Cooperative scheduling method and system for material supply and resource recovery

Abstract

The invention discloses a material supply and resource recovery collaborative scheduling method and system, and belongs to the technical field of logistics scheduling, and the method comprises the steps: obtaining a material supply instruction and a garbage recovery instruction, carrying out the image recognition of a building garbage picture, obtaining the garbage attribute, and generating a demand instruction library; the method comprises the following steps: establishing a feature knowledge base, setting a spatial clustering method, carrying out spatial clustering and time window screening, carrying out geographic coordinate analysis and time sequence sorting on material demand points and recovery points, generating an initial task pool, setting a loading matching method, and generating a material supply and recovery loading schematic diagram according to building material and garbage attributes; a circulation path is planned, a path planning method is set, an ant colony algorithm is used for initial path planning, a monitoring feedback method is set, data are returned in real time, the running state of the vehicle is monitored, and an alarm is given out immediately once abnormity is found; and the sorted and regenerated aggregate data is synchronized to a building material database, a resource feedback method is set, a reverse transportation task is identified, and a material supply and resource recovery closed loop is formed.

Description

TECHNICAL FIELD

[0001] The application relates to a material supply and resource recycling collaborative scheduling method and system, and belongs to the technical field of logistics scheduling. BACKGROUND

[0002] In the field of industrial logistics and resource recycling, the collaborative scheduling of material supply and resource recycling is a core link for realizing low-carbon production and sustainable development. The traditional scheduling mode relies on manual experience or rule-based algorithms, and has problems such as low supply-demand matching efficiency, high vehicle empty running rate, and insufficient resource recycling rate, which cannot meet the requirements of logistics energy efficiency and circular economy under the "double carbon" target. In recent years, with the development of the Internet of Things and AI algorithms, intelligent scheduling systems have initially realized the linkage planning of supply tasks and recycling tasks by integrating order data, vehicle status and path information, and have improved the resource allocation efficiency to a certain extent.

[0003] However, due to the lack of deep multi-source data fusion and dynamic collaborative mechanism, the existing technology still faces multiple challenges; the material supply data and recycling demand lack a unified digital collection and fusion system, and only rely on simple proximity matching based on geographic coordinates, ignoring the physical properties and real-time state of goods, which leads to low coupling of "supply-recycling" tasks, and often causes vehicle load imbalance or insufficient endurance, directly causing transportation safety risks and waste of resource carrying capacity, at the same time, the existing dynamic scheduling algorithm does not deeply integrate real-time traffic conditions, vehicle energy consumption model and recycling site capacity data, and has long planning period, weak anti-interference ability, and lacks load balancing and empty running energy consumption optimization, which is difficult to adapt to the real-time changing transportation environment, in addition, the material supply and recycling links operate independently, and lack of full-link collaborative strategy. After the construction waste at the recycling site is sorted and regenerated into aggregate, it cannot be fed back to the building material supply end in real time, resulting in an increase in repeated procurement rate of the factory, and the supply vehicles often return empty after completing the delivery, without utilizing the return path to perform adjacent recycling tasks, and the recycling rate of renewable resources is low, far from achieving the circular economy goal of building material points, demand points, recycling points, recycling stations and building material points, and the efficient recycling of renewable resources is hindered. SUMMARY

[0004] In view of the deficiencies of the prior art, the purpose of the present application is to provide a material supply and resource recycling collaborative scheduling method and system, which collects instructions and images through a user terminal, performs spatial clustering, time sequence filtering and multi-objective planning by a cloud server, dynamically adjusts the path and loading scheme in combination with real-time feedback, and solves the problems of data fragmentation, insufficient collaboration and resource waste by using a resource feedback mechanism to build a closed loop.

[0005] To achieve the above purpose, the application provides the following technical scheme:

[0006] The material supply and resource recycling collaborative scheduling method comprises:

[0007] Obtain the supply instruction and the garbage recycling instruction, and perform image recognition on the construction waste photos to obtain the garbage attribute, and generate a demand instruction library;

[0008] Establish a feature knowledge base, set a spatial clustering method, perform spatial clustering and time window screening, and perform geographic coordinate analysis and time sequence sorting on the supply points and the recycling points to generate an initial task pool, set a loading matching method, and generate a supply and recycling loading schematic according to the building material and the garbage attribute;

[0009] Plan a circulation path, set a path planning method, perform initial path planning by using an ant colony algorithm, and set a monitoring feedback method to return data in real time and monitor the vehicle operation state, and immediately issue an alarm once an abnormality is found;

[0010] Synchronize the sorting regenerated aggregate data to the building material database, set a resource feedback method, identify reverse transportation tasks, and form a material supply and resource recycling closed loop.

[0011] Specifically, the specific steps of obtaining the garbage attribute include:

[0012] Use a terminal device to take pictures of the construction waste from different angles, and place a reference object beside the construction waste when taking the pictures;

[0013] Perform standardization processing on the taken images to unify the format;

[0014] Replace the last fully connected layer of the ResNet-50 model with a classifier to generate a garbage type recognition model, and obtain a large number of annotated historical construction waste images, and train the garbage type recognition model;

[0015] Use the garbage type recognition model to process the standardized construction waste pictures, output the garbage category, and perform pollution labeling on the images based on the pollution level classification, wherein the pollution level includes no pollution, low pollution, and high pollution, and the pollution labeling includes no pollution and existing pollution;

[0016] Detect the garbage area by using a target detection model, calculate the pixel-to-actual size conversion coefficient in combination with the reference object, and estimate the volume of the construction waste;

[0017] Use the ResNet-50 model as a backbone network, access a multi-label classifier, generate a pollution identification model, and train the pollution identification model based on historical images;

[0018] Use the pollution identification model to process the images with existing pollution, identify the harmful substance label, generate a harmful label set, and determine the pollution level;

[0019] Obtain the garbage area pixel-level mask through semantic segmentation, calculate the area compactness, and determine the stacking mode of the garbage;

[0020] Integrate the garbage type, volume, pollution level, and stacking mode, and save them to the corresponding garbage recycling instruction.

[0021] Specifically, the spatial clustering method comprises:

[0022] Obtain the longitude and latitude coordinates of the demand points and recycling points from the demand instruction library, convert them into a plane coordinate system through Gauss projection, calculate the actual distance between the demand points and recycling points, and generate a distance matrix;

[0023] Set the neighborhood radius and the minimum number of samples, traverse the distance matrix, and identify high-density areas using a dynamic grid clustering algorithm;

[0024] Combine the garbage volume and the weight of building materials to filter and merge nearby point pairs, and output a set of instruction pairs;

[0025] Mark instructions that do not form a clustering cluster as pending, and automatically enter the regional candidate task pool;

[0026] Set a candidate scanning interval to periodically scan the regional candidate task pool, and trigger manual verification when a region in the regional candidate task pool is always a low-density region.

[0027] Specifically, the spatial clustering method further comprises:

[0028] Traverse the supply instruction in the set of instruction pairs, calculate the expected arrival time, and sort it in ascending order to generate a supply list;

[0029] Traverse the garbage recycling instruction in the set of instruction pairs, calculate the loading time based on the stacking mode of the garbage, and sort it in ascending order to generate a recycling list;

[0030] Based on the expected arrival time, add the standard time of unloading operation to calculate the supply completion time, and according to the distance between the recycling points and demand points in the distance matrix, calculate the empty running time of the vehicle from the demand point to the recycling point, to determine the recycling feasible time T R-start ;

[0031] Obtain the time window [T R-begin , T R-end ] in the recycling instruction, and determine whether T R-start falls within the recycling time window; wherein T R-begin and T R-end are the earliest time allowed for the start of the recycling task and the deadline for completion, respectively

[0032] If T R-begin ≤ TR-start ≤T R-end , the instruction is incorporated into the time window matching set; if T R-start <T R-begin , the waiting time is calculated, and is marked as schedulable, and is also incorporated into the time window matching set; if T R-start >T R-end , enters the regional candidate task pool to wait for subsequent aggregation; and outputs the task chain set;

[0033] Based on the feature knowledge base, the weight of the compressed recycled garbage is calculated, and based on the preset load verification condition, the load feasibility flag F w ∈{0, 1} is obtained.

