Intelligent sorting method and platform for annular cross belt

Abstract

The invention discloses an intelligent sorting method and platform for an annular cross belt, and relates to the technical field of intelligent sorting, and the method comprises the following steps: scanning an RFID tag to identify basic information of an article and extract features of the article by using an Internet of Things technology; a 3D visual sensor is used for conducting three-dimensional modeling on an object, and a mechanical arm is connected to conduct self-adaptive grabbing mode adjustment; an article distribution strategy is adjusted in real time through the central control center, and a bidirectional distribution path is matched; the path starting point of the bidirectional distribution path is a switching guide table, a first sliding groove corresponding to the first guide is connected with the forward rotation annular cross belt, and a second sliding groove corresponding to the second guide is connected with the reverse rotation annular cross belt. And implementing a flow direction optimization strategy to intelligently sort the articles. The technical problems that in the prior art, a sorting platform is low in automation and intelligence degree and low in sorting accuracy and efficiency are solved, and the technical effects that the automation and intelligence level of the sorting platform is enhanced, and the sorting accuracy and efficiency are improved are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent sorting, and in particular to an intelligent sorting method and platform for a circular cross belt. Background Art

[0002] In the context of the rapid development of the modern logistics industry, efficient and accurate sorting technology has become the key to improving the efficiency of logistics operations. Circular cross-belt sorting has emerged as an advanced sorting method. The emergence of the circular cross-belt sorting system has changed this situation. It can realize the rapid classification and transportation of a large number of items through a circular conveyor belt and a cross-set sorting port. However, the circular cross-belt sorting system still has some shortcomings. It is not accurate enough in identifying item information, resulting in items being mistakenly assigned to inappropriate sorting ports. At the same time, the speed of the conveyor belt and the coordinated work of the sorting port are not optimized, which affects the overall sorting efficiency, and the level of intelligence and automation is low.

[0003] The existing technology has technical problems such as low automation and intelligence of the sorting platform, and low sorting accuracy and efficiency. Summary of the invention

[0004] The present application provides an intelligent sorting method and platform for a circular cross belt, which is used to solve the technical problems in the prior art of low automation and intelligence of the sorting platform, and low sorting accuracy and efficiency.

[0005] In view of the above problems, the present application provides an intelligent sorting method and platform for a circular cross belt.

[0006] In a first aspect of the present application, there is provided an intelligent sorting method for a circular cross belt, the method comprising: The Internet of Things technology is used to scan RFID tags to identify basic information of items, and in combination with deep learning technology, an image processing module is used to extract item features, wherein the basic information of the items includes destination information, and the item features include item shape features and item size features; based on the item features, a 3D visual sensor is used to perform three-dimensional modeling of the items, and a robotic arm is connected to adaptively adjust the grasping mode, and the grasping mode is determined by the stability screening of the robotic arm; the item distribution strategy is adjusted in real time through the central control center, and a reinforcement learning algorithm is used to dynamically match the bidirectional distribution path in combination with the basic information of the items, and the bidirectional distribution path includes a forward distribution path connected to the forward roundabout inlet and a reverse distribution path connected to the reverse roundabout inlet; the path starting point of the bidirectional distribution path is a switching guide platform, and the switching guide platform includes a first guide and a second guide, the first chute corresponding to the first guide is connected to the forward roundabout, and the second chute corresponding to the second guide is connected to the reverse roundabout; based on the bidirectional distribution path, a flow optimization strategy is implemented, and multimodal sensor data is integrated to perform intelligent sorting of items.

