Transportation path optimization method, system, device and storage medium for automobile parts

Abstract

The application discloses a transportation path optimization method, system, device and storage medium for automobile parts, relates to the technical field of path optimization, and solves the technical problem that the prior art often relies on a single or a small number of data sources, vehicle states, order matching requirements, real-time traffic and other multi-source heterogeneous data are isolated from each other, path planning is based on incomplete and lagging information, and the accuracy and efficiency of the transportation path optimization method are low; a dynamic collaborative transportation network is constructed based on multi-source heterogeneous data; a multi-vehicle collaborative transportation path scheme is obtained by performing collaborative path optimization on the dynamic collaborative transportation network, multi-source heterogeneous data is fused into a digital decision environment of multidimensional data, and information islands are completely broken; on the basis of the dynamic collaborative transportation network, collaborative path optimization is performed, scattered and lagging data is converted into a collaborative decision whole, path planning is based on a complete and real-time information panorama, and the accuracy and efficiency of transportation path optimization are improved.

Description

Technical Field

[0001] This application belongs to the field of route optimization technology, specifically to methods, systems, equipment, and storage media for optimizing the transportation routes of automotive parts. Background Technology

[0002] Optimizing the transportation routes for automotive parts refers to dynamically planning the lowest-cost, most efficient, and least carbon-emission logistics solutions in complex scenarios involving widely distributed suppliers, diverse parts, and extremely time-sensitive demands, using big data and intelligent algorithms while meeting multiple constraints such as time windows, weight limits, and road conditions. Its core significance lies in the fact that it is not only a key tool for cost reduction and efficiency improvement in the automotive manufacturing industry, directly reshaping profit margins by reducing unnecessary mileage and inventory backlog; it is also the cornerstone for ensuring the stable operation of just-in-time production models and enhancing supply chain resilience, effectively avoiding the risk of production line downtime due to parts shortages; and as an important link in green manufacturing, it significantly reduces the carbon footprint of the transportation sector through precise scheduling, driving the automotive industry towards a deep transformation towards digitalization, intelligence, and sustainability.

[0003] Existing technologies often rely on single or a few data sources. Multi-source heterogeneous data such as vehicle status, order fulfillment requirements, and real-time traffic are fragmented and cannot be used uniformly for decision-making. This results in route planning being based on incomplete and lagging information, leading to low accuracy and efficiency of transportation route optimization methods. Therefore, further improvements are still needed for transportation route optimization methods for automotive parts. Summary of the Invention

[0004] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes a method, system, device, and storage medium for optimizing the transportation routes of automotive parts. This addresses the technical problem that the prior art often relies on a single or a few data sources, and that multi-source heterogeneous data such as vehicle status, order fulfillment requirements, and real-time traffic are fragmented and cannot be uniformly used for decision-making. This results in route planning being based on incomplete and lagging information, leading to low accuracy and efficiency of transportation route optimization methods.

[0005] To achieve the above objectives, the first aspect of this application provides a method for optimizing the transportation routes of automotive parts, comprising: Acquire multi-source heterogeneous data; the multi-source heterogeneous data includes static data, dynamic data, task data, and environmental data; A dynamic collaborative transportation network is constructed based on multi-source heterogeneous data; By optimizing the collaborative paths of the dynamic collaborative transportation network, a multi-vehicle collaborative transportation path scheme is obtained.

[0006] Through the steps outlined above, this application first unifies and aggregates four types of heterogeneous data—static foundations, dynamic road conditions, task instructions, and environmental perception—achieving deep integration of all information from the source. Then, it constructs a dynamic collaborative transportation network, merging multi-source data into a digital decision-making environment encompassing real-time vehicle status, multiple task constraints, and time-varying road network characteristics, completely breaking down information silos. Finally, based on a unified, real-time, and comprehensive panoramic network, collaborative path optimization is implemented, enabling multi-vehicle planning to synchronously respond to changes in traffic flow, vehicle operating conditions, and parts availability requirements. This successfully transforms scattered and lagging data into holistic wisdom for collaborative decision-making, significantly improving the accuracy, dynamic adaptability, and overall scheduling efficiency of path planning.

