An Optimization Method for Multi-Objective Path Planning Driven by Implicit Function of a Multi-Channel Network

Abstract

The present invention discloses an optimization method for multi-objective path planning driven by implicit functions of a multi-channel network, belonging to the technical field of algorithm optimization. The method includes obtaining traffic data and performing preprocessing; extracting features of the traffic flow density signal in the traffic data from the time domain dimension and the frequency domain dimension; constructing a three-channel network model and training it; designing an objective function; and using a genetic algorithm to search for the optimal solution. By adopting the above-mentioned optimization method for multi-objective path planning driven by implicit functions of a multi-channel network, different types of data are processed simultaneously through three parallel processing channels, and integrated and output through a fusion layer, accurately reflecting the complexity and dynamic changes of the traffic system, improving the adaptability and accuracy of path planning, and significantly enhancing the system's processing ability and comprehensive expression ability for different data types; by processing complex non-linear relationships in the form of implicit functions and embedding them into the objective function of the optimization algorithm, the accuracy of path planning is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of algorithm optimization, and in particular to a multi-objective path planning optimization method driven by an implicit function of a multi-channel network. Background Art

[0002] With the increasing complexity of urban transportation networks, intelligent optimization algorithms, as an indispensable core technology in intelligent transportation systems, undertake important functions such as path generation, traffic flow optimization and real-time adjustment, which directly affect many aspects of path planning, such as accuracy, efficiency, robustness and cost.

[0003] Traditional path planning algorithms are mostly based on static road network models, ignoring the time series characteristics and dynamic changes of traffic flow. Although some algorithms adjust path selection by updating real-time traffic information, this method cannot fully mine data information, resulting in less than optimal path planning results. How to effectively extract key features from multiple dimensions of massive time series data and establish an accurate prediction model will be the key to improving path planning accuracy.

[0004] At the computational level of optimization algorithms, most optimization methods of traditional path planning algorithms are based on classical mathematical models, such as graph theory and shortest path algorithms. However, these methods often lack adaptability and computational efficiency when faced with complex nonlinear and large-scale data features, making it difficult to achieve ideal path planning results in actual traffic management. Modern optimization algorithms, such as genetic algorithms and particle swarm algorithms, have improved the optimization effect to a certain extent, but their computational overhead is large and they are still insufficient when dealing with complex nonlinear optimization problems. To this end, it is necessary to introduce a more advanced algorithm framework that can effectively explore the potential laws of the traffic system. Although deep learning technology can extract features from large amounts of data, it is not suitable to directly use deep learning models for path planning. How to use deep learning models to assist optimization algorithms is another problem that needs to be solved urgently.

[0005] In addition, the factors affecting path planning are extremely complex and diverse, and the data types involved are different. Conventional network models often lead to excessive model complexity, reduced computational efficiency, and weak generalization capabilities when training all features at once. Due to the different distribution forms of different data types, it is difficult for a single network structure to effectively process these heterogeneous data, which in turn affects the stability and accuracy of path planning. Therefore, how to build a network model that can handle multi-dimensional, heterogeneous data and has strong adaptability has become a key challenge to improve the accuracy and efficiency of path planning. Summary of the invention

[0006] The object of the present invention is to provide an optimization method for multi-objective path planning driven by implicit functions of a multi-channel network, which can effectively integrate and process complex multi-source non-linear data through deep learning technology, optimize the objective function, and improve the accuracy of path planning.

[0007] To achieve the above object, the present invention provides an optimization method for multi-objective path planning driven by implicit functions of a multi-channel network, comprising the following steps:

[0008] S1. Obtain traffic data and perform preprocessing;

[0009] S2. Extract features from the traffic flow density signals in the time domain and frequency domain; D Signal extraction;

[0010] S3. Construct a three-channel network model;

[0011] S4. Train the three-channel network model;

[0012] S5. Design the objective function by combining S2 and S4;

[0013] S6. Use the genetic algorithm to search for the optimal solution.