[0034] The total energy consumption mileage D is calculated, and based on the preset endurance verification condition, the total energy consumption mileage is compared with the remaining endurance of the vehicle to obtain the endurance feasibility flag F e ∈{0, 1}.

[0035] The priority formula is constructed by comprehensively considering the supply urgency, recycling timeliness sensitivity and distance penalty, and the priority of each task chain set is calculated, and the task chain meeting the core screening condition is saved to the constructed initial task pool.

[0036] Specifically, the loading matching method comprises:

[0037] According to the weight and density of the building materials, the volume of the building materials is calculated, and a preliminary loading strategy is generated according to the preset partition rule;

[0038] After loading, the center of gravity coordinates are calculated according to the weight and placement position of each part of the building materials, and are compared with the preset safety interval, and once the safety interval is exceeded, the loading layout of the building materials is automatically adjusted;

[0039] A loading schematic diagram of the supply stage is generated and pushed through the vehicle terminal;

[0040] The load and volume data are returned in real time during the loading process, and the recovery and loading scheme is automatically adjusted when the supply weight deviation exceeds the proportion;

[0041] In the recovery stage, if the actual volume of the garbage deviates from the volume estimated by image recognition beyond the range, the sorting equipment is triggered to adjust;

[0042] A plurality of loading schemes are generated by using a genetic algorithm, and the optimal scheme is selected after comprehensive evaluation of the space utilization rate and the stability of the center of gravity, and an execution scheme is generated.

[0043] Specifically, the path planning method comprises:

[0044] The node coordinate information is extracted from the task pool, and based on the actual distance between nodes and the average speed of the vehicle, the real-time passing time between nodes is calculated to form a passing time matrix;

[0045] Combined with the supply weight, recycled compaction weight and pollution level obtained from the loading scheme, multi-dimensional constraint conditions are established.

[0046] The parameters of the ant colony algorithm are initialized and set, the total passing time, total energy consumption, load balance penalty and empty distance penalty are considered comprehensively, a multi-objective function is constructed, the ant colony algorithm is used to optimize the objective function, and an initial path scheme is generated.

[0047] Specifically, the path planning method further comprises:

[0048] When the recycled aggregate inventory of the recycling station reaches the standard, a reverse supply node returning to the building material point from the recycling station is added in the path;

[0049] An update interval is set, real-time traffic data is accessed based on the update interval, the congestion coefficient of each road section is calculated, and once the congestion coefficient is lower than the set threshold, the Dijkstra algorithm is triggered to re-search the congested road section to generate a locally optimized path.

[0050] The weight coefficients of the objective function are adjusted to generate multiple candidate path schemes with different emphases;

[0051] After the candidate path schemes are comprehensively scored by using the analytic hierarchy process, the optimal path is selected and output.

[0052] Specifically, the monitoring feedback method comprises:

[0053] Vehicle state, task execution and external environment data are collected by a vehicle terminal to construct a three-dimensional data system;

[0054] Multi-dimensional anomaly detection is performed to calculate load deviation, consumption rate and time deviation to determine the anomaly type;

[0055] According to the anomaly type, a hierarchical response is performed, an alarm is sent and the vehicle operation is locked in an emergency, the path is dynamically adjusted in a general anomaly, the anomaly is updated and feedback is optimized;

[0056] When the abnormal situation causes the task to be unable to be performed according to the original plan, the task chain is dynamically reconstructed.

[0057] If multiple time-sensitive abnormal tasks occur in the same area, a regional task merging strategy is triggered.

[0058] Specifically, the resource feedback method comprises:

[0059] Real-time record the weight of recycled garbage, combined with garbage type, volume and compaction coefficient, calculate the theoretical output of recycled aggregate;

[0060] Quality detection of recycled aggregate to determine the quality grade;

[0061] Synchronize actual output, quality grade and storage location to factory building material database;

[0062] Set the safety stock threshold of factory building materials and calculate the availability of recycled aggregate;

[0063] When the trigger condition is met, generate a reverse transportation order from the recycling station to the building material point, and preferentially schedule an empty vehicle to execute;

[0064] Include the reverse transportation node in the original path and optimize it, plan the loading amount and location of recycled aggregate according to the remaining carrying capacity;

[0065] Compare the quality data after transportation, feedback and adjust the sorting equipment parameters, calculate the cost and dynamically adjust the counterproductive strategy.

[0066] The collaborative scheduling system of material supply and resource recycling includes a user terminal, a cloud server, a vehicle-mounted terminal and a zero-carbon factory platform;

[0067] The user terminal is used to receive user inputted supply and garbage recycling instructions, take photos of construction waste and perform standardized preprocessing;

[0068] The vehicle-mounted terminal is used to collect vehicle operation data and loading and unloading site information in real time, receive and execute cloud server instructions, and respond to abnormal situations in stages;

[0069] The zero-carbon factory platform is used to synchronize recycled aggregate inventory and quality data, set inventory thresholds, and adjust resource counterproductive strategies in coordination with production plans;

[0070] The cloud server includes a data acquisition module, a task matching module, a planning module and an updating module;

[0071] The data acquisition module is used to integrate user instruction data, vehicle-mounted terminal real-time data and external environment data, perform image recognition on construction waste photos and extract garbage attributes, and maintain a feature knowledge base;

[0072] The task matching module is used to filter and prioritize supply and recycling instructions through spatial clustering, time window filtering and feasibility evaluation, and generate a task pool;

[0073] The planning module is used to plan paths and loading based on task pool information, trigger reverse supply from the recycling station to the building material point and optimize paths and loading;

[0074] The update module is used for processing abnormal data, updating feature knowledge base parameters, verifying and optimizing resource feedback strategies.

[0075] Advantages of the present application:

[0076] The user terminal realizes standardized collection of instructions and images, the cloud server integrates multidimensional data such as geography, time and cargo characteristics, constructs a feature knowledge base and performs spatial clustering and time sequence screening, generates an initial task pool with high coordination, and solves the problem of low task coupling caused by data fragmentation; path planning integrates multiple objective functions such as load balancing and empty driving penalty, and dynamically adjusts in combination with real-time road conditions, thereby improving the anti-interference ability of planning and reducing transportation risks such as load imbalance and insufficient endurance. The loading matching considers characteristics such as pollution level and stacking form, thereby ensuring transportation safety; the resource feedback mechanism synchronously regenerates aggregate data in real time, triggers a reverse feeding task, forms a closed-loop control including building material supply, recycling and regeneration, and feedback supply, reduces waste such as factory raw material procurement and vehicle empty driving, improves the utilization rate of recycled resources, realizes full-link coordination of material supply and resource recycling, and promotes the achievement of the goal of circular economy. BRIEF DESCRIPTION OF DRAWINGS

[0077] Figure 1 The figure is a structural diagram of the collaborative scheduling method of material supply and resource recycling.

[0078] Figure 2 The figure is a flowchart of the spatial clustering method of the present application.

[0079] Figure 3 The figure is a flowchart of the loading matching method of the present application.

[0080] Figure 4 The figure is a flowchart of the resource feedback method of the present application.

[0081] Figure 5 The figure is a flowchart of the collaborative scheduling system of material supply and resource recycling. DETAILED DESCRIPTION

[0082] The technical solutions of the present application will be described in detail below with reference to the drawings and specific embodiments. It should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solutions of the present application, and are not limitations of the technical solutions of the present application. In the case of no conflict, the technical features in the embodiments and the embodiments can be combined with each other.