[0007] Furthermore, the construction of a dynamic cooperative transportation network based on multi-source heterogeneous data includes: Extract multi-source heterogeneous data; the multi-source heterogeneous data includes static data, dynamic data, task data, and environmental data; the static data includes component supplier data and final assembly plant data; the component supplier data includes supplier ID, supplier geographical location, and warehouse operation time window; the final assembly plant data includes final assembly plant geographical location and receiving time window; the dynamic data includes vehicle ID and its corresponding vehicle information; the task data includes part ID, required quantity, and assembly relationship between parts; the environmental data includes road ID, road intersection location, road traffic speed, and short-term road condition prediction data; A dynamic collaborative transportation network is constructed using supplier geographic location, final assembly plant geographic location, and road intersection location as nodes, and road ID travel time as edge weight; the travel time is obtained through a road condition prediction model. Map the vehicle IDs and their corresponding vehicle information in the dynamic data to intelligent agents in the dynamic cooperative transportation network; A complete set transportation unit is constructed based on the assembly relationship between parts; the complete set transportation unit consists of several part IDs and a fixed transportation time range, and the several part IDs in the complete set transportation unit need to be delivered to the final assembly plant together within the fixed transportation time range. The final dynamic collaborative transportation network is obtained by injecting complete sets of transportation units into the dynamic collaborative transportation network.

[0008] Furthermore, the road condition prediction model consists of a graph structure construction module, a spatiotemporal feature extraction module, and a prediction output module; The graph structure construction module is used to construct the graph data object required by the spatiotemporal feature extraction module. Input data includes a road map and its corresponding road data; output data includes a road node feature matrix and an adjacency matrix. The road map uses road intersections as nodes and road segments corresponding to adjacent road intersections as edges. The road data refers to both static and dynamic data within the road. Static and dynamic features are extracted from the road data as feature vectors for the nodes, forming the road node feature matrix. Static features include road segment level, number of lanes, and speed limit; dynamic features include average vehicle speed and traffic flow. The adjacency matrix refers to the connectivity between several pairs of nodes. The spatiotemporal feature extraction module is composed of several stacked spatiotemporal blocks, and the spatiotemporal blocks include a temporal attention layer and a graph convolutional layer. The time attention layer is used to capture the dependency relationship between the states of the same node at different historical time points. The input data is the feature vector of the node at several historical time points; the output data is the node time feature vector after attention weighting. The graph convolutional layer is used to aggregate the spatial information of neighboring nodes. The input data is the node temporal feature vector and adjacency matrix of several nodes; the output data is the node spatiotemporal feature vector of several nodes after spatial information aggregation. The prediction output module is used to predict the travel time corresponding to several road segments. The input data is the node spatiotemporal feature vector output by the last spatiotemporal block in the spatiotemporal feature extraction module, and the output data is the travel time corresponding to several edges.

[0009] This application systematically integrates heterogeneous data from multiple sources, including parts suppliers, assembly plants, real-time vehicle status, order fulfillment correlations, and real-time traffic, to construct a dynamic collaborative transportation network. This network uses geographical locations as nodes and travel time, incorporating high-precision short-term road condition predictions, as edge weights. Vehicles and fulfillment transportation units are mapped as agents and constraint tasks within the network, respectively. The road condition prediction employs a model composed of a temporal attention mechanism and a graph convolutional network, which deeply captures the complex spatiotemporal dependencies in the road network and accurately predicts road segment travel times. This unifies previously isolated data sources into a real-time, panoramic digital twin environment, enabling subsequent multi-vehicle collaborative path optimization based on comprehensive, accurate, and forward-looking information. This improves the adaptability of transportation path planning to real-time changes, its ability to guarantee production fulfillment requirements, and overall logistics efficiency.

