligent transportation technology, and in particular to a method, system, device and storage medium for constructing a basic traffic flow map for a cruise vehicle. Background Technology

[0002] Traffic flow baseline maps are crucial for analyzing traffic flow characteristics, building traffic models, and optimizing traffic control systems. Obtaining traffic flow baseline maps for Adaptive Cruise Control (ACC) vehicles is essential for predicting their impact on road capacity after large-scale deployment. However, obtaining baseline maps through large-scale real-vehicle road tests during the ACC controller design phase is prohibitively expensive; while conducting only two-vehicle following tests can verify the vehicle's following performance, it cannot assess its impact on overall traffic efficiency after large-scale deployment. If ACC vehicles achieve safe following in real-world applications but lead to a decrease in road capacity, vehicle recalls and algorithm upgrades may be necessary, resulting in significant economic losses.

[0003] Therefore, the core issue at present is that during the verification phase of the ACC vehicle controller, it is difficult to verify its impact on traffic capacity through large-scale road testing due to economic constraints; and if the road traffic capacity is reduced after mass production due to the control algorithm not fully balancing safety and efficiency, it may lead to subsequent recalls and upgrades, causing more serious losses. Summary of the Invention

[0004] To address the aforementioned issues, this application provides a method, system, device, and storage medium for constructing a basic traffic flow map for cruising vehicles.

[0005] The embodiments of this application disclose the following technical solutions:

[0006] The first aspect of this application provides a method for constructing a basic traffic flow graph for a cruising vehicle, including:

[0007] Acquire following data of the cruising vehicle, including speed difference, distance, vehicle speed and acceleration;

[0008] A neural network architecture containing subnetworks is constructed, wherein the input of each subnetwork is the speed difference and the spacing, and the output is the optimal speed; the vehicle's own speed and the optimal speed are respectively connected to the output layer of the neural network through a first weight and a second weight, and together output the acceleration prediction value.

[0009] The neural network is trained using a training dataset to minimize the error between the predicted acceleration and the actual measured acceleration, the training dataset being constructed based on the car-following data;

[0010] Extract the sub-networks from the trained neural network, and construct a symbolic regression training dataset based on the input-output relationship of the sub-networks;

[0011] Based on the symbolic regression training dataset, a mapping relationship between speed difference, spacing and optimal speed is established;

[0012] Based on the mapping relationship, a basic traffic flow graph for the cruise vehicle is constructed.

[0013] In one possible implementation, after acquiring the following data of the cruising vehicle, the method further includes:

[0014] The following data is preprocessed, including removing outliers and noisy data using a moving average method.

[0015] In one possible implementation, establishing the mapping relationship between speed difference, spacing, and optimal speed based on the symbolic regression training dataset includes:

[0016] Using a symbolic regression algorithm based on genetic programming, a mathematical expression representing the relationship between the speed difference, spacing, and optimal speed is searched from a candidate function symbol library. The structure and parameters of the mathematical expression are optimized through genetic operations to improve the data fitting ability of the mathematical expression to the relationship between the speed difference, spacing, and optimal speed.

[0017] In one possible implementation, constructing the basic traffic flow graph of the cruise vehicle based on the mapping relationship includes:

[0018] The traffic flow equilibrium speed of the cruise vehicle is determined based on the mapping relationship.

[0019] Traffic flow density is set as the horizontal axis and traffic volume is set as the vertical axis. The traffic volume is calculated by multiplying the traffic flow density by the equilibrium speed.

[0020] Based on the horizontal and vertical coordinates and the corresponding traffic flow density and traffic volume, a basic traffic flow map of the cruise vehicle is drawn.

[0021] In one possible implementation, the process of constructing the training dataset includes:

[0022] Using the speed difference, distance, and vehicle speed as input data, and the acceleration as output data, a training dataset is constructed for training the neural network.