[0014] Preferably, in S1, the traffic data specifically includes: path types T Containing Classes; environmental conditions C Indicating the influence of external factors, containing Levels; traffic management modes M Containing Types; speed limits V Set to To Discrete values between; traffic flow density D Refers to N Traffic flow density sampling points under different road conditions, sampled at equal intervals within each traffic flow cycle, with the unit of vehicles per hour; and the corresponding actual road loss values L ;

[0015] The preprocessing of traffic data is specifically: filling in missing values, dealing with outliers, ensuring the integrity and accuracy of the data set. The final complete data set includes not less than one sample, and each sample includes path types T , environmental conditions C , traffic management modes M , speed limits V , traffic flow density D , and the corresponding actual road loss values L .

[0016] Preferably, in S2, the traffic flow densityD The time-domain features in the signal include the maximum value, minimum value, sharpness, symmetry, skewness, peak-to-peak value, absolute peak-to-average ratio, average value, and standard deviation. The time-domain features reflect the amplitude, fluctuation degree, and distribution characteristics of the signal; the frequency-domain features include the main frequency, spectral energy, spectral center, frequency bandwidth, harmonic energy ratio, and harmonic energy concentration; there are a total of 15 time-domain and frequency-domain features.

[0017] Preferably, in S3, the construction of the three-channel network model is specifically as follows:

[0018] S31. Normalize the 15 time-domain and frequency-domain features in S2 and the vehicle speed limit in S1 V as numerical features. The calculation formula is:

[0019] ;

[0020] where X represents the original data, is the value after normalization. The normalization interval is [0, 1]. The feature vector form finally input to the numerical channel is where are the 15 time-domain and frequency-domain features and the vehicle speed limit in S1 after their respective normalizations V ;

[0021] S32. The traffic management mode M , path type T , and environmental conditions C are used as categorical features. One-hot encoding is used to convert the values of each category into binary vectors, where each category is represented by an independent dimension;

[0022] The path type T contains categories. After encoding, a binary vector with a length of is obtained; the environmental conditions C contains levels. After encoding, a binary vector with a length of is obtained; the traffic management mode M contains types. After encoding, a binary vector with a length of is obtained; the above three binary vectors are concatenated to obtain the feature vector ，

[0023] input to the categorical channel;

[0024] S33. The traffic flow density DIt has obvious temporality and is classified as a temporal feature. The long short-term memory network is used to capture the long-term dependencies in the time series; the feature vector input to the temporal channel is composed of D the N number of sampling points ; ;

[0025] S34. The numerical channel and the categorical channel process the features input respectively through a fully connected layer,

[0026] ;

[0027] ;

[0028] Among them, is the numerical channel weight matrix, with a dimension of 16×64; is the numerical channel bias term, with a dimension of 64, representing the number of neurons in the output layer; ReLU is the activation function; is the categorical channel weight matrix, with a dimension of ; is the categorical channel bias term, with a dimension of 64; and are the outputs of the fully connected layers of their respective channels, in the form of numerical vectors, representing the feature information after ReLU transformation;

[0029] S35. The temporal channel is processed by the long short-term memory network LSTM to obtain a sequence of hidden states. At each time step , the output hidden state of the LSTM depends on the current input and the state at the previous moment , specifically:

[0030] ;

[0031] Among them, is the cell state, responsible for storing long-term memory; the LSTM layer outputs a sequence of hidden states N with a length of . The size of the hidden layer is set to 64, and the hidden state at the last time step of the LSTM output sequence is selected as the summary representation of the entire temporal data;

[0032] The hidden state at the last time step of the LSTM is mapped to the same feature space dimension as the numerical channel and the categorical channel through a fully connected layer, specifically:

[0033] ;

[0034] Among them, is the output of the time-series channel fully connected layer, and are the weight matrix and bias term of the time-series channel fully connected layer respectively;

[0035] S36. Introduce a cross-channel attention mechanism, calculate the importance of the output features of the numerical channel, categorical channel, and time-series channel, and fuse them in a weighted manner. The fused feature is:

[0036] ;

[0037] Among them, represents the final feature, in the form of a numerical vector, is the attention weight of the three channels, calculated by a scoring network with learning ability;

[0038] S37. The fused feature is mapped to the final output space through a final fully connected layer to obtain the final road loss prediction value :

[0039] ;

[0040] Among them, is the weight matrix of the final fusion layer, with a dimension of 192×1, is the bias term of the final fusion layer.