[0010] Furthermore, the process of optimizing the collaborative path of the dynamic collaborative transportation network to obtain a multi-vehicle collaborative transportation path scheme includes: Step 1: Extract several agents and their corresponding observation spaces from the dynamic cooperative transportation network; the observation space includes the agent's own state, task-related state, and environmental state; the agent's own state includes the current location node, load list, remaining capacity, and list of served tasks; the task-related state includes the progress of the currently processed complete set of transportation units and the fixed transportation time range, as well as the geographical location of the suppliers corresponding to the remaining parts; the environmental state includes the predicted travel time of neighboring road IDs within K hops around the current agent's road ID; where K is an integer, K>0; Step 2: Perform constraint planning and action space pruning operations on several agents and their corresponding observation spaces to obtain the action space corresponding to the agents. Step 3: Based on the observation space and action space corresponding to the agent, perform policy reasoning and action selection to obtain the final action corresponding to the agent; Step 4: Obtain the new observation space corresponding to the agent by executing the final action and performing a state update operation; Step 5: Perform steps 1 to 4 on all agents until the preset stopping condition is reached to obtain a multi-vehicle cooperative transportation route scheme; the preset stopping condition refers to the condition for terminating the cooperative route optimization process.

[0011] Furthermore, the constraint planning and action space pruning operations include: Extract the current location node corresponding to the agent, and obtain several directly reachable neighboring nodes corresponding to the agent as the original candidate action set according to the dynamic cooperative transportation network; the original candidate action set consists of several candidate actions; The feasible action space is obtained by performing feasibility verification on each candidate action in the original candidate action set through a constraint planner; the feasible action space refers to a number of candidate actions that meet the hard constraints corresponding to the current agent, and the feasible action space is a subset of the original candidate action set. The constraint planner includes kitting time window verification and capacity constraint verification; the kitting time window verification refers to verifying the fixed transportation time range corresponding to the kitting transportation unit. The capacity constraint verification refers to verifying the remaining capacity corresponding to the agent.

[0012] Furthermore, the strategy reasoning and action selection include: Extract the observation space corresponding to the current agent and input the observation space into the policy network corresponding to the current agent to obtain the original probability distribution of the agent in the original candidate action set; the policy network is constructed using a multilayer perceptron, whose input data is the observation space corresponding to the agent and whose output data is the original probability distribution of each candidate action in the original candidate action set; Set the probability values ​​of several candidate actions outside the action space in the original candidate action set to zero, sum the probability values ​​of several candidate actions in the action space, and normalize the probability values ​​to obtain the executable probability distribution corresponding to the action space. Select the candidate action corresponding to the maximum probability value from the executable probability distribution as the final action.

[0013] Furthermore, the step of obtaining the new observation space corresponding to the agent by executing the final action and performing a state update operation includes: Extract the agent's final action and submit the final action to the dynamic cooperative transport network for state transition operation to obtain the new observation space corresponding to the agent; The state transition operation refers to the agent advancing the time according to the passage time of the road ID in the dynamic cooperative transportation network until the agent completes the final action, and updating the agent's own state, task-related state and environmental state, thereby obtaining a new observation space for the agent.

[0014] A second aspect of this application provides a transportation route optimization system for automotive parts, comprising: a data acquisition module and a data analysis module; the data acquisition module and the data analysis module are connected together. The data acquisition module acquires multi-source heterogeneous data through data acquisition devices; the multi-source heterogeneous data includes static data, dynamic data, task data, and environmental data. The data analysis module includes a network construction unit and a scheme generation unit; The network construction unit: constructs a dynamic collaborative transportation network based on multi-source heterogeneous data; The scheme generation unit optimizes the collaborative path of the dynamic collaborative transportation network to obtain a multi-vehicle collaborative transportation path scheme.

[0015] A third aspect of this application provides a transportation route optimization device for automotive parts, the transportation route optimization device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the transportation route optimization method for automotive parts as described in the first aspect and any possible implementation thereof.

[0016] Another aspect of this application provides a storage medium, which is a computer-readable storage medium, on which a computer program is stored, which, when executed by a processor, implements the steps of the method for optimizing the transportation path of automotive parts as described in the first aspect and any possible implementation thereof.