[0083] Embodiment 1

[0084] Reference Figures 1 to 4 The present embodiment introduces a collaborative scheduling method of material supply and resource recycling, as shown in Figure 1 The figure is a structural diagram of the collaborative scheduling method of material supply and resource recycling, and the method comprises:

[0085] The user operates the terminal device to upload the supply instruction and the garbage recycling instruction. In terms of the supply instruction, the type, quantity, delivery time requirement and demand point coordinate of the building material are input, and the construction site environment is photographed for subsequent pre-judgment of the unloading space. In terms of the recycling instruction, the building waste photo is uploaded, and the recycling time requirement and recycling point coordinate are marked. The building waste photo is automatically subjected to image recognition to obtain the type, volume, pollution level and stacking form of the waste so as to judge the loading difficulty and generate a demand instruction library;

[0086] A feature knowledge base is established to associate the building material density and the compaction coefficient of different types of building waste for subsequent load planning. A spatial clustering method is set to analyze the geographic coordinates of the demand points and recycling points in the demand instruction library. In order to optimize the task execution sequence, the instructions of similar regions are clustered according to the pre-set spatial range standard in combination with the map information. The supply instruction and the garbage recycling instruction are paired. In the time dimension, the clustered instructions are filtered and sorted according to the time window. The supply instruction and the garbage recycling instruction are sorted according to the time sequence. Between the supply instruction and the garbage recycling instruction, the predicted delivery arrival time of the supply instruction plus a specific time interval is calculated and compared with the recycling time of the garbage recycling instruction. The instruction combinations meeting the time window requirement are selected according to the sequence to ensure the orderly connection of the tasks and avoid task omission. An initial task pool is generated through geographic and time sequence double filtering;

[0087] Before task execution, a loading schematic diagram is generated for the supply and garbage recycling links in the initial task pool. A loading matching method is set. In the supply stage, the loading position of the building material in the vehicle is planned according to the type and quantity of the building material, the carrying capacity of the vehicle and the cargo compartment structure, and the pre-set building material loading principle to generate a supply loading schematic diagram. At the same time, the garbage recycling loading scheme is planned in advance in consideration of the subsequent garbage recycling task. According to the feature knowledge base, the garbage amount that can be carried by each vehicle is calculated in combination with the number of vehicles and the carrying capacity of each vehicle during the supply. For garbage of different pollution levels, a classified loading strategy is adopted, such as arranging garbage of high pollution level in a special sealed area to avoid pollution to other goods. Finally, a garbage recycling loading schematic diagram is generated.

[0088] A circular path is planned for the building material point, demand point, recycling point, recycling station and building material point. According to the type and volume of the garbage in the instruction, a sorting device is matched. A path planning method is set. An ant colony algorithm is used for initial path planning. A load balance penalty term and an empty running distance penalty term are integrated into the objective function to simultaneously optimize the transportation time efficiency, energy cost and load balance degree.

[0089] The vehicle performs tasks according to the planned path, sets a monitoring feedback method, and during the feeding process, the vehicle-mounted camera shoots the unloading site to identify the remaining space, providing a reference for possible secondary feeding, and during the recycling process, the image recognition is used again to verify the type and volume of the garbage, and compared with the estimated value, if the error exceeds the specified range, feedback is given to the scheduling platform in time, and during the driving process, the position, speed, load, and battery level are returned to the scheduling platform in real time, and the scheduling platform monitors the vehicle running state according to these real-time data, and once an abnormality is found, such as the vehicle deviating from the planned path, the load changing abnormally, or the battery level being too low, an alarm is sent in time and corresponding measures are taken, such as re-planning the path or dispatching other vehicles for assistance.

[0090] After the vehicle transports the recycled garbage to the recycling station, the weighing system of the recycling station records the actual recycling weight and cross- verifies the data uploaded by the vehicle, and at the same time, the recycling station starts the corresponding sorting process according to the type of garbage, sorts the construction waste for regeneration into aggregate, and synchronizes the yield and type of the regenerated aggregate to the building material database of the zero-carbon factory in real time, sets a resource feedback method, and according to the building material inventory and regenerated aggregate inventory in the factory database, when the regenerated aggregate meets certain conditions, automatically triggers the transportation task from the recycling station to the building material point, realizes the feedback of recycled resources to the building material supply end, reduces the factory's procurement of raw materials, improves the utilization rate of regenerated resources, and forms a closed loop of material supply and resource recycling.

[0091] Specifically, the specific steps of identifying the construction waste photos include:

[0092] The user uses a terminal device, such as a mobile phone or a camera, to take photos of the construction waste. To ensure the accuracy of subsequent processing, at least 3 photos from different angles, including front view, side view, and overhead view, are taken. When taking photos, a reference object, such as a standard-sized brick, is placed next to the construction waste to facilitate subsequent size calibration.

[0093] The photographed images are subjected to image standardization processing, including scaling and grayscale normalization. The size of the image is uniformly scaled to the input size required by the residual network model, and then the image is converted to RGB three-channel floating-point data, and the pixel value is normalized to a specific interval, eliminating the image differences caused by different shooting devices, and providing uniform format image data for subsequent model processing.

[0094] The last fully connected layer of the ResNet-50 model is replaced with a classifier to generate a garbage type identification model to identify different types of garbage, including concrete, steel, glass, wood, plastic, brick slag, gypsum board, mixed garbage, hazardous garbage (containing paint / batteries), and others.

[0095] A large number of labeled historical building waste images are acquired, and cross-entropy is defined as the loss function. The waste recognition model is trained using the Adam optimizer. The standardized historical images are input into the waste recognition model, and deep features are extracted through 16 residual blocks. The output class probability vector is obtained through global average pooling and fully connected layers. The class with the highest probability is taken as the recognition result. If the highest probability does not reach the preset probability threshold, it is marked as mixed waste and manual review is triggered to determine the waste type.

[0096] The trained waste type recognition model is used to identify user-uploaded images, output specific waste categories, and set pollution level classifications, including no pollution, low pollution, and high pollution. Images that are not classified as hazardous waste are marked as no pollution, while images that are classified as hazardous waste are marked as polluted.

[0097] Using an object detection model such as YOLOv5, garbage regions in the image are detected, and the minimum bounding box coordinates (x1, y1, x2, y2) of the garbage region are obtained. By using a reference object placed next to the garbage, the conversion factor k between the pixel size and the actual size is calculated. con The expression is as follows:

[0098]

[0099] In the formula, L is the actual side length of the reference object, W is the pixel width of the reference object in the image, and (x1,y1) and (x2,y2) are the coordinates of two non-adjacent vertices of the rectangle.

[0100] The 3D volume of construction waste in the image is estimated. The height of the waste is obtained through user-supplied information or a height prediction model. Combined with conversion factors, the actual length and width of the construction waste are calculated using multi-view bounding boxes, thereby calculating the volume of the construction waste. The expression is as follows:

[0101] l = k con ×(x2-x1)

[0102] w = k con ×(y2-y1)

[0103] V = l × w × h

[0104] In the formula, l, w, and h are the actual length, width, and height of the construction waste, V is the estimated volume of the construction waste, and the height prediction model is constructed by a convolutional neural network, trained with a large number of historical images, inputting image features to output the actual height;

[0105] Using the ResNet-50 model as the backbone network, a multi-label classifier is connected to identify specific hazardous substance categories, such as paint, batteries, and chemicals, and defined as a pollution identification model. A large amount of labeled construction waste image data is collected, which covers various scenarios that may contain hazardous substances. The cross-entropy loss function is used as the optimization objective, and the Adam optimizer is used to train the model. The model parameters are continuously adjusted to improve the accuracy of the model in judging whether there are hazardous substances and identifying specific hazardous substances.