[0023] In one possible implementation, the subnetwork is at least one of a propagation neural network and a long short-term memory neural network, used to fit a nonlinear relationship through the input speed difference and spacing to output the optimal speed.

[0024] In one possible implementation, training the neural network using a training dataset includes:

[0025] The neural network is trained using a training dataset based on an optimization algorithm and an error loss function. The optimization algorithm includes at least one of gradient descent and adaptive moment estimation, used to adjust the weights and bias parameters of the neural network. The loss function includes at least one of mean squared error and cross-entropy, used to quantify the error between the predicted acceleration and the actual measured acceleration.

[0026] This application provides a traffic flow basic map construction system for cruising vehicles, comprising:

[0027] The acquisition unit is used to acquire the following data of the cruise vehicle, the following data including speed difference, distance, vehicle speed and acceleration;

[0028] The first building unit is used to build a neural network architecture containing sub-networks. The input of the sub-network is the speed difference and the spacing, and the output is the optimal speed. The vehicle's own speed and the optimal speed are respectively connected to the output layer of the neural network through the first weight and the second weight, and together output the acceleration prediction value.

[0029] A training unit is used to train the neural network using a training dataset to minimize the error between the predicted acceleration and the actual measured acceleration, the training dataset being constructed based on the car-following data;

[0030] An extraction unit is used to extract the sub-network from the trained neural network and construct a symbolic regression training dataset based on the input-output relationship of the sub-network.

[0031] A unit is established to create a mapping relationship between speed difference, spacing and optimal speed based on the symbolic regression training dataset.

[0032] The second construction unit is used to construct a basic traffic flow graph of the cruise vehicle based on the mapping relationship;

[0033] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the traffic flow basic map construction method for a cruise vehicle as described in the first aspect above.

[0034] A fourth aspect of this application provides a computer program product that, when run on a computer, executes the traffic flow basic map construction method for cruise vehicles as described in the first aspect above.

[0035] A fifth aspect of this application provides a computer-readable storage medium storing instructions that, when executed on a terminal device, cause the terminal device to perform the traffic flow basic map construction method for a cruise vehicle as described in the first aspect above.

[0036] Compared with the prior art, this application has the following advantages:

[0037] First, by acquiring car-following data from cruising vehicles, the core data required for modeling can be obtained solely based on two-vehicle car-following scenarios, eliminating the need for costly large-scale road tests. This directly addresses the high cost of large-scale road tests. Next, a neural network architecture containing sub-networks is constructed. This network extracts the optimal speed, representing the core of the vehicle car-following strategy, from speed differences and spacing through its sub-network structure. After training, the input-output relationships of the sub-networks are further extracted as training data for symbolic regression. This transforms the black-box neural network into an interpretable explicit mapping between speed differences, spacing, and the optimal speed, while avoiding the complexity of symbolic regression processing raw high-dimensional data. Finally, based on this explicit mapping, the basic traffic flow graph can be theoretically derived, achieving the goal of predicting traffic flow characteristics after large-scale vehicle deployment from low-cost, small-scale car-following data, while ensuring the interpretability of the model. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 A flowchart illustrating a method for constructing a basic traffic flow map for a cruise vehicle, as provided in this application embodiment;

[0040] Figure 2 This is a schematic diagram of adaptive cruise control vehicle following data provided in an embodiment of this application;

[0041] Figure 3 A neural network structure diagram for extracting the optimal speed provided in the embodiments of this application;

[0042] Figure 4 This is a structural diagram of the neural network-symbolic regression fusion model provided in the embodiments of this application;

[0043] Figure 5 This is a schematic diagram of the basic traffic flow map construction result for adaptive cruise vehicles provided in the embodiments of this application;

[0044] Figure 6 This is a structural diagram of a traffic flow basic map construction system for a cruise vehicle provided in an embodiment of this application. Detailed Implementation

[0045] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0046] To facilitate understanding of the technical solutions provided in the embodiments of this application, the background technology involved in the embodiments of this application will be described below.