[0017] Compared with the prior art, the beneficial effects of this application are: 1. This application constructs a dynamic collaborative transportation network based on multi-source heterogeneous data; it optimizes the collaborative routes of the dynamic collaborative transportation network to obtain multi-vehicle collaborative transportation route schemes, and uniformly acquires four types of heterogeneous data: static, dynamic, task, and environment, integrating all information from the source; it constructs a dynamic collaborative transportation network to integrate the above multi-source data into a digital decision-making environment with vehicle status, task constraints, and time-varying road conditions, completely breaking down information silos; finally, it optimizes collaborative routes on the basis of a unified, real-time, and comprehensive network, enabling multi-vehicle route planning to synchronously respond to real-time traffic changes, vehicle status, and matching constraints, transforming the originally scattered and lagging data into a whole of collaborative decision-making, making route planning based on a complete and real-time information panorama, thereby significantly improving the accuracy, dynamic adaptability, and overall scheduling efficiency of transportation route optimization.

[0018] 2. This application systematically integrates heterogeneous data from multiple sources, including parts suppliers, assembly plants, real-time vehicle status, order completeness correlation, and real-time traffic, to construct a dynamic collaborative transportation network with geographical locations as nodes and travel time incorporating high-precision short-term road condition predictions as edge weights. Vehicles and completeness transportation units are mapped as agents and constraint tasks in the network, respectively. The road condition prediction adopts a model composed of a temporal attention mechanism and a graph convolutional network, which can deeply capture the complex spatiotemporal dependencies in the road network and accurately predict the travel time of road segments. This unifies the originally isolated data sources into a real-time, panoramic digital twin environment, enabling subsequent multi-vehicle collaborative path optimization to be based on comprehensive, accurate, and forward-looking information. This improves the adaptability of transportation path planning to real-time changes, the ability to guarantee production completeness requirements, and overall logistics efficiency.

[0019] 3. This application uses each transport vehicle as a decision-making agent, forming a comprehensive perception based on its real-time location, load, task progress, and surrounding road condition predictions in a dynamic network. A constraint planner is then used to rigorously verify the next possible node, automatically eliminating any invalid choices that may violate time windows or capacity limits, thus ensuring the feasibility of the solution before decision-making. Subsequently, a policy network is used to evaluate and select the optimal action from the remaining feasible options, driving the vehicle to perform movement operations in a simulated environment and updating the global state in real time. This achieves online collaborative decision-making in a multi-constraint dynamic environment, automatically generating transport route solutions that strictly adhere to business rules such as complete delivery and vehicle capacity, while dynamically adapting to traffic changes and achieving better overall efficiency. Attached Figure Description

[0020] 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.

[0021] Figure 1 This is a flowchart of the transportation route optimization method for automotive parts according to this application; Figure 2 This is a schematic diagram illustrating the principle of the transportation route optimization system for automotive parts according to this application. Detailed Implementation

[0022] The technical solutions of this application will be clearly and completely described below with reference to the embodiments. 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.

[0023] Please see Figure 1 The first aspect of this application provides a method for optimizing the transportation route of automotive parts, including: Acquire multi-source heterogeneous data; multi-source heterogeneous data includes static data, dynamic data, task data, and environmental data; A dynamic collaborative transportation network is constructed based on multi-source heterogeneous data; By optimizing the collaborative paths of the dynamic collaborative transportation network, a multi-vehicle collaborative transportation path scheme is obtained.

[0024] The construction of a dynamic collaborative transportation network based on multi-source heterogeneous data in this embodiment includes: Extracting multi-source heterogeneous data; multi-source heterogeneous data includes static data, dynamic data, task data, and environmental data; static data includes component supplier data and final assembly plant data; component supplier data includes supplier ID, supplier geographical location, and warehouse operation time window; final assembly plant data includes final assembly plant geographical location and receiving time window; dynamic data includes vehicle ID and its corresponding vehicle information; vehicle information includes GPS location, load status, remaining capacity, and vehicle status; in this embodiment, vehicle status includes driving, waiting, and loading / unloading; task data in this embodiment includes part ID, required quantity, required deadline, and assembly relationships between parts; environmental data includes road ID, road intersection location, road speed, congestion events, traffic control, and short-term road condition prediction data, etc. A dynamic collaborative transportation network is constructed using supplier geographic location, final assembly plant geographic location, and road intersection location as nodes, and road ID travel time as edge weights; travel time is obtained through a road condition prediction model. Map vehicle IDs and their corresponding vehicle information from dynamic data to intelligent agents in a dynamic collaborative transportation network. A complete set transportation unit is constructed based on the assembly relationship between parts. The complete set transportation unit consists of several part IDs and a fixed transportation time range. Several part IDs in the complete set transportation unit need to be delivered to the final assembly plant together within the fixed transportation time range. The final dynamic collaborative transportation network is obtained by injecting complete sets of transportation units into the dynamic collaborative transportation network.