[0007] The second aspect of the present application provides an intelligent sorting platform for a circular cross belt, the platform comprising: An item feature extraction module, which uses the Internet of Things technology to scan RFID tags to identify basic information about items, and combines deep learning technology to extract item features using an image processing module, wherein the item basic information includes destination information, and the item features include item shape features and item size features; a grasping mode adjustment module, which uses a 3D visual sensor to perform three-dimensional modeling of items based on the item features, and connects to a robotic arm to adaptively adjust the grasping mode, wherein the grasping mode is determined by the stability screening of the robotic arm; a two-way allocation path matching module, which is used to adjust the item allocation strategy in real time through a central control center, and adopts a strong A chemical learning algorithm is used to dynamically match a two-way distribution path in combination with the basic information of the item, wherein the two-way distribution path includes a forward distribution path connected to a forward roundabout belt feed port and a reverse distribution path connected to a reverse roundabout belt feed port; a roundabout belt connection module, wherein the path starting point of the two-way distribution path is a switching guide platform, and the switching guide platform includes a first guide and a second guide, wherein a first chute corresponding to the first guide is connected to the forward roundabout belt, and a second chute corresponding to the second guide is connected to the reverse roundabout belt; an item intelligent sorting module, wherein the item intelligent sorting module implements a flow optimization strategy based on the two-way distribution path, and integrates multimodal sensor data to perform intelligent sorting of items.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: The Internet of Things technology is used to scan RFID tags to identify the basic information of items, and the image processing module is used to extract item features in combination with deep learning technology; based on the item features, a 3D visual sensor is used to perform three-dimensional modeling of the items, and the robotic arm is connected to adaptively adjust the grasping method; the item distribution strategy is adjusted in real time through the central control center, and a reinforcement learning algorithm is used to dynamically match the two-way distribution path in combination with the basic information of the items, and the two-way distribution path includes a forward distribution path connected to the forward roundabout belt feed port and a reverse distribution path connected to the reverse roundabout belt feed port; the path starting point of the two-way distribution path is a switching guide platform, and the switching guide platform includes a first guide and a second guide, the first chute corresponding to the first guide is connected to the forward roundabout belt, and the second chute corresponding to the second guide is connected to the reverse roundabout belt; based on the two-way distribution path, a flow optimization strategy is implemented, and multimodal sensor data is integrated to perform intelligent sorting of items. The technical effect of strengthening the automation and intelligence level of the sorting platform and improving the accuracy and efficiency of sorting is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic flow chart of an intelligent sorting method for a circular cross belt provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of an intelligent sorting platform for a circular cross belt provided in an embodiment of the present application.

[0011] Explanation of the reference numerals: item feature extraction module 10 , grasping mode adjustment module 20 , bidirectional allocation path matching module 30 , circular cross belt connection module 40 , item intelligent sorting module 50 . DETAILED DESCRIPTION

[0012] The present application provides an intelligent sorting method and platform for a circular cross belt, which is used to solve the technical problems in the prior art of low automation and intelligence of the sorting platform, and low sorting accuracy and efficiency.

[0013] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0014] Embodiment 1 like Figure 1 As shown, the present application provides an intelligent sorting method for a circular cross belt, the method comprising: Step S100: Utilize the Internet of Things technology to scan the RFID tag to identify the basic information of the item, and combine the deep learning technology to use the image processing module to extract the item features, wherein the basic information of the item includes the destination information, and the item features include the shape features and the size features of the item.

[0015] Specifically, the radio frequency identification (RFID) technology in the Internet of Things technology is used to scan the RFID tags on the items. When the RFID reader transmits radio waves of a specific frequency, the tag is activated and returns a signal containing basic information about the item, including crucial destination information. This information is received and identified, providing basic guidance for the subsequent sorting and distribution process. At the same time, deep learning technology is introduced to enhance the ability to extract item features. The deep learning model, using a convolutional neural network (CNN), can automatically learn the shape characteristics of an item from a large amount of image data. For example, it can identify whether the item is round, square, long, or other irregular shapes. The image processing module further analyzes the image of the item and accurately measures the size characteristics of the item, including length, width, height, etc. The image processing module uses edge detection, contour extraction and other algorithms to determine the shape boundary of the item, and then obtains the size information by pixel calculation or comparison with a reference object of known size. For example, when processing a package, the destination is obtained by scanning the RFID tag, which is a distribution center in a certain city. At the same time, the deep learning technology and image processing module analyze the shape of the package as a rectangular parallelepiped, and the size characteristics are 50 cm long, 30 cm wide, and 20 cm high. These detailed information enable reasonable sorting and transportation arrangements based on the characteristics of the items, such as selecting a suitable container or transportation tool to load the package to ensure that it is delivered to the destination safely and efficiently. In addition, as more item data is processed, the deep learning model will continue to be optimized and improved, and can more accurately identify and extract the characteristics of items of various complex shapes and sizes, improving the performance and adaptability of the entire sorting system.

[0016] Step S200: Based on the features of the object, a 3D visual sensor is used to perform three-dimensional modeling of the object, and a robotic arm is connected to adaptively adjust a grasping mode, wherein the grasping mode is determined by a stability screening performed by the robotic arm.