[0047] As mentioned earlier, Adaptive Cruise Control (ACC) is an Advanced Driver Assistance System (ADAS) function that, in addition to traditional cruise control, automatically detects the speed and distance of vehicles ahead using sensors integrated at the front (such as radar, cameras, or lidar). During the vehicle controller design phase, the inability to obtain basic traffic flow maps in a low-cost and efficient manner makes it impossible to assess road capacity in advance after large-scale commercialization, potentially leading to production recall losses.

[0048] To address the aforementioned issues, this application's embodiments collect car-following data using vehicle sensors and employ moving averages for noise reduction, thereby obtaining high-quality training data at low cost, replacing high-cost, large-scale road testing. A neural network containing sub-networks is designed to uncover the complex nonlinear relationships between inputs such as speed difference and spacing and acceleration, solving the problem of single methods struggling to handle high-dimensional nonlinearity. Sub-networks are extracted from the trained neural network to construct a training dataset for symbolic regression, avoiding direct processing of raw high-dimensional data by symbolic regression and reducing computational complexity. Genetic programming is used to search for the optimal explicit mathematical expression, balancing data fitting ability and interpretability, overcoming the poor interpretability of pure neural networks. Based on the optimal speed function obtained from symbolic regression, a basic graph is generated with traffic flow density as the x-axis and traffic volume as the y-axis, enabling accurate prediction of the impact on road capacity.

[0049] It should be noted that the method, system, device, and medium for constructing a basic traffic flow map for a patrol vehicle provided in this application can be applied to the field of intelligent transportation. The above are merely examples and do not limit the application areas of the method, system, device, and medium for constructing a basic traffic flow map for a patrol vehicle provided in this application. Furthermore, the embodiments of this application may not limit the executing entity for constructing the basic traffic flow map for a patrol vehicle. For example, the method for constructing a basic traffic flow map for a patrol vehicle in the embodiments of this application can be applied to data processing devices such as terminal devices or servers. The terminal device can be an electronic device such as a computer or a personal digital assistant (PDA). The server can be a standalone server, a cloud server, or a cluster server composed of multiple servers.

[0050] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0051] The following embodiment illustrates a method for constructing a basic traffic flow map for a cruise vehicle provided in this application. See also... Figure 1 ,Should Figure 1 A flowchart of a method for constructing a basic traffic flow map for a cruise vehicle, provided in an embodiment of this application, is included.

[0052] S101. Obtain the following data of the cruise vehicle.

[0053] Using sensors installed on vehicles with adaptive cruise control, such as radar, cameras, and vehicle speed sensors, four key parameters are collected to fully describe the vehicle's following behavior: acceleration characterizes the vehicle's dynamic response, speed reflects its own motion state, distance from the vehicle in front quantifies the ability to maintain a safe distance, and speed difference characterizes the relative motion relationship. These parameters together constitute a complete feature set of the vehicle's following behavior. See also Figure 2 , Figure 2 This is a schematic diagram of adaptive cruise vehicle following data provided in an embodiment of this application.

[0054] This approach eliminates the need for large-scale road testing in terms of manpower and resources. It can obtain the key data required for modeling by relying solely on the low-cost scenario of two vehicles following each other, thus directly avoiding the high cost problem of large-scale road testing from the data source.

[0055] In one possible implementation, the data preprocessing stage employs a moving average method to smooth the raw data. This technique serves a dual purpose: firstly, it effectively filters out outliers caused by sensor transient errors or environmental interference; secondly, it eliminates high-frequency noise components in the data. This preprocessing method ensures the accuracy and reliability of the data required for subsequent modeling and provides stable data input for neural network training, thus laying a solid data foundation for building an accurate traffic flow model. This step transforms raw sensor data into standardized data suitable for model training, a crucial prerequisite for the entire method to accurately reproduce vehicle following behavior.

[0056] S102. Construct a neural network architecture containing sub-networks.

