A Transportation Intelligent Assignment and Scheduling Method and System

Abstract

The present application provides a method and system for intelligent assignment and scheduling of transportation, which relates to the field of logistics management. The method includes: obtaining order information and transportation capacity resource information, where the order information includes cargo attributes, the information of self-owned transportation capacity resources includes vehicle cabin parameters, and the information of external transportation capacity resources includes simplified load parameters. Processing the internal image of the vehicle cabin by using the wavelet transform method to determine the three-dimensional parameters of the abnormal area and generate accurate vehicle cabin parameters. Based on the order information and the accurate vehicle cabin parameters, vehicle and route allocation are performed through a multi-objective scheduling algorithm to generate a candidate scheduling plan; then a three-dimensional loading feasibility check is executed by using a multi-grid construction module to predict the deformation of the cargo and the vehicle cabin and mark the ghost volume. Finally, in response to the passing of the check, the optimal scheduling plan is output. The present invention effectively improves the refinement level of logistics transportation management, avoids space conflicts and overloading risks, and improves the utilization efficiency of transportation capacity resources and transportation safety.

Description

Technical Field

[0001] The present disclosure relates to the field of logistics management, and in particular, to a method and system for intelligent assignment and scheduling of transportation. Background Art

[0002] In the field of modern logistics transportation, with the rapid growth of the number of orders and transportation demands, how to efficiently utilize transportation capacity resources to achieve intelligent scheduling and refined management of orders has become an important issue that enterprises urgently need to solve. Currently, the logistics transportation industry is gradually transforming and upgrading from the traditional extensive management mode to the direction of digitalization, intelligence, and refinement. However, in the actual operation process, traditional transportation scheduling methods often still rely mainly on manual experience or simple rules to assign vehicles and plan routes, usually only matching based on relatively simple and single parameters such as the load, volume, or basic route requirements of vehicles, lacking more comprehensive and accurate consideration. However, this rough matching mode has many drawbacks: First, it cannot effectively identify the spatial adaptability between goods and vehicle compartments, easily causing waste or insufficient utilization of vehicle space, directly reducing transportation efficiency; Second, during the transportation process, due to the lack of refined loading planning, especially in the transportation scenarios of special goods such as dangerous goods, cold chain goods, fragile goods, and precision instrument equipment, problems such as space conflicts, goods damage, overloading, or violation of relevant safety and compliance standards are more likely to occur, greatly increasing transportation risks and costs.

[0003] In addition, with the continuous intensification of competition in the supply chain market, customers' requirements for the timeliness, stability, and safety of logistics services are increasing day by day. Simply relying on traditional extensive scheduling methods has been difficult to meet the diversified, personalized, and refined market demands. Especially in complex application scenarios such as transportation route optimization, precise matching of goods and vehicles, and multi-vehicle collaborative operations, the deficiencies of existing technical means are more prominent. Therefore, there is an urgent need for a more efficient, precise, and intelligent scheduling management method to achieve refined control of the entire logistics transportation process, so as to meet the current practical needs of the development of the logistics industry. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present application provides a method and system for intelligent assignment and scheduling of transportation.

[0005] In a first aspect, the present application provides a method for intelligent assignment and scheduling of transportation, including:

[0006] Obtain order information and transportation capacity resource information, where the order information includes: goods attributes; the transportation capacity resource information includes: vehicle compartment parameters of self-owned transportation capacity and simplified load parameters of external transportation capacity; the vehicle compartment parameters include: internal images of the vehicle compartment;

[0007] For the vehicle cabin parameters of the self-owned transportation capacity, the wavelet transform method is used to process the internal image of the vehicle cabin to determine the three-dimensional parameters of the abnormal area; based on the three-dimensional parameters of the abnormal area and the vehicle cabin parameters, accurate vehicle cabin parameters of the self-owned transportation capacity are generated.

[0008] Based on the order information, the accurate vehicle cabin parameters of the self-owned transportation capacity, and the simplified load parameters of the external transportation capacity, the multi-objective scheduling algorithm is used to allocate vehicles and routes for the pre-processed order set to generate a candidate scheduling plan;

[0009] The preset multi-grid construction module is used to perform three-dimensional loading feasibility verification on the candidate scheduling plan to generate a feasibility verification result; in response to the failure of the feasibility verification, it is rolled back to the multi-objective scheduling algorithm for scheduling plan correction;

[0010] In response to all the feasibility verifications passing, the final scheduling plan is output;

[0011] The three-dimensional loading feasibility verification includes: based on the order information and the transportation capacity resource information, determining the predicted deformation for the goods and the vehicle cabin, and in the multi-grid construction module, mapping the predicted deformation into a ghost volume to be marked in the coordinate units corresponding to different grid layers;

[0012] In response to no overlap between the overall size of the goods and the coordinate units of the ghost volume, the fitting degree between the local surface of the goods and the local units of the ghost volume is compared at the grid layer. In response to the existence of fitting conflicts, it is marked that the loading is not feasible and rolled back to the multi-objective scheduling algorithm. In response to the non-existence of fitting conflicts, it is output that the feasibility verification passes.

[0013] In a second aspect, the present application provides a transportation intelligent assignment and scheduling system, including:

[0014] A receiving module for obtaining order information and transportation capacity resource information, where the order information includes: goods attributes; the transportation capacity resource information includes: vehicle cabin parameters of the self-owned transportation capacity, and simplified load parameters of the external transportation capacity; the vehicle cabin parameters include: internal images of the vehicle cabin;

[0015] The receiving module is further configured to, for the vehicle cabin parameters of the self-owned transportation capacity, use the wavelet transform method to process the internal image of the vehicle cabin to determine the three-dimensional parameters of the abnormal area; based on the three-dimensional parameters of the abnormal area and the vehicle cabin parameters, generate accurate vehicle cabin parameters of the self-owned transportation capacity.

[0016] A scheduling module for, based on the order information, the accurate vehicle cabin parameters of the self-owned transportation capacity, and the simplified load parameters of the external transportation capacity, using the multi-objective scheduling algorithm to allocate vehicles and routes for the pre-processed order set to generate a candidate scheduling plan;

[0017] A detection module, configured to use a preset multi - grid construction module to perform three - dimensional loading feasibility verification on the candidate scheduling plan, generate a feasibility verification result; in response to the failure of the feasibility verification, fallback to the multi - objective scheduling algorithm for scheduling plan correction;

[0018] An output module, configured to output a final scheduling plan in response to all passed feasibility verifications;

[0019] The three - dimensional loading feasibility verification includes: based on the order information and the transport capacity resource information, determining the predicted deformation of the goods and the vehicle cabin, and in the multi - grid construction module, mapping the predicted deformation into a ghost volume to be marked in the coordinate units corresponding to different grid layers;

[0020] In response to no overlap between the overall size of the goods and the coordinate units of the ghost volume, comparing the fitting degree between the local surface of the goods and the local units of the ghost volume at the grid layer. In response to the existence of fitting conflicts, marking the loading as infeasible and falling back to the multi - objective scheduling algorithm. In response to the non - existence of fitting conflicts, outputting that the feasibility verification passes.

[0021] Compared with the prior art, the present invention processes the internal image of the vehicle by using the wavelet transform method, effectively identifies and extracts the three - dimensional parameters of the abnormal area in the vehicle cabin, thereby generating accurate vehicle cabin parameters, significantly improving the accuracy of the space matching between the goods and the vehicle, and avoiding the problems of space waste or mismatch caused by fuzzy parameters in the traditional method. Secondly, the present invention uses a multi - objective scheduling algorithm to perform refined vehicle and route allocation on the pre - processed order set, and cooperates with a three - dimensional loading feasibility verification mechanism to effectively ensure the feasibility and compliance of the scheduling plan. Especially when transporting dangerous goods, cold - chain goods and other special goods, by predicting the deformation between the goods and the vehicle cabin and constructing a ghost volume, accurate local surface fitting degree verification is realized in the multi - grid, greatly reducing the risks of space conflict, overloading and goods damage during transportation. Thirdly, the present invention realizes the efficient coordination between its own transport capacity and external transport capacity, can perform optimal matching and scheduling according to the refined parameters of different types of transport capacity, and significantly improves the overall utilization efficiency of the enterprise's transport capacity resources and the safety and stability of transportation. Brief Description of the Drawings

[0022] Figure 1 It is a flowchart of a transportation intelligent assignment scheduling method provided by an embodiment of the present application;

[0023] Figure 2 It is a structure diagram of transport capacity resource information provided by an embodiment of the present application;

[0024] Figure 3Schematic diagram of the impact of abnormal areas on the placement of goods provided by the embodiments of the present application;

[0025] Figure 4 Schematic diagram of a transportation intelligent assignment and scheduling system provided by the embodiments of the present application. Detailed implementation manners

[0026] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with 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.

[0027] It is found that the existing general order scheduling methods generally adopt manual or semi-automated processing methods, mainly making a rough match for goods based on simple parameters of vehicles (such as load capacity, volume, etc.), and usually lacking refined management and unified coordination of its own transportation capacity and external transportation capacity.

[0028] Therefore, considering the impact of its own transportation capacity and external transportation capacity on order scheduling, this embodiment provides a refined transportation intelligent assignment and scheduling method, which can achieve efficient and automated order processing through the following settings:

[0029] Automatic assignment setting: In this embodiment, by setting conditional parameters in multiple dimensions such as transportation routes, provinces, municipalities, vehicle types, order types, and cargo types, the matching degree of transportation tasks is automatically judged according to the contract agreement. When the order task meets the preset conditions, the system can automatically assign the task to the corresponding external carrier, realizing automatic review and efficient and accurate automatic assignment of the order.

[0030] Automatic scheduling setting: This embodiment considers the characteristics and advantages of its own transportation capacity and provides an automatic scheduling scheme for its own vehicles. Specifically, it includes automatic splitting of orders (for example, splitting orders according to cargo categories, quantities, number of boxes or pallets) and merging of small orders. Through parameter configuration such as transportation routes, provinces, municipalities, vehicle types, order types, and cargo types, when the conditions are met, the system automatically assigns tasks and plans routes for its own vehicles and drivers, so as to achieve the purpose of automatic review and efficient automatic scheduling.

[0031] Automatic assignment and scheduling setting: For scenarios that require coordination of external carriers and their vehicles, this embodiment also considers factors such as transportation routes, provinces, municipalities, vehicle types, order types, and cargo types, and sets a more precise automatic assignment and scheduling function. When the transportation task meets specific conditions, the system can not only automatically assign the task to the designated external carrier, but also directly schedule the transportation task to the external vehicles and drivers designated by the carrier on behalf of the carrier, realizing comprehensive control of external transportation capacity and ensuring a high degree of adaptation and efficient utilization of external transportation capacity and tasks.