[0025] The road condition prediction model in this embodiment consists of a graph structure construction module, a spatiotemporal feature extraction module, and a prediction output module; The graph structure construction module is used to construct the graph data objects required by the spatiotemporal feature extraction module. Input data includes a road map and its corresponding road data; output data consists of a road node feature matrix and an adjacency matrix. The road map uses road intersections as nodes and road segments corresponding to adjacent intersections as edges. Road data refers to both static and dynamic data within the road. Static and dynamic features are extracted from the road data as feature vectors for the nodes, forming the road node feature matrix. Static features include road segment level, number of lanes, and speed limit; dynamic features include average speed and traffic flow. The adjacency matrix refers to the connectivity between several pairs of nodes. The spatiotemporal feature extraction module consists of several stacked spatiotemporal blocks, which include a temporal attention layer and a graph convolutional layer. The temporal attention layer is used to capture the dependencies between the states of the same node at different historical time points. The input data is the feature vectors of the node at several historical time points; the output data is the node temporal feature vector after attention weighting; in this embodiment, the temporal attention layer adopts a self-attention mechanism. The graph convolutional layer is used to aggregate the spatial information of neighboring nodes. The input data is the node temporal feature vector and adjacency matrix of several nodes; the output data is the node spatiotemporal feature vector of several nodes after spatial information aggregation; in this embodiment, the graph convolutional layer adopts a graph convolutional network. The prediction output module is used to predict the travel time corresponding to several road segments. The input data is the node spatiotemporal feature vector output by the last spatiotemporal block in the spatiotemporal feature extraction module, and the output data is the travel time corresponding to several edges. In this embodiment, the prediction output module is composed of several fully connected layers.

[0026] In this embodiment, the multi-vehicle cooperative transportation route scheme is obtained by optimizing the cooperative route of the dynamic cooperative transportation network, including: Step 1: Extract several agents and their corresponding observation spaces from the dynamic cooperative transportation network; the observation space includes the agent's own state, task-related state, and environmental state; the agent's own state includes the current location node, load list, remaining capacity, and list of served tasks; the task-related state includes the progress of the currently processed complete set of transportation units and the fixed transportation time range, as well as the geographical location of the suppliers corresponding to the remaining parts; the environmental state includes the predicted travel time of neighboring road IDs within K hops around the current agent's road ID; where K is an integer, K>0, and the specific value is set according to experience, and in this embodiment it is set to 2; Step 2: Perform constraint planning and action space pruning operations on several agents and their corresponding observation spaces to obtain the action space corresponding to the agents. Step 3: Based on the observation space and action space corresponding to the agent, perform policy reasoning and action selection to obtain the final action corresponding to the agent; Step 4: Obtain the new observation space corresponding to the agent by executing the final action and performing a state update operation; Step 5: Perform steps 1 to 4 on all agents until the preset stopping condition is reached to obtain a multi-vehicle cooperative transportation route scheme; the preset stopping condition refers to the condition for terminating the cooperative route optimization process. In this embodiment, the preset stopping condition is the completion of the agent's transportation task.

[0027] The constraint planning and action space pruning operations in this embodiment include: Extract the current location node corresponding to the agent, and obtain several directly reachable neighboring nodes corresponding to the agent as the original candidate action set according to the dynamic cooperative transportation network; the original candidate action set consists of several candidate actions; in this embodiment, the candidate action refers to the next target node of the agent; The feasible action space is obtained by performing feasibility checks on each candidate action in the original candidate action set through the constraint planner. The feasible action space refers to a number of candidate actions that meet the hard constraints corresponding to the current agent. The feasible action space is a subset of the original candidate action set. The constraint planner includes kitting time window verification and capacity constraint verification. Kitting time window verification refers to verifying the fixed transportation time range corresponding to the kitting transportation unit. In this embodiment, after the current agent moves to the next target node and completes the target node operation, it determines whether it will inevitably lead to the inability to access and complete the transportation of the remaining parts in the current kitting transportation unit within the fixed transportation time range. If it will inevitably lead to the inability to complete, the current candidate action is eliminated. Capacity constraint verification refers to verifying the remaining capacity corresponding to the agent. In this embodiment, after the agent moves to the next target node and completes the task at that target node, it is determined whether the capacity of the parts loaded at the next target node exceeds the remaining capacity. If it does, the current candidate action is eliminated. In this embodiment, within the constraint planner, if at least one of the kitting time window verification and capacity constraint verification is satisfied, the current candidate action is eliminated.