[0057] For example, constructing as Figure 3 The neural network structure shown is Figure 3 This is a neural network structure diagram for extracting the optimal speed provided in an embodiment of this application, showing the speed difference. ,spacing ,enter Figure 3 The subnetwork within the dashed box in the shown neural network structure. In this embodiment, the subnetwork uses a backpropagation neural network structure with 10 hidden layer neurons. The subnetwork output is the optimal speed. .speed Through weight Optimal speed Through weight Together and Figure 3 The output neuron of the last layer of the neural network shown is connected, and its output is acceleration. .

[0058] The sub-network takes the speed difference and distance as input and outputs the optimal speed; that is, the sub-network is specifically designed to handle the two key relative motion parameters of speed difference and distance, and its output is defined as the optimal speed. Essentially, this allows the neural network to learn and simulate the internal decision-making process of the adaptive cruise control. The controller calculates in real time, based on its relative state with the vehicle ahead (speed difference and distance), an ideal target speed that it considers safest and most efficient under the current conditions.

[0059] The vehicle's own speed and the optimal speed output by the sub-network are each connected to a single neuron in the output layer through a trainable weight to generate an acceleration prediction. The acceleration prediction = first weight × vehicle's own speed + second weight × optimal speed. These two weight parameters are optimized and adjusted during model training, essentially allowing the neural network to automatically discover an optimal feedback control strategy. Physically, this structure guides the network to learn how to adjust acceleration to continuously approach the optimal speed determined by the current car-following state.

[0060] It provides a clear learning direction for neural networks, significantly improving training efficiency and model convergence speed; it ensures that the acceleration predictions output by the network conform to the basic principles of vehicle dynamics, avoiding the physically unreasonable results that may be produced by pure black box models.

[0061] S103. The neural network is trained using a training dataset to minimize the error between the predicted acceleration value and the actual measured acceleration.

[0062] For example, the overall input data is the speed difference. ,spacing Adaptive cruise control follows the speed of the vehicle. The output data is the acceleration of the vehicle following the adaptive cruise control. .

[0063] By utilizing a large number of samples in the training dataset and employing optimization algorithms such as gradient descent, all parameters in the neural network (including connection weights, bias terms, and the first and second weights within sub-networks) are automatically adjusted to enable the network to have accurate predictive capabilities. By minimizing the error (such as mean square error) between the predicted acceleration values ​​and the actual measured values, the neural network is forced to learn and mimic the acceleration decisions made by a real adaptive cruise controller under corresponding following conditions (speed difference, distance, and its own speed), thereby ensuring that its output following behavior is statistically consistent with the measured data.

[0064] In one possible implementation, the training process of the neural network is achieved through the following technical solution: Based on the aforementioned car-following data, a structured training dataset is constructed, where speed difference, distance, and vehicle speed are used as input data, and acceleration is used as the target output data. Optimization algorithms such as gradient descent or adaptive moment estimation are employed to automatically adjust the weights and bias parameters in the neural network through a backpropagation mechanism. During training, the model parameters are optimized by minimizing the loss function value. The loss function can be in the form of mean squared error or cross-entropy. The difference between the predicted acceleration value obtained from forward inference and the actual measured acceleration value is used as the optimization objective. Through iterative training, the neural network gradually approximates the real vehicle car-following behavior pattern.

[0065] In another possible implementation, the training dataset is constructed using the following technical solution: a dataset establishing an input-output correspondence based on the collected car-following data. Specifically, the three parameters collected at each sampling moment—speed difference, distance, and vehicle speed—are combined into an input sample, and the acceleration collected at the same moment is used as the corresponding output label, thus forming a complete supervised learning dataset. This dataset maintains the temporal correspondence between state variables and control variables during vehicle car-following, providing the necessary data foundation for the neural network to learn the car-following behavior patterns. The size of the dataset can be determined based on the actual amount of data collected, and the model's generalization ability is ensured by dividing it into training, validation, and test sets.