[0106] The pollution level of the images is determined, and the trained pollution recognition model is used to identify the specific harmful substance labels in the images. All the identified harmful labels are then combined into a harmful label set.

[0107] If the set of hazardous labels consists entirely of low-risk hazardous labels, the pollution level is determined to be low-pollution, and the image is labeled as low-pollution; if the set of hazardous labels contains any high-risk hazardous label, the pollution level is determined to be high-pollution, and the image is labeled as high-pollution; low-risk and high-risk hazardous labels are defined by a pre-set hazardous label library, such as paint fragments and plastic film being labeled as low-risk hazardous, and batteries and chemical coatings being labeled as high-risk hazardous;

[0108] Simultaneously, based on the determined pollution level, risk level mapping is performed to generate corresponding loading strategies. When the construction waste is unpolluted, it can be loaded using ordinary trucks without special treatment. When the pollution level is low, due to the presence of a small amount of low-polluting substances, a leak-proof padding layer needs to be laid inside the truck to prevent leakage or spread of these substances during transportation, providing a certain degree of protection. When the pollution level is high, given the significant hazards of the high-polluting substances, a specialized sealed compartment is used for loading to ensure that hazardous substances do not leak into the external environment. The pollution level and hazardous label set are output to provide an accurate basis for subsequent transportation scheduling and loading operations.

[0109] A U-shaped network model is used to perform semantic segmentation of the garbage region, obtain the pixel-level mask of the garbage region, and calculate the region compactness C to determine the garbage stacking pattern; the expression is as follows:

[0110]

[0111] In the formula, S is the area of ​​the garbage area, and L S The perimeter of the garbage area;

[0112] Set the stacking thresholds as C1 and C2, where C1 > C2, and determine the stacking form of the garbage; if C ≥ C1, it is determined as regular stacking; if C2 < C < C1, it is determined as loose stacking; if C ≤ C2, it is determined as chaotic stacking;

[0113] Integrate and output the recognition results such as garbage type, volume, pollution level, and stacking form, and save them to the corresponding garbage recycling instructions.

[0114] As Figure 2 shown, it is the flowchart of the spatial clustering method of the present invention. Specifically, the spatial clustering method includes:

[0115] Batch obtain the longitude and latitude coordinate data of the material demand points and recycling points from the demand instruction library. Since the longitude and latitude coordinates are based on the earth's spherical coordinate system, which is not conducive to direct distance calculation and spatial analysis, the Gauss projection conversion algorithm is used to uniformly convert all longitude and latitude coordinates to the plane coordinate system. During the conversion process, according to the geographical location of the coordinates, the appropriate projection zone is automatically selected to ensure the accuracy of the converted coordinates. After the conversion is completed, for any material demand point coordinate and any recycling point coordinate, the Euclidean distance formula is used to calculate the actual distance to determine whether the coordinate point is within the set neighborhood radius, so as to identify the high-density area. To improve the calculation efficiency, parallel computing technology is used to simultaneously process the distance calculations of multiple groups of coordinate points, and finally form a complete distance matrix, providing comprehensive and accurate basic data for subsequent clustering analysis;

[0116] Use the dynamic grid clustering algorithm to process the coordinate points, and set the neighborhood radius as R eps and the minimum sample number as N min . Traverse the distance matrix. Taking any point as the center, divide the neighborhood area according to the neighborhood radius, and define the area where the number of points within any neighborhood area exceeds N min as the high-density area, and the points within the high-density area are classified into the same clustering cluster;

[0117] During the clustering process, the garbage volume and building material weight output by image recognition are retrieved in real time. When the garbage volume is less than the preset volume threshold and the building material weight is less than the preset weight threshold, the filtering mechanism is triggered to merge the corresponding points with other adjacent points to avoid uneconomical single vehicle transportation, and an instruction pair set is output, which includes the point pairs within the same cluster and meeting the basic load coordination condition. Among them, the basic load coordination condition is that the weight of the recycled garbage after compaction does not exceed 30% of the remaining load of the vehicle; each pair in the instruction pair set is not a one-to-one fixed pairing. One cluster may contain multiple material demand points and multiple recycling points, forming a many-to-many potential pairing, such as 3 material demand points and 2 recycling points in a residential dense area;

[0118] For coordinate points in low-density areas that have not formed clusters, the corresponding instructions are marked as pending and automatically enter the regional candidate task pool. The candidate scan interval is set, the candidate task pool is scanned, and the regions are divided. The regional candidate task pool is scanned periodically to identify high-density areas. If a certain region is consistently a low-density area within a certain period of time, a manual verification reminder is pushed to confirm whether there are coordinate errors or special requirements, and the task execution instructions are manually configured.

[0119] To rationally arrange the material supply sequence, the estimated arrival time of each material supply instruction in the instruction pair set is calculated, and the material supply instructions are sorted in ascending order according to the estimated arrival time to form an ordered material supply list L. C The expression is as follows:

[0120]

[0121] In the formula, T C-arr T represents the estimated arrival time for delivery. C-dep For the expected trigger time, d P-C The distance between the building materials point and the point where materials are needed. Let t be the average speed of the vehicle. in Standard loading time;

[0122] For each garbage collection instruction in the instruction pair set, the loading time is calculated based on the garbage's stacking pattern and a preset loading time estimation rule. The instructions are then sorted in ascending order according to their loading times, and an ordered collection list L is output. R The loading time estimation rules include: if the garbage is piled up haphazardly, an additional 20 minutes of loading time is required based on experience; if it is piled up loosely, an additional 10 minutes is required; if it is piled up neatly, the standard loading time is used.

[0123] The estimated delivery arrival time of the material supply instruction generated in the single-type instruction sorting is added to the fixed time of the unloading operation to calculate the material supply completion time T. C-end This signifies the completion of the vehicle's work at the material demand point, allowing dispatching to the recycling point to begin. Based on the distances between the recycling point and the material demand point in the distance matrix, the vehicle's idle travel time T from the material demand point to the recycling point is calculated. idle By combining the material supply completion time and the idle time, the earliest time T when the recovery mission can be started is determined. R-start And defined as the feasible time for recycling; the expression is as follows:

[0124] T C-end =T C-arr +T out

[0125]

[0126] TR-start =T C-end +T idle

[0127] In the formula, T out For standard unloading time, d C-R This represents the actual distance from the material-requiring point to the recycling point in the distance matrix;

[0128] Each recycling instruction includes a user-defined time window [T] R-begin ,T R-end ], and determine T R-start Whether it falls within the recycling timeframe; where T R-begin The earliest time that allows a recycling task to begin, such as the start time of waste recycling permitted at a construction site, T R-end Deadlines that must be met for recycling tasks, such as the deadlines for waste collection stipulated by environmental regulations;

[0129] If T R-begin ≤T R-start ≤T R-end If the material supply and recycling instructions form a time-feasible task chain, they are included in the time window matching set; if T R-start <T R-begin If T indicates that the vehicle arrives at the recycling point too early, the vehicle must wait at the recycling point. The waiting time is automatically calculated, and the vehicle is marked as available for scheduling and included in the time window matching set. R-start >T R-end If the vehicle arrives at the recycling point too late, the recycling task is not feasible in terms of time and will enter the waiting task pool to wait for subsequent aggregation.

[0130] Output a set of task chains that match the time window, effectively improving the reliability of task chain timing.