[0017] Specifically, based on the previously acquired object features, the 3D vision sensor is used to model the object in three dimensions. The 3D vision sensor can capture the three-dimensional shape, spatial position and other information of the object, thereby establishing an accurate three-dimensional model. Next, the system will connect the robotic arm and adaptively adjust the gripping method of the robotic arm. Before the robotic arm grasps, it will perform a stability screening to determine the most appropriate grasping method. Stability screening will take into account multiple factors, such as the shape, weight distribution, and surface material of the object. For irregularly shaped objects, the robotic arm will choose to grasp from multiple angles to ensure the stability of the grasping; for heavier objects, the robotic arm will adjust the grasping force and position to prevent the object from slipping or being damaged during the grasping process. For example, if the object is an object with a smooth surface and irregular shape, the robotic arm will try different grasping angles and forces many times, and finally determine a way to stably grasp the object, such as using a special clamp or increasing the contact area of ​​the grasping. Through this process of adaptively adjusting the grasping method, the robotic arm can better adapt to the characteristics of different objects, improve the efficiency and accuracy of grasping, and reduce the risk of damage to objects during the sorting process.

[0018] Step S300: The item distribution strategy is adjusted in real time by the central control center, and a reinforcement learning algorithm is used to dynamically match the bidirectional distribution path in combination with the basic information of the item. The bidirectional distribution path includes a forward distribution path connected to the forward roundabout inlet and a reverse distribution path connected to the reverse roundabout inlet.

[0019] Specifically, the central control center plays a core role. It can receive and process various information in real time, including basic information of items and status information of each link in the system. The central control center will adjust the distribution strategy of items in real time based on this information. In order to achieve more optimized distribution, it adopts reinforcement learning algorithm. Reinforcement learning algorithm is a machine learning algorithm that can learn the optimal strategy by interacting with the environment and based on feedback. In this process, the reinforcement learning algorithm will dynamically match the bidirectional distribution path by combining the basic information of the items, including destination information, item characteristics, etc. The bidirectional distribution path includes a forward distribution path and a reverse distribution path. The forward distribution path is connected to the feed port of the forward roundabout belt to distribute items to the forward roundabout belt for transportation; the reverse distribution path is connected to the feed port of the reverse roundabout belt to distribute items to the reverse roundabout belt for transportation. For example, when the central control center receives the basic information of an item, it will use the reinforcement learning algorithm to decide whether to assign the item to the forward distribution path or the reverse distribution path based on the destination of the item and the operation of the two roundabouts in the current system. If the transportation pressure on the forward roundabout is small and the destination of the item is more suitable for transportation through the forward roundabout, the central control center will assign the item to the forward distribution path. This real-time adjustment and dynamic matching method can improve the efficiency of item sorting and transportation, reduce congestion and delays, and operate more efficiently.

[0020] Step S400: The starting point of the bidirectional distribution path is a switching guide platform, and the switching guide platform includes a first guide and a second guide. The first slide groove corresponding to the first guide is connected to the forward annular cross belt, and the second slide groove corresponding to the second guide is connected to the reverse annular cross belt.

[0021] Specifically, the starting point of the two-way distribution path is the switching guide, which plays a key role in diversion. The switching guide includes two parts: the first guide and the second guide. The first chute corresponding to the first guide is connected to the forward circular cross belt, which means that when the item needs to be allocated to the forward circular cross belt, the switching guide will guide the item to the first chute through the first guide, and then the item enters the forward circular cross belt. Similarly, the second chute corresponding to the second guide is connected to the reverse circular cross belt. When the item needs to be allocated to the reverse circular cross belt, the switching guide will guide the item to the second chute through the second guide, so that the item enters the reverse circular cross belt. For example, when an item arrives at the switching guide, the central control center will determine whether the item should enter the forward circular cross belt or the reverse circular cross belt according to the previously formulated allocation strategy. If it is decided to allocate the item to the forward circular cross belt, the switching guide will adjust the first guide so that the item enters the forward circular cross belt along the first chute; if the item should enter the reverse circular cross belt, the switching guide will adjust the second guide so that the item enters the reverse circular cross belt along the second chute. This design enables items to accurately enter the corresponding circular cross belt, ensuring the efficient operation of the sorting system and the accurate distribution of items.

[0022] Step S500: Based on the bidirectional distribution path, a flow optimization strategy is implemented, and multimodal sensor data is integrated to perform intelligent sorting of items.