[0066] S104. Extract the sub-network from the trained neural network, and construct a symbolic regression training dataset based on the input-output relationship of the sub-network.

[0067] For example, after completing the S103 training, Figure 3 The subnetworks within the dashed boxes in the shown neural network structure are extracted. For example... Figure 4 As shown, Figure 4 This is a structural diagram of the neural network-symbolic regression fusion model provided in the embodiments of this application, which includes a speed difference. ,spacing As input, optimal speed As output, construct the training dataset for symbolic regression.

[0068] In other words, after training, the subnetwork defined in step two is extracted from the larger network. At this point, this subnetwork is an independent model capable of mapping speed differences and spacing to optimal speeds. Using this extracted subnetwork, a large number of speed difference and spacing combinations are input to obtain the corresponding optimal speed output. This process generates a completely new and clean dataset—the symbolic regression training dataset—which contains the correspondence between (speed difference, spacing, and optimal speed). This step transforms the complex and difficult-to-understand mappings implicit in the neural network subnetworks into a structured and standard mathematical regression problem.

[0069] S105. Based on the symbolic regression training dataset, establish the mapping relationship between speed difference, spacing and optimal speed.

[0070] That is, the difference in speed. ,spacing With optimal speed The relationship between them. It completely opens the black box of neural networks, presenting the following decision logic of the adaptive cruise control as a human-readable, analyzable mathematical formula with physical meaning.

[0071] For example, a symbolic regression algorithm based on genetic programming is used to search for a mathematical expression representing the relationship between the speed difference, the spacing and the optimal speed from a candidate function symbol library; wherein, the candidate function symbol library includes mathematical operators, trigonometric functions, polynomials, exponential functions and logarithmic functions;

[0072] During the search process, genetic operations, including selection, crossover, mutation, and tree-like expression search, are used to continuously evolve the structure and parameters of the mathematical expression to improve its ability to fit the data of the relationship between speed difference, spacing, and optimal speed. At the same time, it is ensured that the explicit mathematical expression obtained after evolution has good interpretability and simplicity. Finally, this mathematical expression, which meets the fitting requirements and has both interpretability and simplicity, is used as the mapping relationship between the speed difference, spacing, and optimal speed.

[0073] For example, such as Figure 4 As shown, the input to the symbolic regression algorithm is the velocity difference. and spacing The output is the optimal speed. .

[0074] Step 5.1: Initialize the evolutionary generation of the symbolic regression algorithm ;

[0075] Step 5.2: Initialize the population for the symbolic regression algorithm ;

[0076] Step 5.3: Evaluate each individual in the population adaptability ;

[0077] Step 5.4: Based on individual adaptability From the father generation Choose the next generation;

[0078] Step 5.5: Genes are produced through gene crossover and mutation. ;

[0079] Step 5.6: Evaluation Each individual ;

[0080] Step 5.7: If individual fitness The number of generations exceeds the set adaptive expectation threshold or the evolutionary number of the symbolic regression algorithm. If the number of generations is less than the set maximum number of generations, then the number of generations for the symbolic regression algorithm... Increase the number of generations and return to step 5.4 until an individual with optimal fitness is obtained, i.e., the individual with the best characteristic speed difference is obtained. ,spacing With optimal speed The structure and parameters of the functional expression of the relation.

[0081] In this embodiment, the optimal speed generated by the symbolic regression algorithm The function expression is as follows:

[0082]

[0083] The values ​​of each parameter in the expression are: , , , , , , , , , , , .

[0084] It should be noted that in the above symbolic regression process, the input variable velocity difference and spacing The measured units are meters per second (m / s) and meters (m), respectively. The optimal velocity function V( ) obtained through the symbolic regression algorithm is... , ) is a holistic mathematical mapping relationship, the function and its parameters c1 to c 12 It was obtained by fitting under the current unit system: that is, when the input variable is the velocity difference. The unit is meters per second (m / s), and the spacing is... When the unit is meters (m), the output value calculated by this function has the physical meaning of the optimal speed, with the unit being meters per second (m / s). Where c1 to c... 12 These are the values ​​of the fitting coefficients that fit this unit system. They work together with the function structure to ensure that, under the current unit system, the entire function can correctly map from the input quantity to the output quantity with the correct physical units and dimensions.