[0041] Preferably, in S4, specifically: divide the dataset into a training set and a test set at a ratio of 8:2, and use cross-validation to verify the performance of the three-channel network model. During the training process, the training set is used to update the parameters of the three-channel network model, and the loss function is used to measure the difference between the predicted value and the actual target value of the three-channel network model. The loss function is defined as follows:

[0042] ;

[0043] Among them, U is the number of samples. By minimizing this loss function, the three-channel network model can gradually adjust the parameters to reduce the error between the road loss prediction value and the actual road loss value L .

[0044] Preferably, in S5, the design of the objective function includes the following steps:

[0045] S51. Combine the three-channel network model trained in S4 with the traffic flow density DFeature extraction module, regarded as the calculated value of road loss degree Function F:

[0046] ;

[0047] S52. Calculate traffic mobility , Traffic mobility refers to the degree of vehicle flow in the traffic system, related to the speed limit on the road section V and traffic flow density D peak value within the cycle and is calculated by the formula:

[0048] ;

[0049] S53. Perform weighted combination of traffic mobility and the calculated value of road loss degree to provide a comprehensive evaluation criterion for the optimization algorithm. The design of the comprehensive objective function Z is as follows:

[0050] ;

[0051] Among them, is the weight coefficient, used to control the relative importance of traffic mobility and the calculated value of road loss degree in the path planning optimization and adjust according to the traffic conditions.

[0052] Preferably, in S6, the specific steps are as follows:

[0053] S61. Randomly generate a group of path planning schemes, calculate the objective function Z , and evaluate the fitness;

[0054] S62. Select excellent individuals as parents based on the fitness, and through the crossover operation, recombine the genetic information of the parents to generate new offspring individuals;

[0055] S63. Randomly mutate some individuals, replace the old population with the newly generated offspring to form the next generation population;

[0056] S64. When one of the two conditions of reaching the maximum number of iterations and the fitness reaching the preset threshold is satisfied, the algorithm terminates and outputs the optimal solution, that is, the optimal combination of each feature value.

[0057] Therefore, the present invention adopts the above-mentioned multi-objective path planning optimization method driven by the implicit function of a multi-channel network, and has the following beneficial effects:

[0058] (1)Multi-dimensional feature extraction: The present invention performs multi-dimensional feature extraction on traffic flow density signals to ensure that the extracted features are highly representative and effective. Feature extraction not only covers the basic statistical information of the signals but also includes more advanced morphological features, temporal variations, and fluctuation patterns, etc. These multi-dimensional features provide reliable input data for subsequent optimization algorithms, directly affecting the accuracy and stability of the path planning system, improving the model prediction ability and optimization performance, and providing a more forward-looking decision-making basis for path planning.

[0059] (2)Multi-channel network model construction: By designing three parallel processing channels, including a numerical channel, a categorical channel, and a temporal channel, it can process data from different types simultaneously. Each channel independently processes specific types of data, and then the results of each channel are integrated in the fusion layer to obtain a comprehensive output. This multi-channel structure can more accurately reflect the complexity and dynamic changes of the traffic system, improving the adaptability and accuracy of path planning. Compared with the traditional single-channel neural network model, this method significantly enhances the system's processing ability and comprehensive expression ability for different types of data.

[0060] (3)Implicit function-driven objective optimization: During the optimization process, the neural network model processes complex non-linear relationships through implicit functions and embeds this implicit non-linear relationship into the objective function of the optimization algorithm, which can effectively make up for the deficiencies of traditional optimization algorithms in dealing with complex non-linear problems in the traffic system, thereby improving the accuracy of path planning and fully exploring the potential regularities of the traffic system.

[0061] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Description of the Drawings

[0062] Figure 1 It is a schematic flow chart of a method for optimizing multi-objective path planning driven by an implicit function of a multi-channel network according to the present invention. Detailed Embodiment

[0063] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0064] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meanings as understood by those of ordinary skill in the art to which the present invention pertains. The terms "first", "second" and similar words used in the present invention do not denote any order, quantity or importance, but are only used to distinguish different components. Words such as "comprising" or "including" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. Words such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right", etc. are only used to represent relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0065] Embodiment 1

[0066] The present invention provides an optimization method for multi-objective path planning driven by implicit functions of a multi-channel network. The process is as Figure 1 shown and specifically includes the following steps:

[0067] S1. Obtain traffic data and perform preprocessing;

