An Optimization Method for the Collection and Distribution Route of Unmanned Electric Trucks Driven by Complexity Learning

Abstract

The present invention provides an optimization method for the collection and distribution path of unmanned electric trucks driven by complexity learning, including: comprehensively considering static road attributes, dynamic traffic flow, weather factors, and the interaction characteristics of vehicle flow and road network, constructing a natural driving data set according to n road scenario samples; using the independent weight method MCC to calculate the objective weights of each feature of the road scenario samples, representing the differences in the contributions of different features to complexity; using the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) to quantitatively score the complexity of the road scenario samples; building a neural network model to adaptively learn the internal laws of the road scenario samples and estimate the complexity of dynamic uncertain scenarios; using the natural driving data set to train the neural network model; introducing the Operational Design Domain (ODD) boundary and vehicle endurance constraints, and constructing a complex constraint vehicle path optimization model to achieve path optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned electric truck deployment, and more particularly, to an optimization method for the collection and distribution path of unmanned electric trucks driven by complexity learning. Background Art

[0002] Autonomous driving technology has great potential in improving the safety of logistics transportation, reducing labor costs and improving efficiency. In recent years, governments around the world have successively introduced support policies to encourage the application and testing of autonomous vehicles in controlled environments such as closed parks. Ports have relatively closed and stable road conditions, making them an ideal application scenario for autonomous driving technology. Currently, some large ports at home and abroad have tried to apply Automated Electric Container Trucks (unmanned electric trucks) to the internal logistics operations of ports and cooperate with automated terminal handling equipment to initially achieve unmanned horizontal transportation in ports.

[0003] However, there are still many deficiencies in the existing unmanned electric truck deployment methods. The battery life of unmanned electric trucks is limited and it is difficult to support long-distance transportation operations. The road environment is complex and changeable, and it is difficult for the autonomous driving system to adapt to situations beyond the design operating conditions (ODD) of the autonomous driving system. Therefore, it is necessary to evaluate the complexity of the external roads of the port to judge the applicability of unmanned electric trucks and use a neural network model for path optimization. Summary of the Invention

[0004] In view of the above technical problems, an optimization method for the collection and distribution path of unmanned electric trucks driven by complexity learning is provided. The present invention considers static road attributes, dynamic traffic flow, weather factors, and vehicle-flow-road network interaction characteristics, evaluates the road complexity, and uses a neural network model to construct and optimize a complex constrained vehicle path optimization model to achieve path optimization.

[0005] The technical means adopted by the present invention are as follows:

[0006] An optimization method for the collection and distribution path of unmanned electric trucks driven by complexity learning, comprising:

[0007] Comprehensively considering static road attributes, dynamic traffic flow, weather factors, and vehicle-flow and road network interaction characteristics, and constructing a natural driving data set according to n road scene samples;

[0008] Using the independent weight method MCC to calculate the objective weights of each feature of the road scene samples, representing the differences in the contributions of different features to the complexity;

[0009] Using the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) to quantitatively score the complexity of the road scene samples;

[0010] Build a neural network model to adaptively learn the inherent laws of road scene samples and estimate the complexity of dynamic uncertain scenes;

[0011] Train the neural network model using natural driving datasets;

[0012] Introduce ODD boundaries and vehicle endurance constraints to construct a complex-constrained vehicle path optimization model and achieve path optimization.

[0013] Furthermore, road scene samples of n port collection and distribution roads are extracted from the natural driving dataset, and each sample contains m feature attributes. denotes any road scene sample, is the feature attribute values of this sample, is the value vector of the

[0014] Furthermore, the independence weight method MCC constructs a linear relationship with other features by performing weighted regression on the th feature ; the normalized multiple correlation coefficient is obtained to get the weight of the feature :

[0015]

[0016] The multiple correlation coefficient is used to measure the independence between the th feature and other features.

[0017] Furthermore, in the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS), an ideal most complex scene and an ideal simplest scene are introduced, where , ;

[0018] The weighted Euclidean distances between the road scene sample and the ideal most complex scene and the ideal simplest scene are expressed as:

[0019]

[0020] The weighted Euclidean distances and are transformed into the complexity score of the road scene sample ​ :

[0021]

[0022] Among them, the complexity score takes values in the range of [0, 1].