[0028] The strategy reasoning and action selection in this embodiment include: Extract the observation space corresponding to the current agent and input the observation space into the policy network corresponding to the current agent to obtain the original probability distribution of the agent in the original candidate action set; the policy network is constructed using a multilayer perceptron, whose input data is the observation space corresponding to the agent and whose output data is the original probability distribution of each candidate action in the original candidate action set; Set the probability values ​​of several candidate actions outside the action space in the original candidate action set to zero, sum the probability values ​​of several candidate actions in the action space, and normalize the probability values ​​to obtain the executable probability distribution corresponding to the action space. Select the candidate action corresponding to the maximum probability value from the executable probability distribution as the final action.

[0029] In this embodiment, the new observation space corresponding to the agent is obtained by executing the final action and performing a state update operation, including: Extract the agent's final action and submit it to the dynamic cooperative transport network for state transition operations to obtain the agent's new observation space; State transition operation refers to the process by which an agent advances its time based on the travel duration of road IDs in a dynamic cooperative transportation network until it completes its final action, thereby updating its own state, task-related state, and environmental state to obtain a new observation space for the agent.

[0030] This embodiment innovatively assigns each transport vehicle the identity of an intelligent decision-making agent, enabling it to build a comprehensive perception system based on real-time location, load status, task progress, and road condition prediction. By introducing a constraint planner to pre-verify the next optional nodes, it automatically eliminates any invalid paths that violate time windows or capacity limits, ensuring the feasibility of the solution from the source. Subsequently, a policy network is used to evaluate and select the best action from the remaining feasible solutions, driving the vehicle to move in the simulation environment and updating the global state in real time. This achieves online collaborative decision-making in a multi-constraint dynamic environment, which can not only automatically generate transport solutions that strictly conform to business rules such as complete delivery and load limits, but also flexibly adapt to traffic fluctuations, ultimately achieving the goal of optimal path planning in terms of global efficiency.

[0031] Please see Figure 2 A second aspect of this application provides a transportation route optimization system for automotive parts, comprising: a data acquisition module and a data analysis module; the data acquisition module and the data analysis module are connected. Data acquisition module: Acquires multi-source heterogeneous data through data acquisition devices; multi-source heterogeneous data includes static data, dynamic data, task data, and environmental data; the data acquisition devices in this embodiment include several sensors, etc. The data analysis module includes a network construction unit and a solution generation unit; Network building blocks: Constructing a dynamic collaborative transportation network based on multi-source heterogeneous data; Solution generation unit: Optimizes collaborative paths in a dynamic collaborative transportation network to obtain multi-vehicle collaborative transportation path solutions.

[0032] A third aspect of this application provides a transportation route optimization device for automotive parts, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the transportation route optimization method for automotive parts as described in the first aspect embodiment and any possible implementation thereof.

[0033] Another embodiment of this application provides a storage medium, which is a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the transportation route optimization method for automotive parts as described in the first aspect embodiment and any possible implementation thereof.