[0032] See Figure 1As shown in the figure, it is a flowchart of a transportation intelligent assignment and scheduling method provided by an embodiment of the present application. The method includes steps S101 to S104, where:

[0033] S101: Obtain order information and transportation capacity resource information. The order information includes: cargo attributes; the transportation capacity resource information includes: vehicle cabin parameters of self-owned transportation capacity, and simplified load parameters of external transportation capacity; the vehicle cabin parameters include: internal images of the vehicle cabin. For the vehicle cabin parameters of self-owned transportation capacity, use the wavelet transform method to process the internal images of the vehicle cabin to determine the three-dimensional parameters of abnormal areas; based on the three-dimensional parameters of abnormal areas and the vehicle cabin parameters, generate accurate vehicle cabin parameters of self-owned transportation capacity.

[0034] S102: Based on the order information, the accurate vehicle cabin parameters of self-owned transportation capacity, and the simplified load parameters of external transportation capacity, use a multi-objective scheduling algorithm to allocate vehicles and routes for the preprocessed order set to generate a candidate scheduling plan.

[0035] S103: Use a preset multi-grid construction module to perform three-dimensional loading feasibility verification on the candidate scheduling plan to generate a feasibility verification result; in response to the failure of the feasibility verification, roll back to the multi-objective scheduling algorithm to correct the scheduling plan.

[0036] S104: In response to all passing of the feasibility verification, output the final scheduling plan.

[0037] In the large-scale and multi-type transportation scenarios of the present application, by obtaining basic data including cargo attributes and transportation capacity resource information, using a multi-objective scheduling algorithm to allocate vehicles and routes for orders, and then combining the accurate vehicle cabin parameters of self-owned transportation capacity and the simplified load parameters of external transportation capacity for three-dimensional loading feasibility verification in the follow-up, to achieve automated and accurate transportation assignment for various types of goods including dangerous goods or cold chain.

[0038] For each order in the present application, first generate a candidate scheduling plan by comprehensively considering factors such as freight cost, transportation timeliness, vehicle empty running rate, and compliance requirements for special goods, and then check at the fine-grained three-dimensional loading level whether the goods can actually be loaded or docked in the cabin of the specified vehicle. If the verification result is feasible, directly output the scheduling plan; if the verification fails, roll it back to the algorithm side for plan correction, so as to continuously iterate until the optimal assignment that takes into account both efficiency and safety compliance is obtained.

[0039] Regarding the above S101:

[0040] In specific implementation, first obtain order information and transportation capacity resource information. The order information may include, but is not limited to: cargo attributes, place of shipment, place of receipt, and time requirements.

[0041] Among them, the cargo attributes may include descriptions of size, weight, temperature control requirements, or dangerous goods attributes; the time requirements indicate when the cargo needs to depart or be delivered.

[0042] See Figure 2 , Figure 2 FIG. is a structural diagram of transport capacity resource information provided for an embodiment of the present application. Among them, the transport capacity resource information is divided into two categories: self-owned transport capacity and external transport capacity. For example, self-owned transport capacities A, B, C, etc., and external transport capacities a, b, c, etc. The accurate vehicle cabin position parameters of the self-owned transport capacity include three-dimensional space information inside the cabin, sub-cabin structure information, and protection configuration information, which are used for higher-precision loading inspection of the vehicle interior.

[0043] The so-called three-dimensional space information inside the cabin is data such as the length, width, and height inside the vehicle, as well as concave and convex structures and compartment layouts; the sub-cabin structure information can distinguish cold chain areas, ordinary cabin areas, or explosion-proof cabin areas, etc.; the protection configuration information is used to indicate whether the vehicle has additional dangerous goods transportation capabilities, explosion-proof measures, or temperature control functions.

[0044] The simplified load parameters of the external transport capacity include the maximum loadable weight, the maximum loadable volume, and transport qualification information, which are usually provided by external carriers to quickly determine whether the cargo is overweight or whether there are situations where the dangerous goods or cold chain qualification matching degree is insufficient.

[0045] In the present application, by classifying the transport capacity information into accurate vehicle cabin position parameters and simplified load parameters, the system can flexibly connect different types of vehicles in the subsequent loading link and be compatible with heterogeneous transport capacity environments.

[0046] After obtaining the above information, preprocessing operations can be performed on the order set, such as merging the same origin and destination, splitting large orders, or verifying missing data, to form a preprocessing order set that is more suitable for algorithm processing.

[0047] For the above S102:

[0048] Use a multi-objective scheduling algorithm to allocate vehicles and routes for the order set and generate several candidate scheduling plans.

[0049] As an optional implementation manner, the multi-objective scheduling algorithm includes:

[0050] Establish an objective function in the multi-objective scheduling algorithm; among them, the optimization indicators in the objective function include: freight cost, transportation timeliness, vehicle empty running rate, and compliance requirements for special goods;

[0051] Set hard constraint rules in the multi-objective scheduling algorithm; the hard constraint rules include:

[0052] In response to the cargo attributes including dangerous goods, it can only be assigned to its own transportation capacity with corresponding protective configurations or external transportation capacity with dangerous goods transportation qualifications; in response to the cargo requiring cold chain, it can only be assigned to transportation capacity with corresponding refrigeration functions;

[0053] Set soft constraint rules in the multi-objective scheduling algorithm; the soft constraint rules include:

[0054] In response to the size or weight of the order cargo exceeding the accurate vehicle cabin parameters of the own transportation capacity or the simplified load parameters of the external transportation capacity, add a penalty value to this assignment relationship in the objective function to reduce the fitness of this assignment relationship in the iterative search.

[0055] As an optional implementation manner, the generation of the candidate scheduling plan includes:

[0056] Perform iterative solution on the preprocessed order set to generate multiple candidate solutions, and each candidate solution includes the assignment relationship to the own transportation capacity and / or external transportation capacity, the driving route, and the estimated arrival time;

[0057] In response to detecting an allocation combination that does not meet the constraint rules during the scheduling iteration process, eliminate or mutate this candidate solution in the algorithm iteration until a candidate scheduling plan that meets the requirements of the objective function is output.

[0058] In specific implementation, this multi-objective scheduling algorithm conducts comprehensive evaluation and iterative search on multiple indicators such as freight cost, transportation timeliness, vehicle empty running rate, and compliance requirements for special goods. Among them, if it is dangerous goods, the algorithm preferentially assigns them to vehicles with corresponding explosion-proof cabins or dangerous goods transportation qualifications. If it is cold chain goods, it is assigned to vehicles with corresponding refrigeration functions; if the size or weight of the goods far exceeds the basic load or volume upper limit of the vehicle, a penalty value is imposed on this assignment in the objective function, so that this combination is gradually eliminated during the iteration. In this way, the multi-objective allocation of vehicles and routes can be completed from a large level first, enabling the scheduling layer to give a candidate plan in a relatively short time.

[0059] Regarding the above S103 and S104:

[0060] After the candidate scheduling plan is generated, the system will perform a three-dimensional loading feasibility check on it to check whether each vehicle can accommodate the allocated goods according to the previous assignment. If the vehicle belongs to the own transportation capacity and has accurate vehicle cabin parameters, the three-dimensional stacking or collision detection method can be called, combined with the three-dimensional information in the cabin, the sub-cabin structure, and the protection configuration information, to simulate the placement of the length, width, and height of the goods or more fine-grained external shape data to determine whether the vehicle has enough space, corresponding sub-cabin functions, and protection levels to accommodate the goods.

[0061] If the vehicle is an external transportation capacity, only a simple compliance determination can be made based on the maximum loadable weight, volume, and transportation qualification information. At this time, in the refined verification of the system, if it is found that the goods are incompatible with the vehicle compartment, the requirements for dangerous goods or cold chain cannot be met, the space is insufficient, or the dimensions are seriously mismatched, etc., a result of failed loading verification will be generated. This result will be immediately transmitted back to the multi-objective scheduling algorithm, enabling it to update the assignment plan or replace the vehicle in the next iteration; if all loading verifications pass, the candidate scheduling plan will be marked as feasible. Finally, after all goods and vehicles are successfully matched at the three-dimensional loading level, the system will output the final scheduling plan, which not only ensures the achievement of the macro scheduling goal but also minimizes the number of rollbacks at the loading level as much as possible.

[0062] Exemplarily, in a typical intelligent transportation assignment and scheduling system, the software platform usually consists of an order management module, a transportation capacity management module, a multi-objective scheduling module, a three-dimensional loading verification module, and a data storage and interaction interface.

[0063] First, the order management module receives order data pushed by an external system (such as an enterprise ERP or an online ordering platform) and stores it in the order table in the database. Each order record contains at least the following fields:

[0064] Goods attributes: including length, width, and height data or volume / weight markings, dangerous goods identification (if any), temperature range requirements (if it is a cold chain), etc.;

[0065] Shipping and receiving locations: can be marked as source and destination coordinates in the geocoding library;

[0066] Time requirements: such as the latest departure time, the earliest arrival time window, etc., which are used to generate a feasible time period allocation in the subsequent scheduling algorithm.

[0067] Then, the transportation capacity management module provides transportation capacity resource information for its own vehicles and external vehicles:

[0068] For its own vehicles, accurate vehicle compartment parameters are retained in the "own transportation capacity table" in the database, that is, the three-dimensional space information inside the compartment (such as length, width, height, and local concavity and convexity descriptions), sub-compartment structure information (characterized by partition IDs or partition coordinates), and protection configuration information (cold chain unit model, explosion-proof level, sensors, etc.);

[0069] For external vehicles, only the maximum loadable weight, maximum transportable volume, and transportation qualification fields (indicating whether cold chain or dangerous goods can be transported, etc.) are stored in the "external transportation capacity table" in the database, and are provided in a simplified form for subsequent scheduling calls.

[0070] After obtaining the order and transportation capacity information, the system enters the multi-objective scheduling module to perform preprocessing before scheduling. This may include:

[0071] Order merging or splitting: If the shipping locations, receiving locations, or time windows of multiple orders are similar, they are merged into a larger unit; if a single large order exceeds the load / volume of any one vehicle, it is split.

[0072] Basic verification: For example, remove expired or unavailable vehicles from the available time table of external carriers and the vehicle table, and update the remaining schedulable resources.