[0023] Furthermore, the neural network model includes an input layer, a hidden layer, and an output layer;

[0024] The input layer receives the d-dimensional feature vector of the road scene, and the output layer generates the corresponding complexity estimation value ; The forward propagation process of the neural network is expressed as:

[0025] ;

[0026] Among them, represents the Sigmoid activation function, represents the number of neurons in the input layer, represents the number of neurons in the hidden layer, represents the number of neurons in the output layer; represents the connection weight from the input layer to the hidden layer, represents the connection weight from the hidden layer to the output layer, represents the threshold of the hidden layer neurons, represents the threshold of the output layer neurons.

[0027] Furthermore, the particle swarm optimization algorithm PSO is used to optimize the connection weights in the neural network model, and the optimal solution obtained by the particle swarm optimization algorithm is decoded as the initial connection weights of the neural network model, and the error backpropagation algorithm BP and the gradient descent algorithm are used to finely tune the neural network model;

[0028] Define the mean square error loss function:

[0029]

[0030] Among them, represents the complexity score, represents the complexity estimation value, represents the error between the estimation value and the actual value;

[0031] Use the gradient descent method to iteratively update the neuron connection weights and thresholds:

[0032]

[0033] Among them, is the learning step size, is the neuron connection weight or threshold value.

[0034] Furthermore, the training of the neural network model using the natural driving dataset specifically includes:

[0035] Randomly select representative road scene segments from the natural driving dataset, select road scene samples, and extract the feature attributes of the samples; use the independence weight method and the technique for order preference by similarity to ideal solution to generate the complexity labels for each scene, and randomly divide them into a training set and a test set according to a certain proportion; input the training set into the neural network model for model training, and use the particle swarm optimization algorithm to optimize the parameters of the neural network model.

[0036] Furthermore, the complex-constrained vehicle routing optimization model is based on the following assumptions:

[0037] Assume that the port collection and distribution network is represented as a complete graph , where is the set of nodes, including the port node 0 and the set of customer nodes , is the set of directed arcs between nodes;

[0038] Assume that the import and export container demands of customer nodes are known and can be split, that is, each node can be served by multiple driverless electric trucks multiple times;

[0039] Assume that the driverless electric truck fleet consists of pure electric container trucks with the same rated TEU and cruising range, and all support autonomous driving within the specific ODD boundary;

[0040] Assume that random factors are not considered and the energy consumption per unit distance is constant;

[0041] Assume that the container sizes of each customer node are uniformly standard TEUs, and empty containers can be shared between nodes;

[0042] Assume that customer demands are known and fixed during the planning period, and dynamic changes in demands are not considered.

[0043] Furthermore, introducing the ODD boundary and vehicle cruising range constraints to construct the complex-constrained vehicle routing optimization model specifically includes:

[0044] Consider the multiple pick-up and delivery characteristics of container transportation and the empty container sharing mode; model the complex-constrained vehicle routing optimization model as a mixed integer programming model, expressed as:

[0045]

[0046] ,

[0047] ,

[0048] ,

[0049] ,

[0050] ,

[0051] ,

[0052] ,

[0053] ,

[0054] ,

[0055] ,

[0056] ,

[0057] in, represents the set of demand nodes, represents a node set, where , 0 means port, Indicates the collection of unmanned electric cards. Represents the set of unmanned electric card travel times; Representation Node and The road scene vector of the transport route between them, Representation Node and The distance between Representation Node The export box demand, Representation Node The demand for import containers Indicates the carrying capacity of unmanned electric trucks. Indicates the range of the unmanned electric card. represents the ODD boundary complexity of the unmanned electric card, is a 0-1 variable. No. 1 Unmanned Electric Card The trip was made by Driving If yes, it is 1, otherwise it is 0; express The quantity of export containers loaded during the th trip of the No. driverless electric truck from to is represented by The quantity of import containers loaded during the th trip of the No. driverless electric truck from to

[0058] Compared with the prior art, the present invention has the following advantages:

[0059] The method for optimizing the collection and distribution path of driverless electric trucks driven by complexity learning provided by the present invention comprehensively considers static road attributes, dynamic traffic flow, weather factors, and the interaction characteristics of vehicle flow and road network, and provides a systematic framework for comprehensively evaluating the road network complexity in this scenario. According to n road scenario samples, a natural driving data set is constructed; the objective weights of each feature of the road scenario samples are calculated by the independent weight method MCC to characterize the differences in the contributions of different features to complexity; the TOPSIS method is used to quantitatively score the complexity of the road scenario samples; a non-linear mapping relationship between generalized scenario features and complexity is established by using a machine learning model to realize the adaptive evaluation of dynamic and non-linear road scenarios. A neural network model is built to adaptively learn the internal laws of road scenario samples and estimate the complexity of dynamic uncertain scenarios; the neural network model is trained by using the natural driving data set; the ODD boundary and vehicle endurance constraints are introduced to construct a complex constraint vehicle path optimization model to realize path optimization; considering the multiple pick-up and delivery characteristics of container transportation and the empty container sharing mode, it can effectively optimize the energy consumption cost of driverless electric trucks at different levels of autonomous driving.

[0060] For the above reasons, the present invention can be widely promoted in the field of driverless electric truck deployment. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0062] Figure 1 is a flow chart of the method for optimizing the collection and distribution path of driverless electric trucks driven by complexity learning in the present invention.

[0063] Figure 2 is a technical roadmap in the embodiments of the present invention.

[0064] Figure 3 Schematic diagram of the unmanned electric truck collection and distribution path planning solution in the embodiment of the present invention

[0065] Figure 4 Neural network model structure for evaluating the complexity of road scenarios in the embodiment of the present invention Detailed implementation manners

[0066] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0067] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data used in appropriate cases can be interchanged so that the embodiments of the present invention described herein can be implemented in an order different from those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0068] As shown in Figure 1 , the present invention provides a method for optimizing the collection and distribution path of unmanned electric trucks driven by complexity learning, including:

[0069] Comprehensively considering static road attributes, dynamic traffic flow, weather factors, and the interaction characteristics of vehicle flow and road network, a natural driving dataset is constructed according to n road scenario samples;

[0070] Specifically, as a preferred implementation manner of the present invention, the natural driving dataset extracts n road scenario samples of port collection and distribution roads, and each sample contains m feature attributes, represents any road scenario sample, is the feature attribute values of this sample, is the value vector of the th feature in all samples.

[0071] The objective weights of each feature of the road scene samples are calculated using the independent weight method MCC, which characterizes the differences in the contributions of different features to the complexity;

[0072] Specifically, as a preferred embodiment of the present invention, the independent weight method MCC constructs a linear relationship with other features by performing weighted regression on the th feature ; The normalized multiple correlation coefficient is used to obtain the weight of the feature :

[0073]

[0074] The multiple correlation coefficient is used to measure the independence between the th feature and other features. In implementation, the smaller it is, the weaker the correlation between and other features, and the more independent information it carries.

[0075] The technique for order preference by similarity to an ideal solution (TOPSIS) is used to quantitatively score the complexity of road scene samples;

[0076] Specifically, as a preferred embodiment of the present invention, an ideal most complex scene and an ideal simplest scene are introduced in the technique for order preference by similarity to an ideal solution (TOPSIS), where , ;

[0077] The weighted Euclidean distances between the road scene sample and the ideal most complex scene and the ideal simplest scene and are expressed as:

[0078]

[0079] The weighted Euclidean distances and are converted into the complexity score of the road scene sample :

[0080]

[0081] where the value range of the complexity score is [0, 1].

[0082] Build a neural network model to adaptively learn the internal laws of road scene samples and estimate the complexity of dynamic uncertain scenes;

[0083] In specific implementation, as a preferred implementation manner of the present invention, the neural network model includes an input layer, a hidden layer, and an output layer, as Figure 4 shown;

[0084] The input layer receives the dimensional feature vector , and the output layer generates a corresponding complexity estimation value ; The forward propagation process of the neural network is expressed as:

[0085] ;

[0086] Wherein, represents the Sigmoid activation function, represents the number of neurons in the input layer, represents the number of neurons in the hidden layer, represents the number of neurons in the output layer; represents the connection weight from the input layer to the hidden layer, represents the connection weight from the hidden layer to the output layer, represents the threshold of the hidden layer neurons, represents the threshold of the output layer neurons.