[0085] Therefore, the present application embodiment obtains the above explicit mathematical expression through symbolic regression. This expression, in the form of human-readable mathematical symbols, clearly reveals the quantitative relationship between speed difference, spacing and optimal speed. Compared with the existing neural network black box model that cannot provide internal decision-making logic, its interpretability is fundamentally improved.

[0086] When constructing the basic traffic flow map, the optimal speed value (unit: m / s) calculated by this function must be used. Since the basic traffic flow map usually uses kilometers per hour (km / h) as the unit of speed, and vehicles per kilometer (veh / km) and vehicles per hour (veh / h) as the units of traffic flow density and traffic volume, respectively, the corresponding unit conversions are required when calculating traffic volume.

[0087] Specifically, traffic flow Q (unit: veh / h) is obtained by multiplying traffic flow density K (unit: veh / km) by the equilibrium speed V (unit: km / h), i.e., Q = K × V. The equilibrium speed V_opt calculated by this function is in m / s, and needs to be multiplied by a conversion factor of 3.6 to obtain a speed value in km / h: V(km / h) = V_opt(m / s) × 3.6. Then, substituting this value into the formula Q = K × V yields the correct traffic flow.

[0088] S106. Based on the mapping relationship, construct the basic traffic flow map of the cruise vehicle.

[0089] The correlation between traffic flow equilibrium speed and spacing is determined, that is, the mapping relationship is an explicit mathematical expression constructed by a symbolic regression model that represents the relationship between speed difference, spacing and optimal speed. Since the driving state between vehicles is stable when the traffic flow is in equilibrium and the speed difference tends to be zero, the speed difference in the mapping relationship is set to zero. At this time, the optimal speed output by the expression is the traffic flow equilibrium speed, and thus the correspondence between traffic flow equilibrium speed and spacing is obtained.

[0090] Establish the relationship between traffic flow density and spacing. Based on the basic characteristics of traffic flow, traffic flow density (unit: vehicles / km) represents the number of vehicles per unit road length and is inversely proportional to the average spacing between vehicles (i.e., the higher the traffic flow density, the smaller the average spacing between vehicles, and vice versa). Based on this, establish a quantitative relationship between the two.

[0091] Calculate traffic flow by substituting the equilibrium speed of traffic flow corresponding to each traffic flow density into the core formula: Traffic flow = Traffic flow density × Traffic flow speed, to obtain the traffic flow under different traffic flow densities.

[0092] Traffic flow density is set as the horizontal axis (unit: vehicles / km), and traffic flow is set as the vertical axis (unit: vehicles / hour). Based on the traffic flow density-traffic flow data calculated above, the relationship curve between density and flow is plotted, and finally the basic traffic flow map of the cruise vehicle (adaptive cruise vehicle) is constructed.

[0093] For example, the optimal speed obtained in S105 middle The adaptive cruise vehicle's traffic flow equilibrium speed is obtained. The traffic flow density is then... Set as the x-axis, traffic flow Set the vertical axis as the ordinate. Set the horizontal axis unit to vehicles / km and the vertical axis unit to vehicles / hour. Construct a basic traffic flow map for adaptive cruise control, as shown below. Figure 5 As shown, Figure 5 This is a schematic diagram of the basic traffic flow map construction result for adaptive cruise vehicles provided in this application embodiment, wherein, let Traffic flow basic map, vertical axis: traffic volume Calculated by the following formula: .