[0073] After generating the preprocessed order set, the multi-objective scheduling module starts to perform vehicle and route allocation. The following key steps are included in this module:

[0074] For the definition of the objective function, in an extensible data structure, record the weights of freight costs, transportation timeliness, vehicle empty running rate, and compliance requirements for special goods;

[0075] For constraint setting, enter hard rules (such as dangerous goods can only be assigned to vehicles with protection configurations or qualifications) and soft constraint rules (such as imposing penalty values when the weight or volume of goods exceeds the vehicle limit);

[0076] For iterative solution, initialize several candidate solutions, each candidate solution describes "order ID vehicle ID" and "driving route information", and then perform iterative search through algorithms such as genetic algorithms and ant colony algorithms. Calculate the comprehensive fitness (including freight, timeliness, empty running rate, compliance, etc.) for the candidate solutions of each iteration, and filter or penalize unreasonable assignments according to hard / soft constraints. After multiple iterations of the algorithm, output a batch of suitable candidate scheduling plans.

[0077] After generating the candidate scheduling plans, the system passes these plans to the three-dimensional loading verification module for feasibility verification. In this module, the "order-vehicle allocation" in each candidate plan is checked as follows:

[0078] If assigned to an external vehicle, judge feasibility only according to its maximum loadable weight, volume, and transportation qualifications; if it exceeds the limit or there is no corresponding qualification, mark the plan as infeasible and return it to the scheduling module for fallback and correction;

[0079] If assigned to an owned vehicle, call the three-dimensional information inside the owned vehicle stored in the data structure (such as in the form of a grid or point cloud), the sub-compartment structure, and the protection configuration, and combine the length, width, and height or three-dimensional point cloud of the goods in the order to perform specific placement or collision detection:

[0080] First, check the sub-compartment identification: if it is cold-chain goods, match the cold-chain compartment area; if it is dangerous goods, match the explosion-proof compartment area;

[0081] Then, perform three-dimensional stacking simulation in the corresponding cabin area to determine whether the goods can occupy a certain coordinate space without collision in this cabin area; if sufficient space cannot be found or the shape of the cabin is in serious conflict with the shape of the goods, mark that the loading verification fails.

[0082] The system reads the above loading verification results in real time: if the verification fails, label the vehicle assignment relationship as "invalid" and return it to the scheduling module for fallback or update; if the verification passes, continue to perform the same verification on other assignments until all orders pass the loading verification. Finally, if each vehicle-cargo assignment in the candidate solution meets the loading requirements, the system outputs this scheduling plan as the final result. During this process, for the company's own vehicles that are prone to in-cabin modification or attachment addition, the platform can periodically perform lidar scans to update the three-dimensional data of their cabin positions to avoid misjudgment caused by cabin deformation or attachment installation.

[0083] Through this hierarchical method of first scheduling, then loading verification, fallback and correction in case of verification failure, and outputting the plan in case of verification success, this embodiment can effectively balance the global optimization requirements of large-scale scheduling algorithms for timeliness and vehicle empty running rate, and can also match the refined demands of goods for three-dimensional space, protection requirements, and temperature control requirements in terms of micro loading details. Especially for the accurate vehicle cabin parameters of the company's own transport capacity, higher-precision three-dimensional simulation can be used to ensure accurate placement judgment for dangerous goods, cold chain, or oversized goods; for external transport capacity, only the maximum load and volume are used for simplified processing to accommodate the situation of insufficient transport capacity information of social carriers. In this way, a complete closed-loop is formed from multi-objective scheduling to three-dimensional loading, then to fallback correction and result output. This method realizes the transportation assignment scheduling that takes into account the requirements of efficiency, cost, and safety compliance in the face of complex orders and multiple transport capacities, and reduces the frequent failures and manual operation burdens caused by mismatches or neglect of cabin structures.

[0084] Please refer to Figure 3 , Figure 3 which is a schematic diagram of the impact of an abnormal area on the placement of goods provided by an embodiment of this application. Figure 3 In (a), it is the cabin space without an abnormal area, Figure 3 in (b), it is the cargo hold space with an abnormal area. In Figure 3 in (b), the abnormal area has a significant impact on the placement of goods. Therefore, as an optional implementation manner, obtaining the accurate vehicle cabin parameters of the company's own transport capacity includes:

[0085] Obtain the internal image of the cabin including the key areas; wherein, the key areas include: the walls, floor, corners, and door frames of the cabin;

[0086] Perform two to three layers of two-dimensional discrete wavelet transform on the internal image of the cabin to decompose the low-frequency sub-band and high-frequency sub-band;

[0087] Among them, the low-frequency sub-band is used to characterize the overall structure and background information of the interior image of the vehicle cabin; the high-frequency sub-band is used to characterize the edge, texture, and detail information of the interior image of the vehicle cabin;

[0088] Calculate the amplitude of the wavelet coefficients in the high-frequency sub-band, and apply adaptive threshold processing to obtain abnormal features.

[0089] Use inverse wavelet transform to map the abnormal features back to the original image space, apply morphological closing operation to merge adjacent abnormal points and remove isolated noise points to obtain abnormal region localization information;

[0090] Use a pre-trained deep learning model to predict the depth information of the abnormal region from a single image, generate a depth map, and convert the depth map into three-dimensional parameters of the abnormal region in combination with the camera internal parameters;

[0091] Based on the three-dimensional parameters of the abnormal region, calculate the volume of the vehicle cabin space occupied by the abnormal region, and deduct this volume from the original design size of the vehicle cabin to generate accurate vehicle cabin position parameters of the available transport capacity.

[0092] This embodiment aims to detect and calculate the volume of abnormal regions through image processing and deep learning means for possible deformation or additional accessories in the vehicle cabin after multiple uses, so as to provide more real-time and accurate technical support for obtaining accurate vehicle cabin position parameters in S101 above.

[0093] Its basic principle is as follows: First, use two-dimensional discrete wavelet transform to decompose the interior image of the vehicle cabin in the key region into a low-frequency sub-band and a high-frequency sub-band, and then perform adaptive threshold processing on the wavelet coefficients with larger amplitudes in the high-frequency sub-band to extract "abnormal features" reflecting concavity, convexity, or accessory protrusions. Subsequently, obtain the three-dimensional parameters of the abnormality through morphological and depth estimation prediction, and deduct the abnormal volume from the original vehicle cabin design size to generate accurate vehicle cabin position parameters.

[0094] In practice, several key regions can be determined according to vehicle manufacturers or internal maintenance records, such as the walls, floors, corners, and door frames of the vehicle cabin. These parts are more likely to have local protrusions or depressions due to impacts, loading of special goods, or installation of cold chain units, explosion-proof panels, etc. First, obtain the interior image of each key region. This image can be captured by cameras from multiple angles or regularly obtained by the vehicle cabin internal monitoring system to ensure that the latest changes in the vehicle cabin are captured. Perform two-dimensional discrete wavelet transform on the obtained interior image of the vehicle cabin, and the decomposition level can be selected from two to three layers.

[0095] After decomposition, the low-frequency sub-band and high-frequency sub-band can be obtained. The low-frequency sub-band here is used to characterize the overall structure of the cabin and background information, retaining the regional brightness or chromaticity distribution to reflect the main contours and main planes of the cabin walls or floors; the high-frequency sub-band is used to reveal the details, textures, and edge information of the image, and can magnify the features of certain local anomalies (such as protrusions, cracks, or attachment seams).

[0096] In the high-frequency sub-band, the degree of local variation is characterized by calculating the amplitude of the wavelet coefficients. Adaptive threshold processing is performed on the coefficients whose amplitudes exceed a certain range among several segmented thresholds, so as to distinguish random noise and weak textures from real obvious protrusions or depressions. If the high-frequency coefficient only slightly exceeds the threshold, it is regarded as general texture, and if the high-frequency coefficient far exceeds the threshold, it is marked as an "abnormal feature". Then these marked high-frequency coefficients are mapped back to the original image space during the inverse wavelet transform to determine the distribution of potential abnormal regions in pixel coordinates.

[0097] Usually, these high-frequency anomalies may be caused by newly assembled brackets in the cabin, protruding parts of cold chain units, or other sunken floors. To remove the isolated noise points that may still remain and merge adjacent regions, a morphological closing operation can be performed on the abnormal region obtained by this inverse transform to fill small gaps, merge adjacent pixel clusters, and finally obtain the segmentation result of the connected abnormal region.

[0098] After obtaining the two-dimensional segmentation of the abnormal region, to further quantify its three-dimensional occupied volume, the mapping from the image to depth information is needed.

[0099] In this embodiment, a deep learning model pre-trained on a vehicle or indoor scene can be used to estimate the depth of the abnormal region. The model combines the texture clues, geometric priors of a single image, and the camera internal parameters (including focal length and distortion coefficients, etc.) to predict the depth map of each pixel point.

[0100] Subsequently, by traversing and three-dimensionally projecting the depth map within the abnormal region, a point cloud or three-dimensional contour of the abnormal region in the three-dimensional coordinate system is obtained, and then the abnormal volume is calculated using voxel accumulation or polygon construction methods. Because the model is trained for indoor or vehicle cabin environments, it can better recognize structures such as cabin supports and metal attachments than ordinary general models. The three-dimensional parameters of these abnormal regions represent the newly emerged or increased / decreased protruding parts and local space occupancy in the cabin.

[0101] By deducting the above abnormal region volume from the original design size of the cabin, the actual remaining available space of the cabin can be updated. If there are multiple anomalies in several key regions, repeat this process to calculate the volume of each abnormal block separately and merge or superimpose and deduct them, and finally generate more realistic three-dimensional cabin information, thus forming the "accurate vehicle cabin parameters of its own transport capacity" mentioned in this embodiment.

[0102] This parameter will be used in subsequent 3D stacking or collision detection of goods to determine whether the goods can be placed in the actually available cabin space, rather than relying on outdated original factory CAD or manually filled load data. This significantly improves the ability to automatically identify and update the available space volume after deformation or attachment addition inside the vehicle cabin, providing more reliable cabin basic data for intelligent transportation assignment and scheduling.

[0103] In this way, for the technical problem that "abnormal areas" such as unknown protrusions, depressions, and accessory supports are difficult to accurately identify by traditional measurement methods during multiple uses or additional assembly processes of vehicles, by obtaining internal cabin images, extracting high-frequency subbands, adaptively threshold detecting significantly protruding or depressed points, and then using a deep learning model for 3D reconstruction, it is possible to quickly and automatically deduct these additional or deformed volumes from the original design dimensions of the vehicle, and then generate more accurate vehicle cabin parameters closer to the actual situation.