[0131] Based on the waste volume output by image recognition, the system automatically retrieves the density and compaction coefficient of the corresponding waste type from the feature knowledge base, calculates the weight of the compacted recycled waste, and obtains the load-bearing feasibility flag F based on preset load-bearing verification conditions. w ∈{0,1}; where the load verification condition is W C +W R ≤W up ×(1+W th ), W C For the weight of building materials, W R To recover the weight of the compacted waste, W up W is the upper limit of vehicle load. th The load deviation threshold is used to output the load feasibility flag F if the condition is met. w =1, otherwise F w =0;

[0132] The total energy consumption mileage D is calculated, covering the entire journey from the building material point, through the material demand point and recycling point to the recycling station, and back to the building material point. Based on preset range verification conditions, the total energy consumption mileage is compared with the vehicle's remaining range to obtain the range feasibility indicator F. e ∈{0,1}; where the battery life verification condition is D re ≥D×(1+ε), where D re The remaining range of the vehicle is given by ε, which is the safety redundancy coefficient. If the conditions are met, the range feasibility flag F is output. e =1, otherwise F e =0, ensuring that the vehicle will not be interrupted due to insufficient range during mission execution;

[0133] Taking into account the urgency of material supply, sensitivity to recycling timeliness, and distance penalty, a priority formula is constructed; the expression is shown below:

[0134]

[0135] In the formula, L rank For the priority of the task chain, E C To ensure urgency in material supply, for urgent tasks, E C =1.5, for normal tasks, E C =1,E R To ensure timeliness of recycling, when the waste is highly polluting, E R =1.2, otherwise, E R =1,D p As a distance penalty, d th The penalty is the upper limit, and the penalty increases with the distance. δ is the pollution correction factor. When the garbage is polluted, δ = 1; otherwise, δ = 0.

[0136] The priority of each task chain set is calculated separately, and task chains that meet the core screening criteria are saved to the constructed initial task pool. Each task chain includes supply details, recycling details, and vehicle matching. Supply details include building material type, weight, material requirement point, and delivery time. Recycling details include waste type, volume, pollution level, recycling point, and loading requirements. Vehicle matching includes recommended vehicle type and sorting equipment. The core screening criteria are as follows:

[0137] d C-R ≤d Δ

[0138] F w ∩F e =1

[0139] L rank ≥L Δ

[0140] In the formula, d ΔL is the core distance threshold. Δ The core priority threshold.

[0141] like Figure 3 The diagram shows a flowchart of the loading and matching method of the present invention. Specifically, the loading and matching method includes:

[0142] Based on the weight and density of the building materials, the volume of the building materials is calculated, and a preliminary loading strategy is generated according to the pre-set zoning rules. The zoning rules include: for heavy goods with a density greater than a certain value, they are given priority to be placed in the front area of ​​the cargo hold; for light goods with a lower density, they are placed in the middle of the cargo hold; for fragile items, special protective measures such as adding shock-absorbing pads are taken.

[0143] After the building materials are loaded, the coordinates of the center of gravity (X) after loading are calculated based on the weight and placement of each component. g ,Y g The calculated center of gravity coordinates are compared with a preset safety range. If the coordinates exceed the safety range, the loading layout of building materials is automatically adjusted to ensure that the center of gravity remains within the safe range. The expression is as follows:

[0144]

[0145] In the formula, X g Y g Let X and Y be the horizontal and vertical coordinates of the centroid of the building material in a two-dimensional plane, respectively. i Y is the distance from the center of the cargo to the front of the vehicle. i W is the lateral offset. i For (X) i ,Y i The weight of building materials at point )

[0146] Based on the calculation and planning results, a loading diagram for the material supply stage is generated, including 2D plan view and 3D view. The loading plan information is pushed to the driver through the vehicle terminal, which solves the transportation safety hazard caused by uneven distribution of heavy goods. Through accurate calculation and reasonable layout, the risk of accidents caused by center of gravity shift during transportation is reduced.

[0147] During the loading of building materials and recycling of waste, the vehicle transmits the load distribution and actual volume of waste in real time through pressure sensors and other monitoring equipment; if the actual weight of the supplied materials deviates from the planned weight by more than a certain proportion, the recycling loading plan is automatically adjusted, such as reducing or increasing the estimated amount of recycling waste.

[0148] During the recycling phase, if the actual volume of the waste deviates from the volume estimated by image recognition by more than a certain range, the power of the sorting equipment is adjusted to adapt to the actual waste processing needs. At the same time, a genetic algorithm is used to generate multiple loading schemes, which are comprehensively evaluated from multiple aspects such as space utilization and center of gravity stability, and the optimal loading scheme is selected as the final execution scheme. If the actual situation changes, the evaluation and adjustment are carried out again, which solves the problem of execution deviation caused by rigid loading schemes. Through pre-allocation, real-time feedback and intelligent algorithm optimization, the adaptability and execution efficiency of the loading scheme are improved, ensuring the safety and efficiency of the transportation process.

[0149] Specifically, path planning methods include:

[0150] Extract the geographic coordinates of building material point P, material demand point F, recycling point R, and recycling station J from the task pool, construct the node coordinate set {P,F,R,J}, use map data services such as Gaode API to calculate the actual distance between each node, and calculate the real-time travel time between each node based on the average vehicle speed to form a travel time matrix.

[0151] Combining the supply weight, recycled compaction weight, and pollution level obtained from the loading scheme, multi-dimensional constraints are established, including load constraints and range constraints. The load constraints are used to ensure that the total weight of the supply and recycled goods is within a reasonable range of the vehicle's load limit, while the range constraints are used to ensure that the vehicle's remaining range can meet the needs of the entire journey and reserve a certain safety margin. Consistent with the load verification conditions and range verification conditions, these constraints constitute a feasible path constraint set, providing constraints for subsequent path planning. This solves the problem of rigid planning caused by traditional algorithms that only consider the single factor of distance, and improves the feasibility of path planning.

[0152] The parameters of the ant colony algorithm are initialized, including the number of ants, the initial pheromone concentration, the heuristic factor, the load balance factor, and the empty run penalty coefficient. During the path search process, the ants select the next node based on the state transition probability, giving priority to paths with high pheromone concentration and short distance.

[0153] Taking into account total travel time, total energy consumption, load balancing penalty, and empty travel distance penalty, a multi-objective function is constructed. The ant colony algorithm is then used to optimize the objective function and generate an initial path scheme. The expression of the objective function is shown below:

[0154] Min(Z) = Min(w1×T) total +w2×E total +w3×ΔG+w4×ΔS)

[0155] In the formula, Min(Z) represents the minimum value of a multi-objective function, Z is the multi-objective function, and T is the minimum value of the multi-objective function. total E represents the total travel time.total For total energy consumption, ΔG is the load balance penalty, ΔS is the empty driving distance penalty, and w1, w2, w3, and w4 are the weight coefficients of the objective function.

[0156] When the recycled aggregate inventory at the recycling station reaches a certain standard, a reverse feeding node from the recycling station back to the building material point is forcibly added to the path to form a complete loop path, avoiding empty return trips by vehicles, effectively reducing empty mileage, improving resource utilization efficiency, and solving the problems of load imbalance and high empty mileage in traditional algorithms.

[0157] The system sets an update interval and accesses real-time traffic data based on this interval to calculate the congestion coefficient for each road segment. Once the congestion coefficient falls below a set threshold, it indicates severe congestion on the current road segment. The system then triggers Dijkstra's algorithm to re-search the path portion containing the congested segment, such as the segment from the material demand point to the recycling point, generating a locally optimized path to avoid the congested area and reduce travel time. The expression is shown below:

[0158]

[0159] In the formula, γ is the road segment congestion coefficient, and v real v represents the vehicle's actual speed on the current road segment. max The speed limit for the current road section;

[0160] During the recycling process, if the actual volume of the recycled waste deviates from the previous image recognition result by more than the specified range, the vehicle's remaining range is recalculated. If the remaining range is lower than the safety threshold, the route is immediately adjusted, and the nearest recycling station is selected for unloading to ensure that the vehicle can safely complete the task.