[0094] In summary, the beneficial effects of this application are as follows:

[0095] Neural networks, with their front-end nonlinear fitting capabilities, delve into the complex traffic flow relationships within car-following data; symbolic regression, as a back-end analytical tool, transforms the learned patterns into functional expressions with clear physical meaning. This collaborative architecture overcomes the limitations of single methods. Furthermore, since symbolic regression only needs to process the optimal velocity function refined by the neural network, it avoids directly processing the original high-dimensional data, significantly reducing computational complexity and achieving a simultaneous improvement in modeling accuracy and computational efficiency.

[0096] By generating explicit mathematical formulas through symbolic regression, the traffic flow patterns of adaptive cruise vehicles are expressed in a readable and analytical form, giving the model strong interpretability. This feature enables traffic engineers and researchers to gain a deeper theoretical understanding of the intrinsic relationship between vehicle following behavior and macroscopic traffic flow phenomena, providing an intuitive and reliable theoretical basis for traffic management strategy formulation, road planning and design, and vehicle control algorithm optimization.

[0097] When the control algorithm of an adaptive cruise vehicle is upgraded, the model can be retrained based on the newly collected car-following data, ensuring that the predictive performance of the traffic flow baseline map keeps pace with the controller algorithm. This feature ensures the long-term applicability and practical value of this application in the context of rapid technological iteration, providing a powerful evaluation tool for the continuous development and optimization of adaptive cruise technology.

[0098] The above are some specific implementations of the traffic flow basic map construction method for cruise vehicles provided in the embodiments of this application. Based on this, this application also provides a corresponding system for constructing the traffic flow basic map for cruise vehicles. The system provided in the embodiments of this application will be described below from the perspective of functional modularity. Figure 6 This is a structural diagram of a traffic flow basic map construction system for a cruise vehicle provided in an embodiment of this application.

[0099] The system includes:

[0100] The acquisition unit 110 is used to acquire the following data of the cruise vehicle, the following data including speed difference, distance, vehicle speed and acceleration;

[0101] The first building unit 111 is used to build a neural network architecture containing sub-networks. The input of the sub-network is the speed difference and the spacing, and the output is the optimal speed. The vehicle's own speed and the optimal speed are respectively connected to the output layer of the neural network through the first weight and the second weight, and jointly output the acceleration prediction value.

[0102] Training unit 112 is used to train the neural network using a training dataset to minimize the error between the predicted acceleration value and the actual measured acceleration, the training dataset being constructed based on the car-following data;

[0103] Extraction unit 113 is used to extract the sub-network from the trained neural network and construct a symbolic regression training dataset based on the input-output relationship of the sub-network;

[0104] Establishment unit 114 is used to establish a mapping relationship between speed difference, spacing and optimal speed based on the symbolic regression training dataset;

[0105] The second construction unit 115 is used to construct a basic traffic flow map of the cruise vehicle based on the mapping relationship.

[0106] This application also provides corresponding devices and computer storage media for implementing the traffic flow basic map construction scheme for cruise vehicles provided in this application.

[0107] The device includes a memory and a processor. The memory stores instructions or code, and the processor executes the instructions or code to enable the device to perform the traffic flow basic map construction method for cruise vehicles according to any embodiment of this application.

[0108] The computer storage medium stores code, and when the code is run, the device running the code implements the traffic flow basic map construction method for cruise vehicles according to any embodiment of this application.

[0109] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems or apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and relevant parts can be referred to the method section.