[0023] Specifically, based on the bidirectional distribution path, a flow optimization strategy is implemented to achieve intelligent sorting of items. Specifically, the flow optimization strategy considers the following aspects: Item destination analysis: First, determine the final target location of the item based on the item's destination information. By analyzing the distribution and demand of the destination, plan a more reasonable flow path to reduce the transportation distance and time; Roundabout load monitoring: Real-time monitoring of the load conditions of the forward roundabout and the reverse roundabout, including the number of items, transportation speed, etc. Adjust the distribution of items according to the load conditions to avoid excessive congestion in a certain roundabout and improve the overall transportation efficiency; Transportation time estimation: Combine the item's destination, current location and the running speed of the roundabout to estimate the time required for the item to reach the destination. By optimizing the flow direction, select the fastest transportation path to ensure that the items can be delivered in time; Dynamic adjustment: According to the real-time transportation situation and changes, the system can dynamically adjust the flow strategy. For example, if a roundabout fails or is congested, the system can adjust the item distribution path in time to ensure the smooth progress of the sorting work.

[0024] At the same time, multimodal sensor data is integrated to perform intelligent sorting of items. Specifically: Visual sensor data: The appearance information of the object, such as shape, color, and size, is obtained through visual sensors. This information can help the system identify the category and characteristics of the object, so as to sort it more accurately. For example, for objects with special shapes, the appropriate grasping method and transportation path are selected according to their shape; weight sensor data: The weight sensor can measure the weight of the object. According to the weight of the object, the grasping force of the robot arm and the carrying capacity of the transportation equipment can be adjusted to ensure the safe transportation of the object; position sensor data: The position sensor can track the location information of the object in real time, which helps to accurately control the flow of the object and ensure that the object can be transported and sorted according to the predetermined path. Other sensor data: It also includes temperature sensors, humidity sensors, etc., which are used to obtain environmental information of the object. This information can play an important role in the sorting of some special items, such as temperature-sensitive items or items that require a specific humidity environment; by fusing these multimodal sensor data, it is possible to fully understand the situation of the object and make smarter sorting decisions. For example, when an object enters the sorting system, its destination information is first obtained through the RFID tag, and then the shape and color of the object are identified in combination with the visual sensor, the weight sensor measures the weight of the object, and the position sensor determines the current location of the object. Based on this information, after comprehensive analysis, the best two-way distribution path is selected, and the robot arm is controlled to place the items on the corresponding circular cross belt in an appropriate grasping manner. During the transportation process, the flow direction of the items is adjusted in real time according to the sensor data to ensure that the items can reach the destination accurately and quickly. Intelligent and efficient sorting of items is achieved, and the accuracy and reliability of the sorting system are improved.

[0025] In a possible implementation, step S100 further includes: Step S110: deploy a weight sensor to measure the weight information of the item in real time and associate and bind it with the RFID tag.

[0026] Step S120: using convolutional neural network technology to extract color features of the items.

[0027] Step S130: combining the basic information of the item with the color features of the item, and using a support vector machine classifier to initialize the item sorting configuration.

[0028] Specifically, a suitable weight sensor is selected and installed at the corresponding position of the item sorting to ensure that the weight of the item can be accurately measured. The weight sensor is connected to the control unit to transmit weight data in real time. When the item passes through the weight sensor, the sensor automatically measures the weight of the item and sends the weight information to the control unit. After receiving the weight information, the control unit will associate it with the RFID tag of the item. This can be achieved by establishing a database or data structure to store the weight information and the item information corresponding to the RFID tag together. After association and binding, the weight information of the item can be obtained by querying the RFID tag, so that corresponding processing can be performed according to the weight in the subsequent sorting process, such as selecting appropriate handling equipment, determining the sorting path, etc.

[0029] The specific process of using convolutional neural network (CNN) technology to extract the color features of items is as follows: Data collection: Collect a large amount of image data containing various objects, and annotate these images to mark the color information of the objects; Build a convolutional neural network model: Select a suitable convolutional neural network architecture, such as the common LeNet, AlexNet, VGGNet or ResNet, etc. These models are usually composed of convolutional layers, pooling layers and fully connected layers; Convolutional layer operation: The convolutional layer performs convolution operations on the input image by using multiple convolution kernels (filters) to extract local features of the image. Each convolution kernel will slide on the image and perform convolution operations with the local area of ​​the image to obtain a feature map; pooling layer operation: the pooling layer usually uses maximum pooling or average pooling operations to downsample the feature map output by the convolution layer, reduce the size of the feature map, and retain important feature information; fully connected layer operation: after being processed by multiple convolutional layers and pooling layers, the obtained feature map is flattened into a one-dimensional vector and input into the fully connected layer. The fully connected layer integrates and classifies these features and outputs the color features of the object; training model: use the labeled image data to train the convolutional neural network. During the training process, by adjusting the parameters of the model, the model can accurately predict the color features of the object; feature extraction: in actual applications, the image of the object to be identified is input into the trained convolutional neural network, and the model will automatically extract the color features of the object.