[0161] Based on the center of gravity coordinates of the cargo in the loading plan, and combined with the information on sharp bends in the route, the speed limits of the vehicle when driving on these sections are checked to ensure the stability of the center of gravity and driving safety during the vehicle's journey. At the same time, it is checked whether the arrival time of the recovery point is within the range of the material supply completion time plus the specified loading and unloading time. If it exceeds this range, the task is split up.

[0162] To generate multiple candidate solutions, the weight coefficients of the objective function are adjusted to form three different search directions: shortest time priority, lowest energy priority, and balanced. For shortest time priority, the time factor weight is increased to prioritize the path with the shortest travel time. For lowest energy priority, the energy factor weight is increased to prioritize the smooth road. For balanced, a balanced weight is adopted to take into account both time efficiency and transportation safety. Ant colony algorithm and local optimization are also used to generate multiple candidate path solutions.

[0163] Taking into account time, energy consumption, load balance, and empty runs, the analytic hierarchy process (AHP) is used to comprehensively score candidate routes, select the route with the highest comprehensive score as the optimal route, and output the optimal route, which includes detailed node sequence, estimated arrival time, and real-time monitoring parameters for each segment, such as load changes, remaining range consumption, and sorting equipment start-up nodes. This solves the problem of disconnect between route planning and actual execution, and improves the consistency of route execution and transportation safety.

[0164] Specifically, the monitoring and feedback methods include:

[0165] Data is collected through various devices on the vehicle terminal. Pressure sensors monitor the vehicle's load in real time, battery management devices obtain battery range information, and locate the vehicle's position and speed. At the same time, the vehicle camera captures the unloading and loading scene in real time and connects to external data, including real-time traffic conditions from Gaode Maps, production capacity data from recycling stations, and building material inventory data from factories. This data is integrated with the planned load and planned route to build a three-dimensional data system covering vehicle status, task execution, and external environment. This solves the problem of traditional monitoring data being singular and information fragmentation leading to monitoring lag, and provides comprehensive data support for subsequent anomaly detection.

[0166] Anomaly detection is performed on the task execution process from multiple dimensions; regarding load, the real-time collected load data is compared with the planned load in the loading scheme to calculate the load deviation δ. W If the load deviation exceeds the preset load range, or if the center of gravity calculated based on the real-time load distribution exceeds the safe range, it is determined to be an abnormal load; the expression is as follows:

[0167]

[0168] In the formula, W real For the real-time load of the vehicle, W plan The planned load capacity of the vehicle;

[0169] Regarding battery life, the real-time battery consumption rate γ is calculated. E It compares the consumption rate with the planned value, and triggers a battery life anomaly warning when the consumption rate is too high, exceeding the planned value by a certain percentage; the expression is as follows:

[0170]

[0171] In the formula, E init For continued operation at the start of the mission, E real For real-time battery life data, d travel This represents the distance already traveled.

[0172] In terms of timeliness detection, the time deviation is calculated based on the difference between the actual arrival time and the planned arrival time. If the time deviation exceeds the specified duration, it indicates that the material supply or recycling task is overdue, which will affect the execution of subsequent tasks.

[0173] Through these multi-dimensional detection calculations, various abnormal situations can be identified comprehensively and in a timely manner, significantly improving the accuracy of abnormal detection, realizing the transformation from passive processing to proactive early warning, and outputting the abnormality type;

[0174] For detected anomalies, the system retrieves the classification rules stored in the cloud and performs classification operations based on the anomaly type, categorizing them into emergency, general, and correction. When an anomaly is identified, a Level 1 response is initiated, immediately sending an audible and visual alarm to the driver, displaying a red flashing warning on the vehicle screen, activating the buzzer, and simultaneously issuing a stop and inspection command. When a general anomaly is identified, a Level 2 response is initiated, dynamically adjusting the subsequent route and using a path planning algorithm to replan a detour to minimize the impact on task execution time. When a correction is identified, a Level 3 response is initiated, automatically updating relevant data in the loading plan, such as correcting the garbage weight based on volume deviations and simultaneously revising the estimated remaining battery life. These deviation data are fed back to image recognition, marked as samples to be optimized, and used for subsequent model iteration and optimization to improve recognition accuracy.

[0175] When an abnormal situation causes a task to fail to be executed as originally planned, the task chain is dynamically reconstructed; if a timeliness abnormality causes the recycling task to fail to be completed on time, the task is automatically split into two parts: independent execution of material supply and priority processing of recycling on the next day, and a compensation plan is sent to the user, such as reducing or waiving recycling service fees, to reduce the impact on the user.

[0176] If multiple time-sensitive tasks occur in the same area, the regional task merging strategy is triggered, and large vehicles are dispatched to handle them centrally, thereby improving transportation efficiency.

[0177] like Figure 4 The diagram shows a flowchart of the resource feedback method of the present invention. Specifically, the resource feedback method includes:

[0178] The weight of recycled waste is recorded in real time. Simultaneously, various sensors on the sorting equipment provide real-time feedback on the equipment's operating status and processing progress. Based on the identified waste type and volume information, combined with the corresponding compaction coefficient from the feature knowledge base, the theoretical yield Q of recycled aggregate is calculated. theo The expression is as follows:

[0179] Q theo =V r ×ρ r ×λ r ×η sort

[0180] In the formula, Vr ρ r For the volume and density of the waste, λ r η is the compaction coefficient. sort For sorting and recycling rate;

[0181] After sorting, the recycled aggregate is subjected to quality testing, classified according to particle size, and key indicators such as mud content and strength are tested to determine the quality grade K∈{A,B,H}.

[0182] The actual output, quality grade, and storage location of recycled aggregates are synchronized to the factory's building materials database in real time, triggering the subsequent feedback judgment process. This solves the problem of untimely data updates for recycled aggregates in traditional solutions, and realizes real-time processing from recycling and sorting to data uploading, laying the foundation for rapid response to resource feedback.

[0183] Safety stock thresholds are set for various building material inventories in the factory to measure whether the inventory is at a safe level, and the availability of recycled aggregate is calculated, as shown in the following expression:

[0184] H safe =H day ×α

[0185]

[0186] In the formula, H safe For building materials safety stock, H day Let α be the average daily consumption of building materials, ζ be the guaranteed demand, and Q be the availability rate of recycled aggregates. target The amount of aggregate required for the current production plan;

[0187] Triggering conditions for resource feedback are set based on safety stock thresholds and recycled aggregate availability. When the triggering conditions are met, a reverse transportation order from the recycling station to the building material point is automatically generated, specifying the cargo attributes in the order, including the type and weight of recycled aggregate. Simultaneously, in terms of vehicle scheduling, empty vehicles that have completed recycling tasks and are located close to the recycling station are prioritized for executing reverse transportation orders. This solves the problems of factories repeatedly purchasing virgin materials and vehicles returning empty, achieving precise matching of supply and demand. The triggering conditions are as follows:

[0188] H raw <H safe

[0189] ζ≤ζ th

[0190] K≥B

[0191] In the formula, H raw For original building materials inventory, ζ th The available threshold;

[0192] The reverse transportation node from the recycling station to the building material point is incorporated into the original transportation route to form a complete circular transportation route. The ant colony algorithm is used to recalculate and optimize the new closed loop route, and priority is given to the route scheme that can combine empty return trip and reverse material supply to improve transportation efficiency.

[0193] During the loading process, the loading volume and location of recycled aggregates are rationally planned based on the remaining load space of the vehicle after the feeding and recycling stages. At the same time, corresponding isolation measures are taken to avoid contamination of other goods, taking into account the characteristics of recycled aggregates. This solves the problems of low transportation efficiency and wasted carrying capacity during reverse feeding, making full use of the vehicle's remaining carrying capacity and improving overall transportation efficiency.