[0104] On the one hand, traditional manual measurement or static CAD models often cannot update in real time the temporary brackets, protrusions of cold chain units, or depressed areas that may appear inside the vehicle; on the other hand, simply performing simple filtering on images is also difficult to distinguish real protrusions from noise or texture patterns. However, this application uses multi-level denoising and feature point extraction of two-dimensional discrete wavelet transform plus morphological closing operation to accurately locate the parts that truly protrude or are depressed in the high-frequency subband, and convert them into quantifiable 3D volumes through depth estimation, ultimately making the update of cabin data more dynamic and accurate.

[0105] If other means are tried to solve similar problems of "temporary or dynamic attachments inside the vehicle cabin" identification, common methods include relying on manual on-site measurement, based on fixed laser ranging scripts, or simply using simple Gaussian filtering to remove texture noise. However, these methods are difficult to distinguish real attachments from random noise points and often lack the ability of automatic depth estimation for local protrusions. In contrast, this application focuses on obvious mutation points in the high-frequency subband, excludes non-critical texture noise through adaptive thresholding, and uses morphology and pre-trained depth model prediction to accurately quantify the 3D position and volume of abnormal areas. It can not only effectively ignore fine interference but also retain the main protruding components that affect safety and load, giving full play to the dual advantages of wavelet multi-scale denoising and deep learning scene recognition. This not only reduces the dependence on manual measurement but also avoids the limitations of simple thresholding or static CAD models, with significant improvements in real-time performance and accuracy in a transportation environment with a high data update frequency.

[0106] In addition, compared with directly using a pure prediction model (e.g., an end-to-end image semantic segmentation or depth estimation network) to detect and quantify abnormal areas such as bumps and depressions inside the cabin, in this application, by introducing an adaptive threshold in the high-frequency subband to locate abnormal features with significant amplitudes and removing a large amount of irrelevant textures in the wavelet domain, the input of the subsequent prediction model can be made "purer", reducing the dependence on the depth model and the demanding of the training scale.

[0107] Secondly, when directly relying on an end-to-end network to process cabin images, if there are diverse materials, reflections, or new attachment forms in the cabin, without sufficient training samples, the model may regard some key bumps as background or noise, or judge a normal cabin wall with detailed textures as abnormal. By performing adaptive threshold detection in the wavelet domain, marking the high-frequency responses that truly exceed the threshold as "suspicious bumps or depressions", and then supplementing with morphological operations to merge connected regions, the subsequent depth model can be made to only focus on the real suspected regions, thus significantly reducing the dependence of the end-to-end network on large-scale diverse scenarios.

[0108] Finally, in a business scenario, when the transportation environment, cabin lighting, or image perspective change frequently, a single prediction model often needs to be continuously iteratively optimized or incrementally trained; while the solution combining wavelet analysis and morphological means can filter most external light and low-level texture interferences in a more robust manner, maintaining a stronger robust detection ability for "significant deformations".

[0109] For cabin scenarios that require real-time updates or frequent scans, this method can obtain a high abnormal detection rate without heavily relying on an artificial annotation database, and only call depth prediction in suspected areas, significantly reducing the deployment and maintenance costs. In contrast, if completely relying on a single deep learning model, once new attachment forms or rare concave-convex structures appear inside the cabin and there is a lack of supporting data for iterative training, the model accuracy will decrease significantly.

[0110] Exemplarily, after a self-owned transport vehicle has been used for several months and undergone several modifications, an additional cold chain unit and some metal brackets have been successively installed inside, resulting in changes in height or shape in some areas inside the cabin, and there is a large error between the original CAD data and the actual situation. To maintain an accurate grasp of the space inside the vehicle cabin, the system performs the following operations to update the accurate vehicle cabin position parameters of the self-owned transport capacity:

[0111] When the vehicle returns to the depot for maintenance, maintenance personnel install a movable camera combination in the key areas inside the cabin (including cabin walls, floors, corners, and door frame positions), and obtain multiple images of the inside of the cabin for each key area. Since the camera has the ability to take pictures from multiple angles, it can cover most of the parts protruding due to metal brackets or cold chain units, and ensure that the images can capture the cabin walls that are prone to deformation or impact.

[0112] For each acquired interior image of the vehicle cabin, the system first performs a two-dimensional discrete wavelet transform with the decomposition level set to three layers to obtain the corresponding low-frequency subband and three high-frequency subbands.

[0113] In implementation, the software module uses the Daubechies series of wavelet bases, which can better balance image detail capture and noise robustness. The decomposed low-frequency subband mainly retains the brightness gradients of the large-area structures in the vehicle cabin (such as the cabin wall plane and the main floor plane), while the high-frequency subbands highlight texture details such as the edges of metal brackets, the connections of cold chain units, and the concave traces caused by certain impacts.

[0114] The system then scans the coefficient magnitudes of all high-frequency subbands and excludes the extremely small coefficients as texture noise according to the set multi-segment adaptive thresholds, marks the high-frequency coefficients exceeding the first threshold as general protrusions, and marks the high-frequency coefficients exceeding the second threshold as significant abnormal feature points. For example, the high-frequency coefficient magnitudes at the seams between the metal brackets and the cold chain units are relatively large and are marked as "significantly abnormal" at this stage. During the inverse wavelet transform, these marked high-frequency coefficients are mapped back to the coordinates of the original image, enabling the algorithm to determine the distribution of abnormal regions in the pixel domain.

[0115] To remove isolated noise points and merge adjacent or connected abnormal regions, the system performs a morphological closing operation on the "abnormal segmentation result" obtained after the inverse transform, deletes small noise points based on pixel connectivity judgment, merges adjacent regions into a relatively complete contour, and finally obtains a two-dimensional abnormal segmentation map. At this time, it can be seen that one or more relatively large connected regions are formed near the metal brackets and at the prominent positions of the cold chain units.

[0116] After that, the system calls a depth estimation network pre-trained in an indoor or vehicle cabin scenario to perform depth prediction on this two-dimensional abnormal segmentation map. The network combines the camera internal parameters (including focal length, distortion coefficients, etc.) and depth priors to perform single-image depth inference and outputs a depth map corresponding to the abnormal region. The software module further projects each pixel of the depth map into a three-dimensional coordinate system, thereby generating point cloud data of the suspicious protrusions or depressions in the algorithm memory. A voxel or polygon construction method is performed on this point cloud, and finally the three-dimensional volume size of this abnormal region is calculated.

[0117] Since the vehicle originally has a CAD design size or an "initial three-dimensional volume value of the vehicle cabin" recorded at the factory, the system deducts each of the multiple abnormal volumes calculated above one by one. On the one hand, if there are accessory protrusions, they will occupy the originally available space inside the vehicle cabin. On the other hand, if there are concave or damaged parts (such as a sunken floor), they will also change the distribution of available space. In this way, the three-dimensional space data of the vehicle cabin that is closer to reality than the original CAD data can be updated, and a new "accurate vehicle cabin parameter of its own transport capacity" can be formed, which includes a specific description of the space occupied by the cold chain unit protrusion and the metal bracket.

[0118] In the subsequent transportation assignment process, the scheduling algorithm can accurately avoid metal brackets or protrusions when simulating the stacking of goods by reading the updated three-dimensional cabin parameters, thus avoiding ineffective scheduling. If a certain cargo needs to be placed specially or a safety gap needs to be left (such as the requirements for the explosion-proof area of dangerous goods), more refined placement simulation can also be carried out based on the accurate cabin structure information, greatly reducing the loading conflicts or safety hazards caused by insufficient understanding of cabin deformation.

[0119] In this way, most texture noises and small interferences are excluded in the wavelet domain first, and the real protrusions / depressions are centrally marked for the depth model to process. Even if there are diverse materials and reflective conditions inside the vehicle cabin, a high abnormal detection rate and three-dimensional positioning accuracy can be maintained, effectively supporting the accuracy and safety requirements of large-scale transportation intelligent assignment scheduling at the micro-loading level.

[0120] As an optional implementation manner, the cargo attributes include: three-dimensional shape data of the order cargo; performing three-dimensional loading feasibility verification on the candidate scheduling plan includes:

[0121] Based on the image processing result inside the vehicle cabin, obtaining the three-dimensional shape data of the vehicle cabin;

[0122] Among them, the three-dimensional shape data includes: the overall cabin wall structure and sub-cabin basic information characterized by the low-frequency sub-band, and the three-dimensional parameters of the abnormal area obtained by inverse transformation and merging of the high-frequency sub-band;

[0123] Inputting the three-dimensional shape data of the vehicle cabin and the three-dimensional shape data of the order cargo into the multi-grid construction module respectively, and generating hierarchical discrete expressions of the coarse grid layer and the fine grid layer in sequence;

[0124] In the coarse grid layer, globally screening for collisions between the vehicle cabin and the cargo according to the voxels of the preset unit;

[0125] In response to detecting conflict information between the cargo and the cabin wall or the abnormal area, marking the loading as infeasible and rolling back to the multi-objective scheduling algorithm;

[0126] In response to the absence of detected conflict information, the local concavity and convexity data of the cargo surface and the vehicle compartment are compared one by one at the fine grid layer for stacking simulation. In response to any cell collision detected at the fine grid layer, the loading is marked as infeasible and the algorithm is reverted to the multi-objective scheduling algorithm. In response to no cell collision detected at the fine grid layer, a passed feasibility check is output.

[0127] In specific implementation, to process the three-dimensional data of the cargo and the vehicle compartment simultaneously in the three-dimensional loading feasibility check phase, the order management module maintains a "cargo three-dimensional data record" for each order, which can be named "cargo three-dimensional model". If the cargo has a regular shape, it can be directly described by its length, width, height or simplified volume. If the cargo is an irregular object, its outer surface point cloud or mesh information can be obtained through multi-angle scanning and stored in the "cargo three-dimensional model" entry. This record includes the geometric shape, size and (if any) convex or curved surface features of the cargo, which are used for subsequent collision detection or stacking simulation.

[0128] Meanwhile, the transportation capacity management module stores the "vehicle compartment three-dimensional model" in the database, including the "bulkhead basic information" and "abnormal area information" obtained above.