[0068] In the optimization process of path planning, the road loss degree is crucial. The road loss degree is affected by various factors, mainly including speed limits V , traffic flow density D , management mode M , path type T and environmental conditions C . These factors interact and superimpose on each other, presenting a complex non-linear relationship. The path type T includes categories; the environmental conditions C represent the influence of external factors (such as weather, temperature, etc.) and include levels; the traffic management mode M includes types; the speed limit V directly affects the energy efficiency and traffic capacity of the road and is set as to discrete values; the traffic flow density D is a dynamic variable that changes with time and reflects the usage of the road at different time periods. During peak hours, the traffic flow density will increase significantly, affecting traffic flow and road loss. The traffic flow density D includes N traffic flow density sampling points under different road conditions. Sampling is performed at equal intervals within each traffic flow cycle, with the unit of vehicles per hour. Through the above traffic data, the actual value of the corresponding road loss degree L。

[0069] The preprocessing of traffic data specifically includes: filling in missing values, dealing with outliers, ensuring the integrity and accuracy of the data set. The final complete data set includes no less than one sample, and each sample includes path type T , environmental conditions C , traffic management mode M , speed limit V , traffic flow density D and the corresponding road loss degree L 。

[0070] S2. Extract features from the traffic flow density D signal from the time domain and frequency domain dimensions. Through the combination of multi-dimensional features, the dynamic characteristics of the traffic flow density D signal can be more comprehensively reflected.

[0071] The time domain features include maximum value, minimum value, sharpness, symmetry, skewness, peak-to-peak value, absolute peak-to-average ratio, average value and standard deviation. The time domain features reflect the amplitude, fluctuation degree and distribution characteristics of the signal; the frequency domain features include main frequency, spectral energy, spectral center, frequency bandwidth, harmonic energy ratio and harmonic energy concentration; there are a total of 15 time domain features and frequency domain features.

[0072] S3. Build a three-channel network model, which specifically includes the following steps:

[0073] First, process and transform the traffic data to meet the requirements of the input of the three-channel network model.

[0074] S31. The 15 time domain features and frequency domain features in S2 and the speed limit in S1 V all belong to discrete numerical values within a relatively large value range. They are normalized as numerical features to prevent the situation that gradient descent is difficult to converge or the convergence speed is slow during the training process. The calculation formula is:

[0075] ;

[0076] Among them, X represents the original data, is the value after normalization. The normalization interval is [0, 1]. The final feature vector form input to the numerical channel is , where are the 15 time domain features and frequency domain features after normalization and the speed limit in S1 respectively V .

[0077] S32. Traffic management mode M , path type T and environmental conditions CTake on finite discrete values. As categorical features, use one-hot encoding to transform the values of each category into a binary vector, where each category is represented by an independent dimension;

[0078] Path type T Contains categories, then after encoding, a binary vector of length is obtained; Environmental conditions C Contains levels, then after encoding, a binary vector of length is obtained; Traffic management mode M Contains types, then after encoding, a binary vector of length is obtained; Concatenate the above three binary vectors to obtain the feature vector input to the categorical channel ，

[0079] .

[0080] S33, Traffic flow density D Has obvious temporal characteristics and is classified as a temporal feature. Use a long short-term memory network to capture the long-term dependencies in the time series; The feature vector input to the temporal channel Consists of N sampling points of the traffic flow density and .

[0081] Then integrate different types of data into different input channels of the three-channel model.

[0082] S34, The numerical channel and the categorical channel process the features input respectively through a fully connected layer,

[0083] ;

[0084] ;

[0085] Among them, Is the weight matrix of the numerical channel, with a dimension of 16×64; Is the bias term of the numerical channel, with a dimension of 64, representing the number of neurons in the output layer; ReLU is the activation function; Is the weight matrix of the categorical channel, with a dimension of ; Is the bias term of the categorical channel, with a dimension of 64; And is the output of the fully connected layer for each respective channel, in the form of a numerical vector, representing the feature information after ReLU transformation; since the data input to the numerical channels is normalized and the data input to the categorical channels is one-hot encoded, there is no structural relationship or natural sequential relationship among the data features. Therefore, it is efficient and simple to use a single fully connected layer to process both numerical and categorical features, mapping the numerical and categorical features to a suitable space through a single fully connected layer.