[0194] Once the recycled aggregate is transported to the factory, the factory's quality inspection department will conduct random inspections and compare the actual quality data obtained with the theoretical data. If the deviation exceeds a certain range, it will be fed back to the recycling station to prompt the recycling station to adjust the sorting equipment parameters to ensure the stable quality of the recycled aggregate.

[0195] Meanwhile, the factory's finance department calculates the cost of reverse transportation orders, assesses the economic benefits of resource feedback, and dynamically adjusts the feedback trigger threshold and strategy according to different seasons and production conditions. For example, during the off-season for construction, the feedback trigger threshold for recycled aggregates is lowered to avoid inventory backlog. When the factory production line malfunctions, the relevant feedback tasks are automatically suspended to ensure normal production. Through continuous verification, adjustment, and optimization, the stability and reliability of the resource feedback chain are ensured, the utilization rate of recycled resources is continuously improved, and an efficient and sustainable closed-loop economic system of recycling, regeneration, and utilization is formed, solving the problems of easy breakage and incomplete circulation in the feedback chain.

[0196] Example 2

[0197] Please see Figure 5 Another embodiment of the present invention provides a collaborative scheduling system for material supply and resource recycling, comprising: a user terminal, a cloud server, a vehicle terminal, and a zero-carbon factory platform;

[0198] The user terminal is used to receive user input instructions for material feeding and waste recycling, take photos of construction waste and perform standardized preprocessing, solve the problems of non-standard manual data entry and inconsistent image data formats, and provide a standardized data foundation for subsequent instruction processing and image recognition.

[0199] The vehicle-mounted terminal is used to collect vehicle operation data and loading and unloading site information in real time, receive and execute instructions from the cloud server, and respond to abnormal situations in a graded manner to solve the problems of lack of monitoring during transportation and untimely handling of abnormalities, thus ensuring the safe execution of tasks.

[0200] The zero-carbon factory platform is used to synchronize recycled aggregate inventory and quality data, set inventory thresholds, coordinate production plans to adjust resource feedback strategies, solve the problems of data disconnect between the factory and the recycling process and inflexible resource allocation, and promote the low-carbon and circular development of the construction industry.

[0201] The cloud server includes a data acquisition module, a task matching module, a planning module, and an update module;

[0202] The data acquisition module is used to integrate user command data, real-time data from the vehicle terminal, and external environmental data. It performs image recognition on construction waste photos, extracts waste type, volume, and pollution level, maintains a feature knowledge base, and solves the problems of scattered multi-source data and large errors in manual judgment of waste attributes, thereby achieving efficient data acquisition and accurate analysis.

[0203] The task matching module is used to identify high-density geographical areas through spatial clustering algorithms. Combined with time windows and feasibility assessments, it filters and prioritizes material supply and recycling instructions to solve problems such as unreasonable task matching, chaotic execution order, and resource waste, and generates an efficient and feasible task pool.

[0204] The planning module is used to perform path planning based on task pool information and ant colony algorithm, generate initial and candidate paths and select the optimal one, perform loading planning based on building materials and waste attributes, generate loading diagrams and adapt them to the actual situation, trigger reverse material supply from recycling station to building material point when conditions are met, optimize reverse transportation path and loading, and solve the problems of path optimization, reasonable loading layout and reverse resource utilization.

[0205] The update module is used to process abnormal data reported by the vehicle terminal, update the feature knowledge base parameters, verify the resource feedback effect and dynamically optimize the strategy, solve the problems of poor system adaptability and unstable resource feedback, and improve the accuracy and sustainability of the system.

[0206] Specifically, the planning module includes a path planning unit, a loading and equipment unit, and a reverse drive unit;

[0207] The path planning unit is used for initial path planning based on the ant colony algorithm. It dynamically adjusts the path by combining real-time traffic conditions and abnormal situations, generates multiple candidate paths and selects the optimal solution. This solves the problem that traditional path planning only considers a single factor and lacks a dynamic adjustment mechanism, thus achieving intelligent optimization and efficient execution of the path.

[0208] The loading equipment unit is used to plan the loading scheme for the material supply and recycling stages based on the properties of building materials and waste. It optimizes the loading layout through algorithms, provides real-time feedback and adjustments, and solves problems such as unreasonable loading schemes, major transportation safety hazards, and low space utilization, thereby ensuring transportation safety and loading efficiency.

[0209] The reverse drive unit is used to monitor the inventory of recycled aggregates, trigger the reverse transportation task from the recycling station to the building material point, optimize the reverse path and loading plan, solve the problems of repeated purchase of virgin materials by the factory, high vehicle empty running rate and resource waste, and realize the recycling of resources and efficient transportation.

[0210] In summary, this invention uploads material supply instructions and waste recycling instructions via terminal devices, takes photos of the construction site environment and construction waste, obtains relevant information about the waste through image recognition, generates a demand instruction library, and establishes a feature knowledge base. It then performs geographic coordinate analysis and spatial clustering on material demand points and recycling points, and combines this with time window filtering and sorting to generate an initial task pool. Before task execution, a material supply and recycling loading diagram is generated based on the attributes of building materials and waste. An ant colony algorithm is used to plan a loop path including building material points, material demand points, recycling points, and recycling stations. Vehicles execute tasks according to the planned path, transmitting data back to the scheduling platform in real time. Any anomalies are promptly alerted and addressed. After the recycled waste is transported to the recycling station, it is sorted and recycled into aggregate. Data is synchronized to the factory database. When conditions are met, a reverse transportation task from the recycling station to the building material point is automatically triggered, achieving resource feedback and forming a closed loop of material supply and resource recycling.

[0211] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for coordinated scheduling of material supply and resource recovery, characterized in that, include: Obtain material supply instructions and waste recycling instructions, and perform image recognition on construction waste photos to obtain waste attributes and generate a demand instruction library; Establish a feature knowledge base, set up a spatial clustering method, and use spatial clustering and time window filtering to analyze the geographic coordinates and sort the time sequence of the material demand points and recycling points to generate an initial task pool. Set up a loading matching method and generate a material supply and recycling loading diagram based on the building materials and waste attributes. Plan a loop path, set up a path planning method, use the ant colony algorithm for initial path planning, and set up a monitoring and feedback method to transmit data back in real time, monitor the vehicle's operating status, and issue an alarm immediately if any abnormality is detected. The data of sorted and recycled aggregates is synchronized to the building materials database, resource feedback methods are set up, reverse transportation tasks are identified, and a closed loop of material supply and resource recycling is formed.

2. The method for coordinated scheduling of material supply and resource recovery according to claim 1, characterized in that: The specific steps to obtain garbage attributes include: Use terminal equipment to take pictures of construction waste from different angles, and place reference objects next to the construction waste while taking pictures; Standardize the captured images to unify their format; The last fully connected layer of the ResNet-50 model was replaced with a classifier to generate a waste type recognition model. A large number of labeled historical construction waste images were then acquired to train the waste type recognition model. The standardized construction waste images are processed using the waste type recognition model to output waste categories, and the images are labeled with pollution levels. The pollution levels include no pollution, low pollution, and high pollution, and the pollution labels include no pollution and pollution present. The garbage area is detected by the target detection model, and the conversion coefficient between pixels and actual size is calculated by combining the reference object, and the volume of construction waste is estimated. Using the ResNet-50 model as the backbone network, a multi-label classifier is connected to generate a pollution identification model, and the pollution identification model is trained based on historical images; The pollution identification model is used to process images containing pollution, identify hazardous substance labels, generate a set of hazardous labels, and determine the pollution level. Pixel-level masks of garbage regions are obtained through semantic segmentation, the compactness of the regions is calculated, and the garbage stacking pattern is determined. The waste type, volume, pollution level, and stacking method are integrated and saved to the corresponding waste recycling instructions.