[0129] Specifically, the "bulkhead basic information" corresponds to the overall structure and compartment parameters characterized by the low-frequency sub-band, and the "abnormal area information" comes from the accessory protrusions or depressions detected after the inverse transformation of the high-frequency sub-band. The system manages the internal coordinate distribution, bulkhead position, compartment function, and "abnormal area coordinates" of metal brackets, cold chain units, etc. that may exist in the form of a "vehicle three-dimensional data table" in the database.

[0130] When performing a three-dimensional loading feasibility check on the candidate scheduling plan, the scheduling module sends the "vehicle compartment three-dimensional model" and the "cargo three-dimensional model" to the "multi-grid construction module" respectively. This module first generates a "coarse grid layer" according to the preset space segmentation rule (for example, the voxel size is 10 cm), and then generates a "fine grid layer" with smaller units (for example, 2 cm or 1 cm) within the range that requires fine simulation, so as to form a "two-layer or multi-layer hierarchical discrete representation".

[0131] At the coarse grid level, the system performs a "global collision screening" on the vehicle compartments and the goods, dividing the vehicle compartments into several large-sized cubic units (referred to as "coarse voxels"); the basic information of the bulkheads or the information of abnormal areas is marked. If a cubic unit is regarded as a "bulkhead unit" or an "abnormal unit", its coordinate range is unavailable; conversely, an "available unit" means that loading can be carried out. The three-dimensional model of the goods is also divided proportionally, mapping the shape of the goods into corresponding coarse voxels. The system compares the coordinate intervals of each cubic unit of the goods and the vehicle one by one. If it is found that the goods unit overlaps with the "bulkhead unit" or "abnormal unit" of the vehicle, a conflict is marked at this level, "loading infeasible" is output, and a fallback instruction is sent to the multi-objective scheduling algorithm; if there is no overlap in all comparisons, the coarse grid level initially determines feasibility and allows entry into the fine grid level for more accurate simulation.

[0132] After entering the fine grid level, the system divides both the vehicle compartments and the goods more finely. For example, smaller granularity voxels or polygon mesh units are further subdivided within the original coarse voxels. At this time, the "abnormal area information" of the vehicle compartments can further restore the true contours of the metal brackets or sunken floors, and the local curved surfaces in the shape of the goods are also discretized at a high resolution in the same way. The system compares the surface coordinates of the goods with the coordinates of the vehicle compartments unit by unit: if any fine unit overlaps or interpenetrates, a collision is marked and the loading is determined to be infeasible; only when no collision is detected at the fine grid level, it is finally determined that the goods can be safely and undisturbedly placed in the vehicle compartments, and the loading feasible result is output.

[0133] During the implementation process, an "occupied flag" or "idle flag" can be set for each voxel unit, and its coordinates or numbers are recorded; when the goods unit and the vehicle unit have the same coordinate number and one of them is "occupied", a conflict occurs; when there is no such situation, the comparison continues to the next stage. When the vehicle contains a cold chain compartment or a dangerous goods compartment, the fine grid level will also attach a "compartment type flag" to the corresponding unit. If the goods have the attributes of cold chain or dangerous goods, they need to match it, otherwise a non-compliance conflict is determined.

[0134] Through this double-layer discretization method, first use the coarse grid screening to quickly filter out large-scale spatial mismatches, and then the fine grid level detects the fit of local concavities and convexities and compartment configurations cell by cell, greatly reducing the high-complexity full-scale fine collision calculation amount, and providing a feasible and fast loading determination mechanism for large-scale orders and multi-type vehicle scenarios. If the fine grid still detects a conflict, immediately return "loading infeasible" to the multi-objective scheduling algorithm to re-correct the plan; if there is no conflict at the fine grid level, it indicates that safe loading is also possible at a higher accuracy, and finally a "feasibility verification passed" result is given.

[0135] It should be emphasized that in this embodiment, if the three-dimensional shape data of the ordered goods is only the simple length, width, and height, the module can quickly construct it into an approximate cuboid voxel; if there is a more refined shape (such as a circular metal device), point cloud or STL mesh files can be used at the goods end for fine grid discretization; if the goods are dangerous chemicals, corresponding safety gaps should be left in the grid structure or special markings should be made according to the requirements of explosion-proof sub-cabin to strictly avoid approaching ordinary goods or vehicle cabin units without explosion-proof configuration during local surface fitting. Through this multiple grid construction and collision detection, the safety and feasibility of the micro-loading link of intelligent transportation assignment and scheduling are ensured in a relatively high-precision and efficient manner throughout the process.

[0136] As an alternative embodiment, performing three-dimensional loading feasibility verification on the candidate scheduling plan further includes:

[0137] In response to data on predicted deformation or additional attachments of the vehicle cabin or goods, map the predicted volume or deformation area into a ghost volume during multiple grid construction and mark it in the corresponding coordinate units of the coarse grid layer and the fine grid layer;

[0138] During the global collision screening, in response to the overall size of the goods overlapping with the coordinate units of the ghost volume, it is determined that there will be no available space after subsequent deformation occurs, mark the loading as infeasible and roll back to the multi-objective scheduling algorithm;

[0139] In response to the overall size of the goods not overlapping with the coordinate units of the ghost volume, further compare the local surface of the goods with the local units of the ghost volume at the fine grid layer. In response to the existence of fitting conflicts, mark the loading as infeasible and roll back to the multi-objective scheduling algorithm. In response to the absence of fitting conflicts, output that the feasibility verification has passed.

[0140] In this embodiment, in order to anticipate early empty space or conflicts for "predicted deformation or additional attachments" during the three-dimensional loading feasibility verification stage, the system introduces the concept of "ghost volume" during the multiple grid construction process and marks it in the corresponding coordinate units of the coarse grid layer and the fine grid layer to simulate the space that may be occupied by the vehicle cabin or the volume that the goods may expand during the subsequent journey.

[0141] In specific implementation, there are entries of "data for predicting deformation or additional attachments" at the cargo end and the vehicle compartment end respectively. If the vehicle end predicts that a cold chain unit, bracket or other attachments will be added in the next stage, or there is a bulge or depression somewhere due to metal fatigue, then this part will be given an "attachment addition identifier" and a coordinate area; if the cargo end predicts that it will expand or extend in shape by a certain proportion under environmental conditions such as temperature and humidity, the system will generate "expanded size" or "expanded volume" information in the three-dimensional data of the cargo. To ensure a response to this prediction during the scheduling and loading process, when the system constructs multiple grids, these predicted volumes (or local coordinate areas) are mapped as "phantom volumes" and loaded together in the grid coordinate system.

[0142] Subsequently, the system first completes a "global collision screening" at the coarse grid level: if any voxel unit after the discrete coarse grid of the overall size of the cargo overlaps with the voxel unit of the phantom volume, it can be determined that there will be no available space for the cargo during subsequent deformation or attachment addition. The system immediately determines that the loading is not feasible and sends a fallback signal to the multi-objective scheduling algorithm to avoid investing more computing power in the subsequent fine grid level for precise judgment; if no overlap is detected at the coarse grid level, it indicates that there is no obvious conflict between the cargo and the phantom volume at the overall scale, and it can enter the fine grid level for a more accurate local fitting check.

[0143] At the fine grid level, the system makes a finer-grained cell division of the three-dimensional shapes of the vehicle compartment and the cargo (for example, further dividing into 2-cm or 1-cm voxels within the upper 10-cm voxel), and also marks the coordinate range of the "phantom volume" in the corresponding coordinate area. At this time, a "fitting degree comparison" or "collision detection" will be performed for the local part of the cargo surface and the phantom volume: if the cargo coordinates of any fine grid cell overlap with the voxel unit of the phantom volume, it means that the cargo will inevitably collide or occupy the area where the attachment is added during the future period or when the predicted deformation occurs, and the system marks the loading as not feasible; only when there is no overlap in all cell comparisons at the fine grid level, is it considered that the cargo can still coexist safely with the interior space of the vehicle cabin after the predicted deformation or attachment addition, so the feasibility check is output as passed.

[0144] Through this hierarchical marking method of double-layer grids and "phantom volumes", this implementation method can not only quickly eliminate most of the solutions with no available space at the coarse grid level, but also perform a higher-precision conflict check on the cargo and the future occupied space (such as cold chain units, raised panels or the self-expansion area of the cargo) at the fine grid level, reducing the problem of space extrusion caused by temporary installation or deformation during subsequent trips. If any conflict is detected, it immediately falls back to the multi-objective scheduling algorithm for assignment correction or re-splitting of the cargo; if the detection passes, it is marked as feasible, enabling the system to still obtain a stable and effective loading allocation result when facing the dynamically changing interior environment of the vehicle cabin and the cargo that will deform.

[0145] Exemplarily, after the initial coarse grid processing, the system has divided the vehicle cabin area into "voxels" or "small unit grids", and each unit is recorded in the "vehicle cabin fine grid table".

[0146] If there is a predicted deformation or an additional attachment in the vehicle cabin, the system first writes a "ghost volume identifier" and corresponding "occupied" or "to be left empty" marks on the corresponding coordinate unit in the "vehicle cabin fine grid table", and can also attach a safety clearance value (such as a 3 cm distance needs to be reserved).

[0147] For the bulkhead body or abnormal protrusion areas (such as metal brackets), "unavailable" or "to be detected" marks are also made on the corresponding units in the "vehicle cabin fine grid table" to distinguish them from available space units.

[0148] The system also adopts the same fine grid resolution in the "3D cargo model", divides it into several small units according to each coordinate, and records them in the "cargo fine grid table"; each cargo unit contains a "grid ID" (such as 3D indexes i, j, k), "boundary coordinates" (such as the lowest / highest xyz), etc.

[0149] If the cargo has its own predicted expansion, its shape is directly expanded according to the "expanded contour" during the fine grid generation stage, so that the divided voxel units cover the part that will grow in the future, so that the cargo shape is larger during subsequent collision detection.

[0150] At this time, the "cargo fine grid table" and the "vehicle cabin fine grid table" have a comparable coordinate system (usually under the same unit length or corresponding coordinate system).

[0151] The system traverses each unit (denoted as C_g) in the "cargo fine grid table" and extracts its central coordinates or spatial boundaries therein;

[0152] Then, in the "vehicle cabin fine grid table", the corresponding coordinate unit (denoted as C_v) is located according to the central coordinates or boundaries; if there is any overlap between the unit boundary of the cargo and the "ghost volume identifier" unit or "unavailable" unit of the vehicle cabin, it is recorded as a collision.

[0153] The Axis-Aligned Bounding Box (AABB) principle can be used to judge the overlap: if the projections of two cubes have interval overlaps in each dimension of X, Y, and Z, a collision can be determined.