[0086] S35. The temporal channels are processed by a long short-term memory network (LSTM) to obtain a sequence of hidden states. At each time step , the output hidden state of the LSTM depends on the current input and the state at the previous time step , specifically:

[0087] ;

[0088] where is the cell state, responsible for storing long-term memory; the LSTM layer outputs a sequence of hidden states N of length . Setting the size of the hidden layer to 64, the hidden state at the last time step of the LSTM output sequence is selected as the summary representation of the entire temporal data, carrying all the information of the temporal data;

[0089] To obtain a unified feature representation, the hidden state at the last time step of the LSTM is mapped to the same feature space dimension as the numerical and categorical channels through a fully connected layer, specifically:

[0090] ;

[0091] where is the output of the fully connected layer for the temporal channels, and are the weight matrix and bias term of the fully connected layer for the temporal channels, respectively.

[0092] S36. To enable the three-channel model to adaptively allocate information from the three channels and help the three-channel model adjust its contribution according to the importance of each channel, a cross-channel attention mechanism is introduced to calculate the importance of the output features of the numerical, categorical, and temporal channels and fuse them in a weighted manner. The fused feature is:

[0093] ;

[0094] where Represents the final feature, in the form of a numerical vector, which are the attention weights for three channels and are calculated by a learnable scoring network.

[0095] S37. The fused feature is mapped to the final output space through a final fully connected layer to obtain the final predicted value of the road loss degree :

[0096] ;

[0097] Among them, is the weight matrix of the final fusion layer, with a dimension of 192×1, and is the bias term of the final fusion layer.

[0098] S4. Train the three-channel network model;

[0099] The dataset is divided into a training set and a test set in a ratio of 8:2, and cross-validation is used to verify the performance of the three-channel network model. During the training process, the training set is used to update the parameters of the three-channel network model, and the loss function is used to measure the difference between the predicted value and the actual target value of the three-channel network model. The loss function is defined as follows:

[0100] ;

[0101] Among them, U is the number of samples. By minimizing this loss function, the three-channel network model can gradually adjust the parameters to reduce the error between the predicted value of the road loss degree and the actual value of the road loss degree L . The Adam optimizer is used to optimize the parameters of the three-channel network model, which can effectively handle high-dimensional data and accelerate convergence.

[0102] S5. Design the objective function by combining S2 and S4;

[0103] In the design of the objective function, traffic mobility and the calculated value of the road loss degree are two core indicators that affect path selection and directly affect the direction and accuracy of the optimization result. The calculated value of the road loss degree represents the energy loss or the decrease in traffic capacity caused by various road conditions during the use of the road, and is closely related to the speed limit V , traffic flow density D , management mode M , path type T and environmental conditions C . The design of the objective function includes the following steps:

[0104] S51. Incorporate the three-channel network model trained in S4 and combine it with the traffic flow density D Feature extraction module, regarded as the calculated value of road loss of function F:

[0105] ;

[0106] Embed the deep learning model F as an implicit function into the objective function, process different types of traffic data, and calculate the calculated value of road loss with high precision to improve the accuracy of the optimization objective.

[0107] S52. Calculate traffic mobility , where traffic mobility refers to the degree of vehicle flow in the traffic system, and is related to the speed limit on the road section V and the traffic flow density D peak value within the cycle . Improving traffic mobility helps to reduce congestion, improve traffic efficiency, and accelerate the passing speed of the path. The calculation formula is:

[0108] .

[0109] S53. Improving traffic mobility contributes to the efficiency of path planning, but excessive optimization or insufficient optimization may both lead to an increase in the calculated value of road loss . Combine traffic mobility and the calculated value of road loss by weighted combination to provide a comprehensive evaluation criterion for the optimization algorithm. The design of the comprehensive objective function Z is:

[0110] ;

[0111] where is the weight coefficient, used to control the relative importance of traffic mobility and the calculated value of road loss in path planning optimization, and adjust according to the traffic situation.

[0112] S6. Use the genetic algorithm to search for the optimal solution;

[0113] In path planning optimization, the goal is to simultaneously maximize traffic mobility and minimize the calculated value of road loss , that is, to maximize the comprehensive objective function Z . The constraint conditions are mainly the value range limitations of each feature. The genetic algorithm is used to search for the optimal solution.