[0034] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0035] The working principle of this application is as follows: First, multi-source heterogeneous data is acquired. Then, a dynamic collaborative transportation network is constructed based on this data. Next, collaborative path optimization is performed on the dynamic collaborative transportation network to obtain multi-vehicle collaborative transportation route solutions. This involves uniformly acquiring four types of heterogeneous data: static, dynamic, task, and environmental, thus integrating all information from the source. The construction of the dynamic collaborative transportation network integrates the aforementioned multi-source data into a digital decision-making environment with vehicle status, task constraints, and time-varying road conditions, completely breaking down information silos. Finally, collaborative path optimization is performed on a unified, real-time, and comprehensive network foundation, enabling multi-vehicle route planning to synchronously respond to real-time traffic changes, vehicle status, and kitting constraints. This transforms the originally scattered and lagging data into a cohesive collaborative decision-making whole, making route planning based on a complete and real-time information panorama. This significantly improves the accuracy, dynamic adaptability, and overall scheduling efficiency of transportation route optimization, avoiding the problems of existing technologies that often rely on single or a few data sources. Furthermore, the fragmented nature of multi-source heterogeneous data, such as vehicle status, order kitting requirements, and real-time traffic, prevents unified use for decision-making, resulting in route planning based on incomplete and lagging information, leading to low accuracy and efficiency in transportation route optimization methods.

[0036] The above embodiments are only used to illustrate the technical methods of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of this application without departing from the spirit and scope of the technical methods of this application.

Claims

1. A method for optimizing the transportation routes of automotive parts, characterized in that, include: Acquire multi-source heterogeneous data; the multi-source heterogeneous data includes static data, dynamic data, task data, and environmental data; A dynamic collaborative transportation network is constructed based on multi-source heterogeneous data; Multi-vehicle collaborative transportation route schemes are obtained by optimizing the collaborative routes of a dynamic collaborative transportation network.

2. The method for optimizing the transportation route of automotive parts according to claim 1, characterized in that, The construction of a dynamic collaborative transportation network based on multi-source heterogeneous data includes: Extract multi-source heterogeneous data; the multi-source heterogeneous data includes static data, dynamic data, task data, and environmental data; the static data includes component supplier data and final assembly plant data; the component supplier data includes supplier ID, supplier geographical location, and warehouse operation time window; the final assembly plant data includes final assembly plant geographical location and receiving time window; the dynamic data includes vehicle ID and its corresponding vehicle information; the task data includes part ID, required quantity, and assembly relationship between parts; the environmental data includes road ID, road intersection location, road traffic speed, and short-term road condition prediction data; A dynamic collaborative transportation network is constructed using supplier geographic location, final assembly plant geographic location, and road intersection location as nodes, and road ID travel time as edge weight; the travel time is obtained through a road condition prediction model. Map the vehicle IDs and their corresponding vehicle information in the dynamic data to intelligent agents in the dynamic cooperative transportation network; A complete set transportation unit is constructed based on the assembly relationships between parts; the complete set transportation unit consists of several part IDs and a fixed transportation time range. The final dynamic collaborative transportation network is obtained by injecting complete sets of transportation units into the dynamic collaborative transportation network.

3. The method for optimizing the transportation route of automotive parts according to claim 2, characterized in that, The road condition prediction model consists of a graph structure construction module, a spatiotemporal feature extraction module, and a prediction output module. The graph structure construction module is used to construct the graph data object required by the spatiotemporal feature extraction module. Input data includes a road map and its corresponding road data; output data includes a road node feature matrix and an adjacency matrix. The road map uses road intersections as nodes and road segments corresponding to adjacent road intersections as edges. The road data refers to both static and dynamic data within the road. Static and dynamic features are extracted from the road data as feature vectors for the nodes, forming the road node feature matrix. Static features include road segment level, number of lanes, and speed limit; dynamic features include average vehicle speed and traffic flow. The adjacency matrix refers to the connectivity between several pairs of nodes. The spatiotemporal feature extraction module is composed of several stacked spatiotemporal blocks, and the spatiotemporal blocks include a temporal attention layer and a graph convolutional layer. The time attention layer is used to capture the dependency relationship between the states of the same node at different historical time points. The input data is the feature vector of the node at several historical time points; the output data is the node time feature vector after attention weighting. The graph convolutional layer is used to aggregate the spatial information of neighboring nodes. The input data is the node temporal feature vector and adjacency matrix of several nodes; the output data is the node spatiotemporal feature vector of several nodes after spatial information aggregation. The prediction output module is used to predict the travel time corresponding to several road segments. The input data is the node spatiotemporal feature vector output by the last spatiotemporal block in the spatiotemporal feature extraction module, and the output data is the travel time corresponding to several edges.