[0154] If a collision is detected, the system immediately determines that the loading is not feasible and rolls back. If no collision is found during the entire traversal, it means that the cargo is compatible with the vehicle cabin at this resolution.

[0155] If a certain safety gap needs to be left, for example, a 3-cm gap needs to be left around the cold chain unit, the system will expand several grids outward for the relevant ghost volume units in the fine grid of the vehicle compartment to form a "safety zone", which is marked as "untouchable".

[0156] Exemplarily, in the fine grid table of the vehicle compartment, around the coordinate units added for the ghost volume or accessories, an additional mark of "safety zone radius = " is added;

[0157] During collision detection, the system not only simply checks for coordinate overlap, but also checks whether the distance between the cargo unit and the center of the "safety zone" unit is less than ; if so, it is regarded as a conflict.

[0158] After the system finishes traversing all units and no conflicts or insufficient fits are found, it outputs "feasibility check passed", ends the check, and confirms the available solution.

[0159] In addition, to reduce excessive calculation, the fine grid operation is only performed when large-area conflicts have been excluded in the coarse grid; if the cargo volume is huge, the octree or octal hierarchical dissection method can be used to improve the search efficiency.

[0160] In actual application, for irregular cargoes (such as large mechanical components) or chemical containers that will expand at extreme temperatures, this combination of fine grid and safety gap inspection can significantly reduce the risk of extrusion or breakage during subsequent driving.

[0161] As an alternative implementation, performing a three-dimensional loading feasibility check on the candidate scheduling plan further includes:[[]]

[0162] At the coarse grid layer, perform a global collision screening on the cargo and the vehicle compartment based on the ghost volume;

[0163] In response to any voxel overlap detected in the global collision screening for any coarse grid unit, mark the loading as infeasible and roll back to the multi-objective scheduling algorithm;

[0164] In response to no voxel overlap detected in the global collision screening for any coarse grid unit, at the fine grid layer, compare the local surface of the cargo under the ghost volume with the concave and convex areas of the vehicle compartment grid by grid. In response to a fitting conflict, mark the loading as infeasible and roll back to the multi-objective scheduling algorithm. In response to no fitting conflict at the fine grid layer, confirm that the cargo can still be loaded smoothly after its expansion, and output that the feasibility check has passed.

[0165] In this embodiment, to further improve the hierarchical screening of 3D loading feasibility verification in the face of the "phantom volume" scenario, the system performs collision or fitting degree judgment on the coarse grid layer and the fine grid layer respectively to ensure that the goods can still achieve safe loading with high precision even after experiencing expansion or additional accessories in the vehicle compartment. The purpose is to quickly screen out large-scale spatial conflicts through brief global collision detection, and then accurately judge the microscopic relationship between the outer surface of the goods and the concave and convex parts inside the vehicle compartment with the help of fine fitting verification, so as to reduce the squeezing problem caused by deformation during transportation.

[0166] In terms of the specific principle, when the system detects the existence of goods expansion prediction information or vehicle compartment accessory addition identification, during the multi-grid construction stage, the corresponding voxel units in the coordinate system will be marked with the "phantom volume"; if the goods will also have a volume increase under conditions such as temperature and humidity, their 3D models will first expand their shapes before discretization and record them in a larger range in the grid table. Thus, the coarse grid layer can perform global collision screening with larger-sized voxels (such as 10 cm cubes): if any voxel unit of the goods overlaps with the phantom volume unit, the system immediately determines that there is no available space in the subsequent journey, marks the loading as infeasible and returns to the scheduling algorithm. This operation corresponds to "performing global collision screening of the goods and the vehicle compartment based on the phantom volume on the coarse grid layer". If any overlap of voxels is detected in any coarse grid unit, it will directly return, avoiding waste of subsequent operations.

[0167] When no conflict is seen in the coarse grid, the system enters the fine grid layer and performs voxel or grid discretization with a smaller granularity (such as 2 cm or 1 cm) on the same area. At this time, the vehicle compartment coordinates under the phantom volume are compared with the local surface of the goods grid by grid; because the fine grid can capture factors such as protrusions, depressions or safety gaps, the system will perform higher-precision fitting degree or collision detection at this level. If any overlap occurs in any fine grid unit at this time, that is, a fitting conflict occurs, the system marks the loading as infeasible and returns to the scheduling; only when no conflict is seen after the traversal of the fine grid layer, it is confirmed that the goods can still fit into the vehicle compartment after expansion. In this way, after the local verification of the fine grid is completed, since there is no fitting conflict, the system outputs "feasibility verification passed", enabling the goods to complete the loading smoothly. In engineering applications, this "double-layer screening" combined with the "phantom volume identification" has significant advantages: the coarse grid stage can quickly eliminate large-scale mismatches, and the fine grid ensures the microscopic safety space of the predicted deformation.

[0168] In one example, if the vehicle compartment predicts the installation of a cold chain unit bulge in the next stage, the coordinate range is marked as a ghost volume unit; the volume of the goods is estimated to increase by 5% due to chemical properties. In the coarse grid stage, global collisions are performed with 10-cm cubic voxels. If the overall coordinates of the goods overlap with the ghost volume, it is immediately determined to be infeasible; if there is no overlap, then in the fine grid stage, the surface boundaries of the goods and the coordinates of the ghost volume are compared one by one with 2-cm voxels; if a fitting conflict or insufficient safety gap is detected at any unit, it is determined to be infeasible and rolled back, otherwise it is finally determined to be feasible. Through this process, appropriate space can be reserved under the dual interference of predicting accessory occupancy and goods expansion, preventing the risk of overcrowding or impact during driving.

[0169] In this way, the coarse grid can quickly filter to avoid a large number of unnecessary fine operations, and the fine grid's precise fitting detection can strictly control the "space occupation" problem caused by temperature expansion or accessory installation, ensuring transportation safety and effective utilization rate.

[0170] As an alternative implementation, the transportation intelligent assignment and scheduling method further includes:

[0171] At the vehicle compartment end, a preset vehicle compartment deformation prediction model is periodically called. Based on the vehicle usage duration, the vehicle compartment metal fatigue coefficient, and the historical modification records, time series inference is performed to output data on the predicted deformation or accessory addition within the next stage itinerary of each free vehicle;

[0172] The data on the predicted deformation or accessory addition is marked with an accessory addition identifier or a metal plate bend, and the coordinate range is stored in the deformation data table;

[0173] At the goods end, based on the order information, the volume expansion rate or structural deformation probability information based on the goods material or chemical properties is extracted. Combining the predicted values of temperature, humidity, and pressure conditions during transportation, the goods deformation prediction module calculates the possible new outer contour dimensions of the goods and records them in the deformation data table in the form of expanded volume or extended dimensions;

[0174] In the multi-grid construction stage, based on the accessory addition identifier of the vehicle compartment, the metal plate bend of the vehicle compartment, the expanded volume of the goods, and the extended dimensions of the goods registered in the deformation data table, the predicted deformation area or extended form is written into the corresponding ghost volume coordinate unit respectively.

[0175] In this embodiment, in order to truly utilize the "predicted deformation / accessory addition data at the vehicle compartment end or goods end" in the loading feasibility verification, the system establishes a set of "deformation data table" and "ghost volume writing" mechanisms between the database and the multi-grid construction process.

[0176] In a specific implementation, at the vehicle cabin end, the "vehicle cabin deformation prediction model" is periodically called. This model can run on the server side or the local maintenance side. Using time series analysis or machine learning methods, it reads:

[0177] The vehicle usage mileage or duration (obtained from vehicle sensors or mileage data sheets), the vehicle cabin metal fatigue coefficient (obtained from historical sensor logs or metal stress tests), and the historical modification records (such as newly added brackets, modified cold chain modules, etc.).

[0178] Based on these input parameters, the model uses algorithms such as ARIMA (Autoregressive Integrated Moving Average) or LSTM (Long Short-Term Memory Network) to predict the sheet metal deformation amount of the vehicle cabin in the next stage (such as the next 2000 kilometers or the next maintenance cycle), and at the same time identify whether there are planned attachment addition coordinates (such as the height and installation position of the cold chain unit). If it is predicted that a certain area may have dents, bulges, or attachment occupancy, it outputs the approximate spatial coordinate range (usually in the form of a three-dimensional cuboid or point cloud) and the "attachment addition type", which is stored in the "vehicle cabin deformation data table" in this implementation. The system annotates each record in this table with information such as "Vehicle ID", "coordinate start and end", "specific deformation amount or attachment size", etc.; if the metal plate may be bent, it also records the "bending amplitude or center point".

[0179] Secondly, at the cargo end, the system has a "cargo deformation prediction module". By the "cargo material, chemical properties, temperature sensitivity" marked in the order information, and combined with environmental prediction data such as the average temperature, humidity, or altitude pressure of the route, it infers the volume expansion or deformation that the cargo may generate during transportation. If the cargo is a chemical barrel, the module will retrieve the internal "material - expansion rate mapping table" and calculate the coordinate distribution of the "maximum expansion volume" or "local surface expansion" in combination with the route temperature curve.

[0180] For example, if the top of the barrel will bulge 5 cm with the increase in temperature, the module will record this additional bulge as an "expansion coordinate" and store it in the "cargo deformation data table". The system makes the same record for the potential shape changes of each batch of cargo, so that space can be reserved in advance during the subsequent loading stage.

[0181] After the above two stages are completed, the information in the "vehicle cabin deformation data table" and the "cargo deformation data table" will be integrated into the "deformation data table" through the data exchange interface. This table may contain fields:

[0182] VehicleID / OrderID: Indicates which vehicle or which cargo;

[0183] ShapeType: Indicates whether it is "attachment addition", "sheet metal bending", or "cargo expansion";

[0184] CoordRanges: Three-dimensional coordinate ranges, such as x1x2, y1y2, z1~z2;

[0185] Margin: If a safety margin is required, a minimum margin value will be recorded.

[0186] TimeWindow: An optional field if the prediction only occurs during specific time periods or temperatures.

[0187] When the system enters the "multi-grid construction phase" and discretizes the vehicle compartments and cargo (see the relevant description of multi-grid construction above), it will first read all the records in the "deformation data table" and write them into the coordinate layer one by one. For example:

[0188] If the record indicates that vehicle ID = V123 will install accessory "Unit A" at (X:5070, Y:1020, Z:0~30), then in the vehicle compartment grid coordinates, the corresponding voxel unit will be marked as "phantom volume - accessory addition". If Margin = 3cm, a 3 cm range will also be expanded around it to form a "safety zone" unit, which is uniformly marked as "untouchable" or "empty area".