[0114] S61. Randomly generate a set of path planning schemes, calculate the objective function Z , and evaluate the fitness. The higher the fitness, the closer the path planning scheme is to the optimal solution.

[0115] S62. Select excellent individuals as parents based on the fitness, and through the crossover operation, recombine the genetic information of the parents to generate new offspring individuals.

[0116] S63. Randomly mutate some individuals. The mutation operation changes some gene values of the individuals, thereby exploring a wider solution space, and replacing the old population with the newly generated offspring to form the next generation population.

[0117] S64. When at least one of the two conditions of reaching the maximum number of iterations and the fitness reaching the preset threshold is satisfied, the algorithm terminates and outputs the optimal solution, that is, the optimal combination of the values of each feature.

[0118] Through S6, under the given constraint conditions, obtain the feature combination that maximizes the comprehensive objective function Z , so as to achieve the optimal path planning. In practical applications, the speed limit V , management mode M and environmental conditions C are usually prior values, and these values can be set according to the actual situation or obtained by system collection. The traffic flow density D can be obtained in real time through the roadside unit. Therefore, by fixing the prior features, the optimization system can be used to search for the path type that optimizes the core indicators T , so as to achieve the optimal path planning.

[0119] Therefore, the present invention adopts the above-mentioned multi-objective path planning optimization method driven by the implicit function of a multi-channel network, processes different types of data simultaneously through three parallel processing channels, and performs integrated output through the fusion layer, accurately reflecting the complexity and dynamic changes of the traffic system, improving the adaptability and accuracy of path planning, and significantly enhancing the system's processing ability and comprehensive expression ability for different data types. During the parameter optimization process, complex non-linear relationships are processed in the form of implicit functions and embedded into the objective function of the optimization algorithm, effectively making up for the deficiencies of traditional optimization algorithms in dealing with complex non-linear problems in the traffic system, improving the accuracy of path planning, and fully exploring the potential regularities of the traffic system.

[0120] 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 them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions of the present invention or make equivalent replacements, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An optimization method for multi-objective path planning driven by implicit functions of a multi-channel network, characterized in that, It includes the following steps: S1. Obtain traffic data and perform preprocessing. Each sample includes the path type T , environmental conditions C , traffic management mode M , speed limit V , traffic flow density D , and the actual value of the road loss degree corresponding thereto L ; S2. Extract features from the traffic flow density D signal in the time domain and frequency domain. The time domain features in the traffic flow density D signal include maximum value, minimum value, sharpness, symmetry, skewness, peak-to-peak value, absolute peak-to-average ratio, average value, and standard deviation. The time domain features reflect the amplitude, fluctuation degree, and distribution characteristics of the signal; The frequency-domain features include the main frequency, spectral energy, spectral center, frequency bandwidth, harmonic energy ratio, and harmonic energy concentration; There are a total of 15 time-domain features and frequency-domain features; S3. Take the time-domain features and frequency-domain features in S2 and the speed limit in S1 V as numerical features, the traffic management mode in S1 M , path type T and environmental conditions C as categorical features, and the traffic flow density in S1 D as a time-series feature to construct a three-channel network model, specifically: S31. Normalize the numerical features, and the calculation formula is: ； Among them, X represents the original data, is the value after normalization, and the normalization interval is [0, 1]. The feature vector form finally input into the numerical channel is , where are respectively the 15 time-domain features, frequency-domain features after their respective normalizations, and the vehicle speed limit in S1 V ; S32. Use one-hot encoding for the categorical features to convert the values of each category into binary vectors, where each category is represented by an independent dimension; Path type T contains classes, and after encoding, a binary vector with a length of is obtained; environmental conditions C contains levels, and after encoding, a binary vector with a length of is obtained; traffic management modes M contains types, and after encoding, a binary vector with a length of is obtained; the above three binary vectors are concatenated to obtain the feature vector of the input classification channel ， ； S33. For time-series features, use a long short-term memory network to capture long-term dependencies in the time series; the feature vectors input into the time-series channel are composed of D the N number of sampling points and ; S34. The numerical channel and the categorical channel process the features input by each through a fully connected layer respectively, ； ； Among them, is a numerical channel weight matrix with a dimension of 16×64; is a numerical channel bias term with a dimension of 64, representing the number of neurons in the output layer; ReLU is an activation function; is a categorical channel weight matrix with a dimension of ; is a categorical channel bias term with a dimension of 64; and are the outputs of the fully connected layers of their respective channels, in the form of numerical vectors, representing the feature information after ReLU transformation; S35. The time-series channel is processed by a long short-term memory network (LSTM) to obtain a sequence of hidden states. At each time step , the output hidden state of the LSTM depends on the current input and the state at the previous time step , specifically: ； Among them, is the cell state, responsible for storing long-term memories; the output of the LSTM layer is a hidden state sequence of length N ; set the size of the hidden layer to 64, and select the hidden state at the last time step of the LSTM output sequence as the summary representation of the entire time series data; ​ The hidden state of the last time step of the LSTM is mapped through a fully connected layer to the same feature space dimension as the numerical channels and categorical channels, specifically: ； Among them, is the output of the time-series channel fully connected layer, and are the weight matrix and bias term of the time-series channel fully connected layer respectively; S36. Introduce a cross-channel attention mechanism to calculate the importance of the output features of the numerical channel, the categorical channel, and the time-series channel, and fuse them in a weighted manner. The fused features are: ； Among them, represents the final feature, in the form of a numerical vector, which are the attention weights for three channels and are calculated by a learnable scoring network; S37, the fused features are mapped to the final output space through a final fully connected layer to obtain the final predicted value of the path loss degree : ； Among them, is the weight matrix of the final fusion layer, with a dimension of 192×1, is the bias term of the final fusion layer; S4. Train the three-channel network model; S5. Design the objective function by combining S2 and S4; specifically, it includes the following steps: S51. Combine the three-channel network model trained in S4 with the traffic flow density D Feature extraction module, regarded as the calculated value of road loss Function F of ； S52. Calculate traffic mobility , traffic mobility refers to the degree of vehicle flow in the traffic system, and is related to the speed limit on the road section V and traffic density D the peak value within the cycle is related, and the calculation formula is as follows: ； S53. Combine the traffic mobility and the calculated value of road loss degree through weighted combination to provide a comprehensive evaluation criterion for the optimization algorithm. The design of the comprehensive objective function Z is as follows: ； Among them, is a weight coefficient used to control traffic mobility and the calculated value of road loss in the relative importance of path planning optimization and is adjusted according to traffic conditions; S6. Use a genetic algorithm to search for the optimal solution.