3. The method for coordinated scheduling of material supply and resource recovery according to claim 2, characterized in that: The spatial clustering method includes: The latitude and longitude coordinates of the material demand point and the recycling point are obtained from the demand instruction library, converted into a plane coordinate system through Gaussian projection, and the actual distance between the material demand point and the recycling point is calculated to generate a distance matrix. Set the neighborhood radius and minimum number of samples, traverse the distance matrix, and use a dynamic grid clustering algorithm to identify high-density regions. By combining the volume of waste and the weight of building materials, nearby point pairs are filtered and merged, and a set of instruction pairs is output. Instructions that have not formed clusters are marked as pending and automatically enter the regional candidate task pool; Set a candidate scan interval to periodically scan the regional candidate task pool. When a region in the regional candidate task pool is consistently a low-density region, manual verification is triggered.

4. The coordinated scheduling method for material supply and resource recovery according to claim 3, characterized in that: The spatial clustering method further includes: Iterate through the feeding instructions in the instruction pair set, calculate the estimated arrival time, sort them in ascending order, and generate a feeding list; The garbage collection instructions in the instruction set are traversed, and the loading time is calculated based on the garbage stacking pattern. The instructions are then sorted in ascending order to generate a collection list. Based on the estimated arrival time and the standard time for unloading operations, the material supply completion time is calculated. Simultaneously, based on the distances between the recycling point and the material demand point in the distance matrix, the vehicle's empty travel time from the material demand point to the recycling point is calculated, thereby determining the feasible recycling time T. R-start ; Get the timeout window in the recycling command [T] R-begin ,T R-end ], and determine T R-start Whether it falls within the recycling timeframe; where T R-begin and T R-end These are the earliest time that the recycling task can start and the deadline for completion, respectively. If T R-begin ≤T R-start ≤T R-end The instruction pair will be included in the time window matching set; if T R-start <T R-begin Calculate the waiting time and mark it as schedulable, also including it in the time window matching set; if T R-start >T R-end Enter the regional reserve task pool to await subsequent aggregation; and output the task chain set; Based on a feature knowledge base, the weight of compacted recycled waste is calculated. Based on preset load-bearing verification conditions, a load-bearing feasibility flag F is obtained. w ∈{0,1}; Calculate the total energy consumption mileage D. Based on preset range verification conditions, compare the total energy consumption mileage with the vehicle's remaining range to obtain the range feasibility indicator F. e ∈{0,1}; Taking into account the urgency of material supply, the sensitivity of recycling timeliness, and distance penalty, a priority formula is constructed to calculate the priority of each task chain set. The task chains that meet the core screening conditions are saved to the initial task pool.

5. The coordinated scheduling method for material supply and resource recovery according to claim 4, characterized in that: The loading matching method includes: Calculate the volume of building materials based on their weight and density, and generate a preliminary loading strategy according to preset zoning rules. After loading, the center of gravity coordinates are calculated based on the weight and placement of each building material and compared with the preset safe range. If the safe range is exceeded, the loading layout of the building materials is automatically adjusted. Generate a loading diagram for the material supply stage and push it through the vehicle terminal; During loading, load and volume data are transmitted back in real time, and the recycling loading plan is automatically adjusted when the deviation of the feeding weight exceeds the proportion. During the recycling phase, if the actual volume of the waste deviates significantly from the volume estimated by image recognition, the sorting equipment will be adjusted. Multiple loading schemes are generated using a genetic algorithm. The optimal scheme is selected after comprehensively evaluating space utilization and center of gravity stability, and an execution scheme is generated.

6. The coordinated scheduling method for material supply and resource recovery according to claim 5, characterized in that: The path planning method includes: Node coordinate information is extracted from the task pool. Based on the actual distance between each node and the average speed of the vehicle, the real-time travel time between each node is calculated to form a travel time matrix. By combining the feed weight, recycled compaction weight, and pollution level obtained from the loading scheme, multi-dimensional constraints are established. The parameters of the ant colony algorithm are initialized, and a multi-objective function is constructed by comprehensively considering the total travel time, total energy consumption, load balance penalty and empty travel distance penalty. The objective function is then optimized using the ant colony algorithm to generate an initial path scheme.

7. The coordinated scheduling method for material supply and resource recovery according to claim 6, characterized in that: The path planning method also includes: When the recycled aggregate inventory at the recycling station reaches the target, add a reverse feeding node from the recycling station back to the building material point in the path; Set the update interval, and access real-time traffic data based on the update interval to calculate the congestion coefficient of each road segment. Once the congestion coefficient is lower than the set threshold, trigger the Dijkstra algorithm to re-search the congested road segments and generate a locally optimized path. Adjust the weight coefficients of the objective function to generate multiple candidate path schemes with different emphases; After comprehensively evaluating the candidate path schemes using the Analytic Hierarchy Process (AHP), the optimal path is selected and output.

8. The coordinated scheduling method for material supply and resource recovery according to claim 7, characterized in that: The monitoring feedback method includes: A three-dimensional data system is constructed by collecting vehicle status, task execution, and external environment data through in-vehicle terminals. Perform multi-dimensional anomaly detection, calculate load deviation, range consumption rate and time deviation to determine the anomaly type; The system responds in a tiered manner based on the type of anomaly. In case of an emergency anomaly, an alarm is sent and vehicle operation is locked. In case of a general anomaly, the path is dynamically adjusted, the loading data is updated and feedback is provided for optimization. When an abnormal situation causes a task to fail to execute as originally planned, the task chain is dynamically restructured. If multiple time-sensitive tasks occur in the same region, the region task merging strategy will be triggered.

9. The method for coordinated scheduling of material supply and resource recovery according to claim 8, characterized in that: The resource feedback method includes: The weight of recycled waste is recorded in real time, and the theoretical yield of recycled aggregate is calculated by combining the waste type, volume and compaction coefficient. The recycled aggregate is subjected to quality testing to determine its quality grade; Synchronize actual output, quality grade, and storage location to the factory's building materials database; Set a safety stock threshold for factory building materials and calculate the availability of recycled aggregates; When the triggering conditions are met, a reverse transportation order from the recycling station to the building materials point is generated, and empty vehicles are prioritized for execution. Incorporate and optimize the reverse transport nodes into the original route, and plan the loading volume and location of recycled aggregate based on the remaining load space; After transportation, compare quality data, provide feedback to adjust sorting equipment parameters, calculate costs, and dynamically adjust feedback strategies.

10. A collaborative scheduling system for material supply and resource recovery, used to implement the collaborative scheduling method for material supply and resource recovery as described in any one of claims 1-9, characterized in that, include: User terminals, cloud servers, vehicle terminals, and zero-carbon factory platforms; The user terminal is used to receive user input instructions for material supply and waste recycling, take photos of construction waste, and perform standardized preprocessing. The vehicle-mounted terminal is used to collect vehicle operation data and loading and unloading site information in real time, receive and execute cloud server instructions, and provide graded responses to abnormal situations. The zero-carbon factory platform is used to synchronize recycled aggregate inventory and quality data, set inventory thresholds, and coordinate with production plans to adjust resource feedback strategies. The cloud server includes a data acquisition module, a task matching module, a planning module, and an update module; The data acquisition module is used to integrate user command data, real-time data from the vehicle terminal, and external environmental data to perform image recognition on construction waste photos, extract waste attributes, and maintain a feature knowledge base. The task matching module is used to filter and prioritize material supply and recycling instructions through spatial clustering, time window filtering, and feasibility assessment to generate a task pool. The planning module is used to perform path planning and loading planning based on task pool information, trigger reverse material supply from the recycling station to the building material point, and optimize the path and loading. The update module is used to process abnormal data, update feature knowledge base parameters, and verify and optimize resource feedback strategies.

Images