[0189] If the record shows that the cargo ID = G456 has an additional 5 cm height at the top, then when discretizing the grid, the coordinates of the cargo Z: (H - 5)~H + 0 will be expanded or new grid units will be added and marked as "expansion area". Similarly, it will be stored as a phantom volume unit so that the cargo shape is larger during subsequent collision detection.

[0190] After the system finishes writing each deformation record, the coordinate tables of the vehicle compartment grid and the cargo grid will have the corresponding phantom volume units. In this way, when the three-dimensional loading feasibility check enters the collision detection stage (including global screening of the coarse grid layer and fitting degree judgment of the fine grid layer), any overlap of these phantom volume coordinate units will be regarded as "no available space in the subsequent journey" or "risk of impact after expansion", thus triggering the loading infeasible rollback. If no conflicts are detected in the double-layer grid finally, it indicates that even after installing accessories in the vehicle compartment or the cargo expands, it can still be safely loaded.

[0191] In this way, the system first calculates the "possible deformations or accessory coordinates" during the prediction stage, writes them into the "phantom volume coordinate units" during multi-grid discretization, and finally uses these coordinates in the collision detection stage to make spatial conflict judgments. The beneficial effect is to reduce unexpected occupation or conflicts during driving, help the scheduling algorithm output a safer and feasible assignment plan more stably, and also improve the adaptability to complex temperature-sensitive goods or frequent vehicle modifications. Especially in large-scale order scenarios, this process can save the number of subsequent iterative scheduling, improving the overall operation efficiency and safety guarantee.

[0192] As an alternative implementation, the multi-objective scheduling algorithm further includes:

[0193] Based on the predicted deformation or the data of additional accessories, an additional preset penalty value is added to the allocation relationship of assigning this vehicle in the objective function;

[0194] Based on the inflated volume or extended size of the goods, during the scheduling phase, the inflated or extended size is regarded as the reference volume of the goods, and a preset high penalty value is set for the allocation exceeding this reference volume.

[0195] In this implementation, in order to reduce the risk that the three-dimensional loading becomes infeasible due to the deformation of the vehicle compartment or the inflation of the goods volume during future trips, the system further introduces two "penalty value setting" steps inside the multi-objective scheduling algorithm, enabling the algorithm to preferentially exclude or reduce the assignment of those high-risk combinations before the vehicle and the goods enter the real collision detection phase. Specifically, a "vehicle predicted deformation information table" and a "goods inflation information table" are newly added to the scheduling data structure of the system, and several configurable "preset penalty values" are added to the objective function or constraint rules.

[0196] In specific implementation, if a certain vehicle has been determined by the vehicle-side prediction model to deform or install accessories in the next stage of use (for example, a cold chain unit is newly added to the left bulkhead, or local protrusions occur due to metal fatigue), the system will record the "deformation volume (or area)" and its corresponding coordinate interval in the "vehicle predicted deformation information table". When the scheduling algorithm initializes or updates the candidate solution, once it is found that the order assignment attempts to assign the goods to this vehicle, a "vehicle deformation penalty value" is added to this allocation relationship in the objective function, and this value can be calculated based on the deformation volume, the remaining available space of the vehicle, and the business-side priority. For example, the system can adopt:

[0197] Deformation penalty value = α × Predicted deformation volume (or accessory occupied volume)

[0198] Where α is a set of adjustable penalty coefficients. If the deformation of the vehicle is larger or the future accessory occupancy is more obvious, then this vehicle will be more penalized during the algorithm search process, making its assignment relationship with the goods more tend to a lower score in the overall fitness ranking, thereby reducing the chance of being selected during the iteration process. However, if there are indeed no other vehicles with corresponding transport capacity or timeliness advantages, the algorithm can still retain this assignment combination, but its initial fitness is relatively low.

[0199] Furthermore, for the inflated or extended size of the goods side, the estimated inflation rate or the newly added external dimensions of each piece of goods are also recorded in the "goods inflation information table".

[0200] For example, if the volume of a chemical barrel increases by 10% when the temperature is above 30°C, the system will directly update the cargo volume by (1+expansion rate) times to the "base volume after expansion" during dispatch, so that when calculating "size or load constraints" or even freight or energy consumption, this larger size will be used for calculation. If the expansion deformation of a certain cargo does not absolutely prohibit allocation, but it is a high-risk scenario, a "cargo expansion penalty value" can also be added to the objective function for this combination, such as:

[0201] Expansion penalty = β × (expanded volume - original volume)

[0202] β is another adjustable coefficient. If the expansion is higher, it means that the risk of occupying space is greater, and the scheduling algorithm tends to avoid allocating this cargo to vehicles with less space. In actual calculations, this inflation penalty value, together with indicators such as freight, timeliness, and empty driving rate, constitutes the objective function, allowing the algorithm to automatically eliminate those candidate solutions that cause deep conflicts due to cargo inflation during evolution or iterative search.

[0203] At the algorithm operation level, every time the algorithm generates or improves an "order-vehicle" mapping relationship, it queries the "vehicle predicted deformation information table" and the "cargo expansion information table" to determine whether the vehicle is in an "estimated deformation" state and whether the cargo has "volume expansion". If either party meets the requirements, the corresponding penalty value is added to the fitness of the mapping relationship, and feasibility screening or selection / mutation operations are performed. In this way, before the actual execution of the three-dimensional loading collision detection, the system has already eliminated or downgraded these potential high-risk and high-conflict combinations from the candidate solutions step by step through the penalty items, thereby reducing the rollback rate in the subsequent three-dimensional loading stage and improving the overall iteration efficiency and solution quality.

[0204] In this way, the system can learn about possible dynamic changes in the vehicle compartment or cargo end in advance without having to wait until the three-dimensional collision detection stage to make a discovery; on the other hand, the algorithm imposes additional penalties on combinations involving high deformation or high expansion during multiple rounds of iterations, so that the scheduling solution is more in line with the actual needs of future trips at the macro level, greatly reducing the number of repeated rollbacks and resource waste in the entire process. At the same time, for special cases, such as dangerous goods with high expansion rates, multiple penalties can be superimposed in the scheduling objective function to ensure that safety requirements take precedence over ordinary operating indicators. Therefore, this method is particularly suitable for highly dynamic transportation scenarios (such as long-distance cold chain, chemical transportation, or frequent vehicle modifications), and can effectively alleviate the safety and efficiency problems caused by the traditional scheduling-loading split.

[0205] Based on the same inventive concept, an embodiment of the present disclosure also provides a transportation intelligent assignment and scheduling system corresponding to a transportation intelligent assignment and scheduling method. Since the principle of problem-solving of the system in the embodiment of the present disclosure is similar to that of the above-mentioned transportation intelligent assignment and scheduling method in the embodiment of the present disclosure, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be elaborated.

[0206] Referring to Figure 4 As shown, it is a schematic diagram of a transportation intelligent assignment and scheduling system provided by an embodiment of the present application. The system includes:

[0207] A receiving module 10, configured to obtain order information and transport capacity resource information. The order information includes: cargo attributes; the transport capacity resource information includes: vehicle cabin parameters of self-owned transport capacity and simplified load parameters of external transport capacity; the vehicle cabin parameters include: internal images of the vehicle cabin.

[0208] The receiving module 10 is further configured to, for the vehicle cabin parameters of the self-owned transport capacity, process the internal image of the vehicle cabin by using the wavelet transform method to determine the three-dimensional parameters of the abnormal area; based on the three-dimensional parameters of the abnormal area and the vehicle cabin parameters, generate accurate vehicle cabin parameters of the self-owned transport capacity.

[0209] A scheduling module 20, configured to perform vehicle and route allocation on the preprocessed order set by using a multi-objective scheduling algorithm based on the order information, the accurate vehicle cabin parameters of the self-owned transport capacity, and the simplified load parameters of the external transport capacity, and generate a candidate scheduling plan.

[0210] A detection module 30, configured to perform three-dimensional loading feasibility verification on the candidate scheduling plan by using a preset multi-grid construction module to generate a feasibility verification result; in response to the failure of the feasibility verification, fallback to the multi-objective scheduling algorithm for scheduling plan correction.

[0211] An output module 40, configured to output a final scheduling plan in response to all passing of the feasibility verification.

[0212] The three-dimensional loading feasibility verification includes: determining a predicted deformation for the cargo and the vehicle cabin based on the order information and the transport capacity resource information, and mapping the predicted deformation into a ghost volume in the multi-grid construction module to be marked in the coordinate units corresponding to different grid layers.

[0213] In response to no overlap between the overall size of the cargo and the coordinate units of the ghost volume, compare the fitting degree between the local surface of the cargo and the local units of the ghost volume at the grid layer. In response to the existence of a fitting conflict, mark the loading as infeasible and fallback to the multi-objective scheduling algorithm. In response to the non-existence of a fitting conflict, output that the feasibility verification passes.

[0214] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in the present invention can be implemented by electronic hardware, or by a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

Claims

1. A transportation intelligent assignment and scheduling method, characterized in that, Including: Obtain order information and transportation capacity resource information, where the order information includes: cargo attributes; the transportation capacity resource information includes: vehicle cabin parameters of self-owned transportation capacity and simplified load parameters of external transportation capacity; the vehicle cabin parameters include: internal images of the vehicle cabin; For the vehicle cabin parameters of the self-owned transportation capacity, use the wavelet transform method to process the internal images of the vehicle cabin to determine the three-dimensional parameters of the abnormal area; based on the three-dimensional parameters of the abnormal area and the vehicle cabin parameters, generate accurate vehicle cabin parameters of the self-owned transportation capacity; Based on the order information, the accurate vehicle cabin parameters of the self-owned transportation capacity, and the simplified load parameters of the external transportation capacity, use a multi-objective scheduling algorithm to allocate vehicles and routes for the preprocessed order set to generate a candidate scheduling plan; Use a preset multi-grid construction module to perform three-dimensional loading feasibility verification on the candidate scheduling plan to generate a feasibility verification result; in response to the failure of the feasibility verification, roll back to the multi-objective scheduling algorithm for scheduling plan correction; In response to all passing the feasibility verification, output the final scheduling plan; The three-dimensional loading feasibility verification includes: based on the order information and the transportation capacity resource information, determine the predicted deformation for the cargo and the vehicle cabin, and in the multi-grid construction module, map the predicted deformation into a ghost volume to be marked in the coordinate units corresponding to different grid layers; Wherein, the ghost volume is a virtual three-dimensional volume constructed with grid coordinate units as the smallest discrete unit in the multi-grid construction module based on the predicted deformation, and is used to simulate the extended size and / or expanded volume generated by the cargo and the vehicle cabin due to the predicted deformation; In response to no overlap between the overall size of the cargo and the coordinate units of the ghost volume, compare the fitting degree between the local surface of the cargo and the local units of the ghost volume at the grid layer. In response to the existence of fitting conflicts, mark the loading as infeasible and roll back to the multi-objective scheduling algorithm. In response to the non-existence of fitting conflicts, output that the feasibility verification passes.