[0087] In specific implementation, as a preferred implementation manner of the present invention, the particle swarm optimization algorithm PSO is used to optimize the connection weights in the neural network model, the optimal solution obtained by the particle swarm optimization algorithm is decoded as the initial connection weights of the neural network model, and the error backpropagation algorithm BP and the gradient descent algorithm are used to finely tune the neural network model;

[0088] Define the mean square error loss function:

[0089]

[0090] Wherein, represents the complexity score, represents the complexity estimation value, represents the error between the estimation value and the actual value;

[0091] Use the gradient descent method to iteratively update the neuron connection weights and thresholds:

[0092]

[0093] Wherein, is the learning step size, is the neuron connection weight or threshold.

[0094] Train a neural network model using a natural driving dataset;

[0095] In specific implementation, as a preferred implementation manner of the present invention, the training of the neural network model using the natural driving dataset specifically includes:

[0096] Randomly select representative road scene segments from the natural driving dataset, select road scene samples, and extract the feature attributes of the samples; generate complexity labels for each scene using the independent weight method and the technique for order preference by similarity to ideal solution, and randomly divide them into a training set and a test set according to a ratio; input the training set into the neural network model for model training, and use the particle swarm optimization algorithm to optimize the parameters of the neural network model.

[0097] Introduce ODD boundaries and vehicle endurance constraints, construct a complex-constrained vehicle routing optimization model, and achieve route optimization.

[0098] In specific implementation, as a preferred implementation manner of the present invention, the complex-constrained vehicle routing optimization model is based on the following assumptions:

[0099] Assume that the port collection and distribution network is represented as a complete graph , where is the set of nodes, including the port node 0 and the set of customer nodes , is the set of directed arcs between nodes;

[0100] Assume that the import and export container demands of customer nodes are known and can be split, that is, each node can be served by multiple driverless electric trucks multiple times;

[0101] Assume that the driverless electric truck fleet consists of pure electric container trucks with the same rated TEU and endurance mileage, and all support autonomous driving within a specific ODD boundary;

[0102] Assume that random factors are not considered and the energy consumption per unit distance is constant;

[0103] Assume that the container sizes of each customer node are uniformly standard TEU, and empty containers can be shared between nodes;

[0104] Assume that customer demands are known and fixed during the planning period, and dynamic changes in demands are not considered.

[0105] In specific implementation, as a preferred implementation manner of the present invention, the introduction of ODD boundaries and vehicle endurance constraints, and the construction of a complex-constrained vehicle routing optimization model specifically include:

[0106] Consider the multiple pick-up and delivery characteristics of container transportation and the empty container sharing mode; model the complex-constrained vehicle routing optimization model as a mixed integer programming model, expressed as:

[0107] Minimize the total driving distance of the driverless electric truck fleet:

[0108]

[0109] Restrict the driverless electric truck to execute autonomous driving only on paths with a complexity not exceeding the ODD boundary value :

[0110] ,

[0111] Denote the function that maps the road attribute vector to a complexity score; this function can be implemented by a neural network model obtained through training;

[0112] Ensure that each customer node is visited at least once and there are no sub-circuits in a single trip:

[0113] ,

[0114] ,

[0115] Vehicle driving range limit:

[0116] ,

[0117] Traffic flow balance equation to ensure that the number of driverless electric trucks entering and leaving each node is equal:

[0118] ,

[0119] Cargo flow balance equation to ensure that the loading and unloading volume matches the node demand:

[0120] ,

[0121] ,

[0122] ,

[0123] ,

[0124] Restrict the cargo volume on the arc not to exceed the vehicle capacity:

[0125] ,

[0126] Define the value range of decision variables:

[0127] ,

[0128] wherein, represents the set of demand nodes, represents the set of nodes, where , 0 represents the port, represents the set of driverless electric trucks, represents the set of travel times of driverless electric trucks; represents the road scenario vector of the transportation route between node and , represents the distance (km) between node and , represents the export container demand (TEU) of node , represents the import container demand (TEU) of node , represents the carrying capacity (ft) of the driverless electric truck, represents the cruising range (km) of the driverless electric truck, represents the ODD boundary complexity of the driverless electric truck, is a 0-1 variable. If the th trip of the th driverless electric truck travels from to then it is 1, otherwise it is 0; represents the th trip of the th driverless electric truck travels from to and represents the amount of export containers (TEU) loaded on the way, represents the th trip of the th driverless electric truck travels from to

[0129] Embodiment

[0130] As Figure 1 shown, the present invention provides an optimization method for the collection and distribution path of driverless electric trucks driven by complexity learning, Figure 3A simplified schematic diagram of port collection and distribution operations is given. There are multiple container shippers distributed within the hinterland of a certain port. The driverless electric truck needs to perform container transportation tasks between the port and each shipper. The driverless electric truck departs from the port, transports 2 imported heavy containers to customer A via route r1, and after loading 1 exported heavy container and 1 empty container, it was originally planned to return directly to the port via r2. However, after optimization, it can go to customer B via r3 to load the second exported heavy container and then return via r4.

[0131] The transportation process includes two stages: transporting imported heavy containers from the port to each shipper, and transporting exported heavy containers from each shipper back to the port. In order to improve the comprehensive transportation efficiency and make full use of the return load capacity, the driverless electric truck can, after visiting the imported shipper, load and transport the generated exported heavy containers or idle empty containers to other shippers in need, and then pull the exported heavy containers back to the terminal for loading onto the ship to complete a business cycle.

[0132]

[0133] A quantitative method integrating multi-attribute decision-making and machine learning is proposed. This method annotates the complexity of representative scenarios through multi-attribute decision-making and uses a machine learning model to establish a non-linear mapping relationship between the generalized scenario features and the complexity, so as to realize the adaptive evaluation of dynamic and non-linear road scenarios.

[0134] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0135] In the above embodiments of the present invention, the descriptions of each embodiment have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0136] In several embodiments provided in the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only illustrative. For example, the division of the units can be a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings or direct couplings or communication connections shown or discussed with each other can be through some interfaces. The indirect couplings or communication connections of the units or modules can be in electrical or other forms.

[0137] The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0138] In addition, in each embodiment of the present invention, each functional unit can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0139] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The foregoing storage medium includes: various media such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs that can store program codes.

[0140] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of each embodiment of the present invention.

Claims

1. An optimization method for the collection and distribution path of unmanned electric trucks driven by complexity learning, characterized in that, Including: Considering static road attributes, dynamic traffic flow, weather factors, and the interaction characteristics between traffic flow and road network, according to n road scene samples to construct a natural driving dataset; Using the independent weight method MCC to calculate the objective weights of each feature of road scene samples, characterizing the differences in the contributions of different features to complexity; Using the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) to quantitatively score the complexity of road scene samples; Building a neural network model to adaptively learn the internal laws of road scene samples and estimate the complexity of dynamic uncertain scenarios; Using a natural driving dataset to train the neural network model; Introducing the Operational Design Domain (ODD) boundary and vehicle endurance constraints, constructing a vehicle routing optimization model with complex constraints to achieve route optimization; The vehicle routing optimization model with complex constraints is based on the following assumptions: Suppose the port collection and distribution network is represented as a complete graph , where is the set of nodes, including the port node 0 and the set of customer nodes , is the set of directed arcs between nodes; Assume that the import and export container demands of customer nodes are known and can be split, that is, each node can be served by multiple unmanned electric trucks multiple times; Assume that the unmanned electric truck fleet consists of pure electric container trucks with the same rated TEU and endurance mileage, and all support autonomous driving within a specific ODD boundary; Assume that random factors are not considered and the energy consumption per unit distance is constant; Assume that the container sizes of each customer node are uniformly standard TEU, and empty containers can be shared among nodes; Assume that customer demands are known and fixed during the planning period, without considering the dynamic changes in demands; The introduction of the ODD boundary and vehicle endurance constraints to construct a vehicle routing optimization model with complex constraints specifically includes: Considering the multiple pick-up and delivery characteristics of container transportation and the empty container sharing mode; modeling the vehicle routing optimization model with complex constraints as a mixed-integer programming model, expressed as: ， ， ， ， ， ， ， ， ， ， ， Among them, represents the set of demand nodes, represents the set of nodes, where 0 represents the port, represents the set of driverless electric cards, represents the set of travel times of driverless electric cards; represents node and the road scene vector of the transportation route between them, represents node and the distance between them, represents the demand for export containers at node , represents the demand for import containers at node , represents the carrying capacity of the driverless electric card, represents the cruising range of the driverless electric card, represents the ODD boundary complexity of the driverless electric card, is a 0-1 variable. If the th trip of the th driverless electric card goes from to then it is 1, otherwise it is 0; represents the th trip of the th driverless electric card goes from to and the quantity of export containers loaded on the way, represents the th trip of the th driverless electric card goes from to and the quantity of import containers loaded on the way.