[0110] It should be understood that in this application, "at least one" refers to one or more items, and "more" refers to two or more items. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one" or similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, "at least one" of a, b, or c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0111] It should be understood that the terms center, longitudinal, transverse, up, down, front, back, left, right, vertical, horizontal, top, bottom, inside, outside, etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0112] It should be noted that, unless otherwise explicitly specified and limited, the terms installation, connection, and linking should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0113] It should also be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the statement "comprising a..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0114] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0115] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for constructing a basic traffic flow map for cruising vehicles, characterized in that, include: Acquire following data of the cruising vehicle, including speed difference, distance, vehicle speed and acceleration; A neural network architecture containing subnetworks is constructed, wherein the input of each subnetwork is the speed difference and the spacing, and the output is the optimal speed; the vehicle's own speed and the optimal speed are respectively connected to the output layer of the neural network through a first weight and a second weight, and together output the acceleration prediction value. The neural network is trained using a training dataset to minimize the error between the predicted acceleration and the actual measured acceleration, the training dataset being constructed based on the car-following data; Extract the sub-networks from the trained neural network, and construct a symbolic regression training dataset based on the input-output relationship of the sub-networks; Based on the symbolic regression training dataset, a mapping relationship between speed difference, spacing and optimal speed is established; Based on the mapping relationship, a basic traffic flow graph for the cruise vehicle is constructed.

2. The method according to claim 1, characterized in that, After acquiring the following data of the cruise vehicle, the process also includes: The following data is preprocessed, including removing outliers and noisy data using a moving average method.

3. The method according to claim 1, characterized in that, The process of establishing a mapping relationship between speed difference, spacing, and optimal speed based on the symbolic regression training dataset includes: Using a symbolic regression algorithm based on genetic programming, a mathematical expression representing the relationship between the speed difference, spacing, and optimal speed is searched from a candidate function symbol library. The structure and parameters of the mathematical expression are optimized through genetic operations to improve the data fitting ability of the mathematical expression to the relationship between the speed difference, spacing, and optimal speed.

4. The method according to claim 1, characterized in that, The construction of the basic traffic flow map for the cruise vehicle based on the mapping relationship includes: The traffic flow equilibrium speed of the cruise vehicle is determined based on the mapping relationship. Traffic flow density is set as the horizontal axis and traffic volume is set as the vertical axis. The traffic volume is calculated by multiplying the traffic flow density by the equilibrium speed. Based on the horizontal and vertical coordinates and the corresponding traffic flow density and traffic volume, a basic traffic flow map of the cruise vehicle is drawn.

5. The method according to claim 1, characterized in that, The process of constructing the training dataset includes: Using the speed difference, distance, and vehicle speed as input data, and the acceleration as output data, a training dataset is constructed for training the neural network.

6. The method according to claim 1, characterized in that, The sub-network is at least one of a propagation neural network and a long short-term memory neural network, used to fit a nonlinear relationship through the input speed difference and spacing to output the optimal speed.

7. The method according to claim 1, characterized in that, Training the neural network using a training dataset includes: The neural network is trained using a training dataset based on an optimization algorithm and an error loss function. The optimization algorithm includes at least one of gradient descent and adaptive moment estimation, used to adjust the weights and bias parameters of the neural network. The loss function includes at least one of mean squared error and cross-entropy, used to quantify the error between the predicted acceleration and the actual measured acceleration.

8. A traffic flow basic map construction system for cruising vehicles, characterized in that, include: The acquisition unit is used to acquire the following data of the cruising vehicle, the following data including speed difference, distance, vehicle speed and acceleration; The first building unit is used to build a neural network architecture containing sub-networks. The input of the sub-network is the speed difference and the spacing, and the output is the optimal speed. The vehicle's own speed and the optimal speed are respectively connected to the output layer of the neural network through the first weight and the second weight, and together output the acceleration prediction value. A training unit is used to train the neural network using a training dataset to minimize the error between the predicted acceleration and the actual measured acceleration, the training dataset being constructed based on the car-following data; An extraction unit is used to extract the sub-network from the trained neural network and construct a symbolic regression training dataset based on the input-output relationship of the sub-network. A unit is established to create a mapping relationship between speed difference, spacing and optimal speed based on the symbolic regression training dataset. The second construction unit is used to construct a basic traffic flow map of the cruise vehicle based on the mapping relationship.

9. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method for constructing a basic traffic flow map of a cruise vehicle as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed on a terminal device, cause the terminal device to perform the traffic flow basic map construction method for a cruise vehicle as described in any one of claims 1-7.

Images