2. The method according to claim 1, wherein The internal images of the vehicle cabin include internal images of key areas of the vehicle cabin; the key areas include: the walls, floor, corners, and door frames of the vehicle cabin; The generation of accurate vehicle cabin parameters of the self-owned transportation capacity includes: Perform two-dimensional discrete wavelet transform on the internal images of the vehicle cabin for 2 to 3 layers to decompose into low-frequency subbands and high-frequency subbands; Wherein, the low-frequency subband is used to characterize the overall structure and background information of the internal images of the vehicle cabin; the high-frequency subband is used to characterize the edges, textures, and detail information of the internal images of the vehicle cabin; Calculate the amplitude of the wavelet coefficients in the high-frequency subband and apply adaptive threshold processing to obtain abnormal features; Use inverse wavelet transform to map the abnormal features back to the original image space, apply morphological closing operation to merge adjacent abnormal points and remove isolated noise points to obtain abnormal area positioning information; Use a pre-trained deep learning model to predict the depth information of the abnormal area from a single image to generate a depth map, and combine the camera internal parameters to convert the depth map into three-dimensional parameters of the abnormal area; Calculate the volume of the vehicle cabin space occupied by the abnormal area based on the three-dimensional parameters of the abnormal area, and deduct this volume from the original design dimensions of the vehicle cabin to generate accurate vehicle cabin parameters for available transport capacity.

3. The method according to claim 2, characterized in that, The cargo attributes include: three-dimensional shape data of the ordered goods; performing three-dimensional loading feasibility verification on the candidate scheduling plan includes: Obtain the three-dimensional shape data of the vehicle cabin based on the accurate vehicle cabin parameters for available transport capacity. Among them, the three-dimensional shape data includes: the overall cabin wall structure and sub-cabin basic information characterized by the low-frequency sub-band, and the three-dimensional parameters of the abnormal area obtained by inverse transformation and merging according to the high-frequency sub-band. Input the three-dimensional shape data of the vehicle cabin and the three-dimensional shape data of the ordered goods into the multi-grid construction module respectively to generate hierarchical discrete expressions of the coarse grid layer and the fine grid layer in sequence. In the coarse grid layer, perform global collision screening on the vehicle cabin and the goods based on the voxels of the preset unit. In response to detecting conflict information between the goods and the cabin wall or abnormal area, mark the loading as infeasible and roll back to the multi-objective scheduling algorithm. In response to not detecting conflict information, compare the local concave and convex data of the goods surface and the vehicle cabin one by one in the fine grid layer for stacking simulation. In response to finding any cell collision in the fine grid layer, mark the loading as infeasible and roll back to the multi-objective scheduling algorithm. In response to not finding any cell collision in the fine grid layer, output that the feasibility verification passes.

4. The method according to claim 3, characterized in that, Performing three-dimensional loading feasibility verification on the candidate scheduling plan further includes: In response to data on predicted deformation or attachment addition of the vehicle cabin or goods, map the predicted volume or deformation area into a ghost volume during multi-grid construction and mark it in the corresponding coordinate cells of the coarse grid layer and the fine grid layer. In the global collision screening, in response to the overall size of the goods overlapping with the coordinate cells of the ghost volume, it is determined that there is no available space after subsequent deformation occurs, mark the loading as infeasible and roll back to the multi-objective scheduling algorithm. In response to the overall size of the goods not overlapping with the coordinate cells of the ghost volume, further compare the local surface of the goods with the local cells of the ghost volume in the fine grid layer for fitting degree. In response to the existence of fitting conflicts, mark the loading as infeasible and roll back to the multi-objective scheduling algorithm. In response to the non-existence of fitting conflicts, output that the feasibility verification passes.

5. The method according to claim 4, wherein Performing three-dimensional loading feasibility verification on the candidate scheduling plan further includes: In the coarse grid layer, perform global collision screening on the goods and the vehicle cabin based on the ghost volume. In response to any coarse grid cell detecting voxel overlap in the global collision screening, mark the loading as infeasible and roll back to the multi-objective scheduling algorithm. In response to no coarse grid cell detecting voxel overlap in the global collision screening, compare the local surface of the goods under the ghost volume with the concave and convex areas of the vehicle cabin cell by cell in the fine grid layer. In response to the occurrence of fitting conflicts, mark the loading as infeasible and roll back to the multi-objective scheduling algorithm. In response to no fitting conflicts in the fine grid layer, confirm that the goods can still be loaded smoothly after its expansion, and output that the feasibility verification passes.

6. The method according to claim 5, characterized in that, It also includes: At the vehicle cabin end, a preset vehicle cabin deformation prediction model is periodically called. Based on the vehicle usage duration, the vehicle cabin metal fatigue coefficient, and the historical modification records, time series inference is performed to output the predicted deformation or the data of additional accessories within the next stage of the journey for each free vehicle. The data of the predicted deformation or the additional accessories is marked with an additional accessory identifier or a metal plate bend, and the coordinate range is stored in the deformation data table. At the cargo end, based on the order information, the volume expansion rate or the structural deformation probability information based on the cargo material or chemical properties is extracted. Combining the predicted values of the temperature, humidity, and pressure conditions during the transportation process, the cargo deformation prediction module calculates the possible additional outer contour dimensions of the cargo and records them in the deformation data table in the form of the expanded volume or the extended dimensions. In the multi-grid construction stage, based on the additional accessory identifier of the vehicle cabin, the metal plate bend of the vehicle cabin, the expanded volume of the cargo, and the extended dimensions of the cargo registered in the deformation data table, the predicted deformation area or the extended form is written into the corresponding ghost volume coordinate unit respectively.

7. The method according to claim 6, wherein The simplified load parameters include: the maximum loadable weight, the maximum loadable volume, and the transportation qualification information; the multi-objective scheduling algorithm includes: Establish an objective function in the multi-objective scheduling algorithm; among them, the optimization indicators in the objective function include: freight cost, transportation timeliness, vehicle empty running rate, and compliance requirements for special goods. Set hard constraint rules in the multi-objective scheduling algorithm; the hard constraint rules include: In response to the cargo attribute including dangerous goods, it can only be assigned to the self-owned transport capacity with the corresponding protection configuration or the external transport capacity with the dangerous goods transportation qualification; in response to the cargo requiring cold chain, it can only be assigned to the transport capacity with the corresponding refrigeration function. Set soft constraint rules in the multi-objective scheduling algorithm; the soft constraint rules include: In response to the size or weight of the order cargo exceeding the accurate vehicle cabin parameters of the self-owned transport capacity or the simplified load parameters of the external transport capacity, add a penalty value to the current assignment relationship in the objective function to reduce the fitness of this assignment relationship in the iterative search.

8. The method according to claim 7, characterized in that, The generation of the candidate scheduling plan includes: Perform iterative solution on the preprocessed order set to generate multiple candidate solutions, and each candidate solution includes the assignment relationship to the self-owned transport capacity and / or the external transport capacity, the driving route, and the estimated arrival time. In response to detecting an allocation combination that does not meet the constraint rules during the scheduling iteration process, eliminate or mutate this candidate solution in the algorithm iteration until a candidate scheduling plan that meets the requirements of the objective function is output.

9. The method according to claim 8, wherein The multi-objective scheduling algorithm further includes: Based on the data of the predicted deformation or the additional accessories, add an additional preset penalty value to the assignment relationship of assigning this vehicle in the objective function. Based on the expanded volume or the extended dimensions of the cargo, during the scheduling stage, regard the expanded or extended dimensions as the reference volume of the cargo, and set a preset high penalty value for the allocation exceeding this reference volume.

10. A transportation intelligent assignment and scheduling system, characterized in that Include: A receiving module, configured to obtain order information and transportation capacity resource information, where the order information includes: cargo attributes; and the transportation capacity resource information includes: vehicle cabin parameters of self-owned transportation capacity and simplified load parameters of external transportation capacity; and the vehicle cabin parameters include: internal images of the vehicle cabin. The receiving module is further configured to, for the vehicle cabin parameters of the self-owned transportation capacity, process the internal images of the vehicle cabin by using a wavelet transform method to determine three-dimensional parameters of abnormal areas; and generate accurate vehicle cabin parameters of the self-owned transportation capacity based on the three-dimensional parameters of the abnormal areas and the vehicle cabin parameters. A scheduling module, configured to perform vehicle and route allocation on a preprocessed order set by using a multi-objective scheduling algorithm based on the order information, the accurate vehicle cabin parameters of the self-owned transportation capacity, and the simplified load parameters of the external transportation capacity, and generate a candidate scheduling plan. A detection module, configured to perform three-dimensional loading feasibility verification on the candidate scheduling plan by using a preset multi-grid construction module, and generate a feasibility verification result; and in response to the failure of the feasibility verification, fallback to the multi-objective scheduling algorithm to correct the scheduling plan. An output module, configured to output a final scheduling plan in response to the passing of all the feasibility verifications. The three-dimensional loading feasibility verification includes: determining a predicted deformation for the cargo and the vehicle cabin based on the order information and the transportation capacity resource information, and mapping the predicted deformation into a ghost volume in the multi-grid construction module to be marked in coordinate units corresponding to different grid layers. Wherein, the ghost volume is a virtual three-dimensional volume constructed with grid coordinate units as the smallest discrete unit in the multi-grid construction module based on the predicted deformation, and is used to simulate the extended size and / or expanded volume generated by the cargo and the vehicle cabin due to the predicted deformation. In response to no overlap between the overall size of the cargo and the coordinate units of the ghost volume, compare the fitting degree between the local surface of the cargo and the local units of the ghost volume at the grid layer. In response to the existence of a fitting conflict, mark the loading as infeasible and fallback to the multi-objective scheduling algorithm. In response to the non-existence of a fitting conflict, output that the feasibility verification has passed.

Images