2. The complexity learning-driven unmanned electric truck collection and distribution route optimization method according to claim 1, wherein, The following road scene samples of the port collection and distribution roads are extracted from the natural driving dataset, and each sample contains n feature attributes. m Let represent any road scene sample, be the feature attribute values of this sample, and be the value vector of the i-th feature in all samples.

3. The complexity learning-driven optimization method for the collection and distribution path of unmanned electric trucks according to claim 2, characterized in that The independence weight method MCC constructs a linear relationship with other features by performing weighted regression on the th feature ; Normalized complex correlation coefficient Obtain features Weights of : Multiple correlation coefficient Used to measure the independence between the th feature and other features.

4. The complexity learning-driven optimization method for the collection and distribution path of unmanned electric trucks according to claim 3, characterized in that Introduce the ideal most complex scenario into the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) and the ideal simplest scenario , where , ; The road scene sample and the ideal most complex scene and the ideal simplest scene weighted Euclidean distance and are expressed as: Convert the weighted Euclidean distance and into the complexity score of road scene samples : Among them, the complexity score ranges from [0, 1].

5. The complexity learning-driven optimization method for the collection and distribution path of unmanned electric trucks according to claim 1, wherein The neural network model includes an input layer, a hidden layer, and an output layer; The input layer receives the dimensional feature vector , and the output layer generates a corresponding complexity estimation value ; The forward propagation process of the neural network is expressed as: ； Among them, represents the Sigmoid activation function, represents the number of neurons in the input layer, represents the number of neurons in the hidden layer, represents the number of neurons in the output layer; represents the connection weights from the input layer to the hidden layer, represents the connection weights from the hidden layer to the output layer, represents the threshold of the hidden layer neurons, represents the threshold of the output layer neurons.

6. The complexity learning-driven optimization method for the collection and distribution path of unmanned electric trucks according to claim 5, wherein Using the Particle Swarm Optimization (PSO) algorithm to optimize the connection weights in the neural network model, decoding the optimal solution obtained by the PSO algorithm as the initial connection weights of the neural network model, and using the Back Propagation (BP) algorithm and the gradient descent algorithm to finely tune the neural network model; Defining the mean squared error loss function: wherein, represents the complexity score, represents the complexity estimated value, represents the error between the estimated value and the actual value; Using the gradient descent method to iteratively update the neuron connection weights and thresholds: Among them, is the learning step size, is the neuron connection weight or threshold value.

7. The complexity learning-driven optimization method for the collection and distribution path of unmanned electric trucks according to claim 1, wherein The use of a natural driving dataset to train the neural network model specifically includes: Randomly selecting representative road scene segments from the natural driving dataset, selecting road scene samples, and extracting the feature attributes of the samples; using the independent weight method and the TOPSIS method to generate complexity labels for each scene, and randomly dividing them into a training set and a test set according to a certain proportion; inputting the training set into the neural network model for model training, and using the PSO algorithm to optimize the parameters of the neural network model.

Images