2. The multi-objective path planning optimization method driven by implicit function of a multi-channel network according to claim 1, characterized in that: In S1, the traffic data specifically includes: path type T including categories; environmental conditions C indicating the influence of external factors, including levels; traffic management mode M including types; speed limit V set to to discrete values between; traffic flow density D referring to N traffic flow density sampling points under different road conditions, sampled at equal intervals within each traffic flow cycle, with the unit of vehicle / hour; and the actual value of the corresponding road loss degree L ; The preprocessing of traffic data is specifically as follows: fill in the missing values and process the outliers to ensure the integrity and accuracy of the dataset. The final complete dataset includes no less than one sample.

3. The multi-objective path planning optimization method driven by implicit function of a multi-channel network according to claim 2, characterized in that: In S4, specifically: divide the dataset into a training set and a test set at a ratio of 8:2, and use cross-validation to verify the performance of the three-channel network model. During the training process, the training set is used to update the parameters of the three-channel network model, and the loss function is used to measure the difference between the predicted value and the actual target value of the three-channel network model. The loss function is defined as follows: ； Among them, U is the number of samples. By minimizing this loss function, the three-channel network model can gradually adjust the parameters to reduce the error between the predicted value of the road loss degree and the actual value L of the road loss degree.

4. The multi-objective path planning optimization method driven by implicit function of a multi-channel network according to claim 1, characterized in that: In S6, the specific steps are: S61. Randomly generate a set of path planning solutions, calculate the objective function Z , and evaluate the fitness; S62. Select excellent individuals as parents based on fitness, and through crossover operations, recombine the genetic information of the parents to generate new offspring individuals; S63. Randomly mutate some individuals, replace the old population with the newly generated offspring, and form the next-generation population; S64. When one of the two conditions of reaching the maximum number of iterations and the fitness reaching the preset threshold is met, the algorithm terminates and outputs the optimal solution, that is, the optimal combination of the values of each feature.

Images