4. The method for optimizing the transportation route of automotive parts according to claim 1, characterized in that, The process of optimizing the collaborative path of the dynamic collaborative transportation network to obtain a multi-vehicle collaborative transportation path scheme includes: Step 1: Extract several agents and their corresponding observation spaces from the dynamic cooperative transportation network; the observation space includes the agent's own state, task-related state, and environmental state; the agent's own state includes the current location node, load list, remaining capacity, and list of served tasks; the task-related state includes the progress of the currently processed complete set of transportation units and the fixed transportation time range, as well as the geographical location of the suppliers corresponding to the remaining parts; the environmental state includes the predicted travel time of neighboring road IDs within K hops around the current agent's road ID; where K is an integer, K>0; Step 2: Perform constraint planning and action space pruning operations on several agents and their corresponding observation spaces to obtain the action space corresponding to the agents. Step 3: Based on the observation space and action space corresponding to the agent, perform policy reasoning and action selection to obtain the final action corresponding to the agent; Step 4: Obtain the new observation space corresponding to the agent by executing the final action and performing a state update operation; Step 5: Perform steps 1 to 4 on all agents until the preset stopping condition is met to obtain a multi-vehicle cooperative transportation route plan.

5. The method for optimizing the transportation route of automotive parts according to claim 4, characterized in that, The constraint planning and action space pruning operations include: Extract the current location node corresponding to the agent, and obtain several directly reachable neighboring nodes corresponding to the agent as the original candidate action set according to the dynamic cooperative transportation network; the original candidate action set consists of several candidate actions; The feasible action space is obtained by performing feasibility verification on each candidate action in the original candidate action set through a constraint planner; the feasible action space refers to a number of candidate actions that meet the hard constraints corresponding to the current agent, and the feasible action space is a subset of the original candidate action set. The constraint planner includes kitting time window verification and capacity constraint verification; the kitting time window verification refers to verifying the fixed transportation time range corresponding to the kitting transportation unit. The capacity constraint verification refers to verifying the remaining capacity corresponding to the agent.

6. The method for optimizing the transportation route of automotive parts according to claim 4, characterized in that, The strategy reasoning and action selection include: Extract the observation space corresponding to the current agent and input the observation space into the policy network corresponding to the current agent to obtain the original probability distribution of the agent in the original candidate action set; the policy network is constructed using a multilayer perceptron, whose input data is the observation space corresponding to the agent and whose output data is the original probability distribution of each candidate action in the original candidate action set; Set the probability values ​​of several candidate actions outside the action space in the original candidate action set to zero, sum the probability values ​​of several candidate actions in the action space, and normalize the probability values ​​to obtain the executable probability distribution corresponding to the action space. Select the candidate action corresponding to the maximum probability value from the executable probability distribution as the final action.

7. The method for optimizing the transportation route of automotive parts according to claim 4, characterized in that, The process of obtaining the new observation space corresponding to the agent by executing the final action and performing a state update operation includes: Extract the agent's final action and submit the final action to the dynamic cooperative transport network for state transition operation to obtain the new observation space corresponding to the agent; The state transition operation refers to the agent advancing the time according to the passage time of the road ID in the dynamic cooperative transportation network until the agent completes the final action, and updating the agent's own state, task-related state and environmental state, thereby obtaining a new observation space for the agent.

8. A transportation route optimization system for automotive parts, characterized in that, include: A data acquisition module and a data analysis module; the data acquisition module and the data analysis module are connected together. The data acquisition module acquires multi-source heterogeneous data through data acquisition devices; the multi-source heterogeneous data includes static data, dynamic data, task data, and environmental data. The data analysis module includes a network construction unit and a scheme generation unit; The network construction unit: constructs a dynamic collaborative transportation network based on multi-source heterogeneous data; The scheme generation unit optimizes the collaborative path of the dynamic collaborative transportation network to obtain a multi-vehicle collaborative transportation path scheme.

9. A transportation route optimization device for automotive parts, characterized in that, The vehicle component transportation route optimization device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the vehicle component transportation route optimization method as described in any one of claims 1-7.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the transportation route optimization method for automotive parts as described in any one of claims 1-7.

Images