[0030] The basic information of the item (such as the destination information, etc.) is organically combined with the color characteristics of the item. This can be achieved by combining this information into a feature vector, which contains the relevant data of the basic information and color characteristics of the item. Next, the support vector machine (SVM) classifier is used to process these feature vectors. SVM is a powerful machine learning algorithm that can find the optimal classification boundary in a high-dimensional space to achieve data classification. When initializing the item sorting configuration, the trained SVM classifier is applied to the new item feature vector. The SVM classifier will determine which sorting category the item should belong to based on the input feature vector and output the corresponding initialization sorting configuration. For example, assuming that the basic information of the item includes the destination of city A and the color feature of red, after combining this information into a feature vector and inputting it into the SVM classifier, the SVM classifier will determine that the item should be assigned to the sorting route to city A based on previous training experience, and will group it with other items with similar characteristics or take a specific treatment method based on the color feature of red. In this way, by utilizing the combination of basic information and color features of items, as well as the powerful classification capability of the SVM classifier, the item sorting configuration can be initialized, providing an important basis and guidance for subsequent further sorting and processing.

[0031] In a possible implementation, step S130 further includes: Step S131: on the forward circular cross belt and the reverse circular cross belt, based on historical sorting instances, key nodes are located, and edge computing servers are deployed on the key nodes.

[0032] Step S132: using a streaming data processing framework to interact with the support vector machine classifier and make sorting decisions.

[0033] Specifically, a comprehensive analysis is conducted on the forward and reverse circular cross belts, and the data of historical sorting instances are consulted. These data record various information about items in the sorting process in the past, such as the type, quantity, flow direction, and processing time at various locations of the items. Through in-depth mining and analysis of these historical data, nodes that play a key role in the sorting process can be identified. These key nodes are locations with large item flows, or areas with more complex sorting operations, or places prone to congestion or bottlenecks. Once these key nodes are determined, edge computing servers will be deployed at these locations. The edge computing servers have powerful computing and data processing capabilities and can process large amounts of data generated at key nodes in real time. For example, if it is found in historical sorting instances that a large number of different types of items often converge at a certain location, and complex sorting judgments and operations are required, then this location will be determined as a key node. After the edge computing server is deployed at this key node, when the item arrives at this location, the edge computing server can quickly process and analyze the relevant information of the item, and make accurate sorting decisions in a timely manner, thereby improving the efficiency and accuracy of the entire sorting system.

[0034] The streaming data processing framework plays a key role. It can receive and process data streams from the forward and reverse roundabouts in real time. These data streams contain various real-time information about the items during the sorting process, such as the current position, speed, direction, and features of the items. When the streaming data processing framework receives this data, it interacts with the support vector machine classifier, which is a machine learning-based classifier that has learned how to make sorting decisions based on the features of the items through previous training. The streaming data processing framework passes the real-time data to the support vector machine classifier, which classifies and judges the items based on this data to determine which position or path the item should be sorted to. For example, when an item moves on the roundabout, the streaming data processing framework obtains the position and feature information of the item in real time and passes this information to the support vector machine classifier, which makes sorting decisions based on the features of the item, such as color, shape, size, etc., as well as the previously learned patterns, such as sorting the item to a specific sorting port or conveying it to a specific roundabout area. In this way, the interaction between the streaming data processing framework and the support vector machine classifier can achieve real-time sorting decisions, ensuring that items can be sorted to the correct location quickly and accurately, improving the efficiency and accuracy of sorting.

[0035] In a possible implementation, step S131 further includes: Step S1311: Based on the forward-rotating circular cross-belt, a forward-sorting exit sequence is determined, wherein the forward-sorting exit sequence includes M forward-rotating circular cross-belt sorting exits.

[0036] Step S1312: Based on the reverse roundabout, determine a reverse sorting exit sequence, wherein the reverse sorting exit sequence includes N reverse roundabout sorting exits.

[0037] Step S1313: Connect to the logistics center, set up a temporary storage area, and use the edge computing server to dynamically manage the positive sorting export sequence and the reverse sorting export sequence with circulation efficiency as the optimization goal.

[0038] Specifically, a detailed analysis and study of the forward circular cross belt is conducted, and its structure, layout, and the law of item flow are considered. Based on these factors, a series of exit positions are determined to form a forward-sorted exit sequence, which contains a total of M forward circular cross belt sorting ports. For example, assuming that the length of the forward circular cross belt is a certain value, sorting ports are set at different positions according to the type of item, destination, or other preset classification rules. After comprehensive analysis and planning, M specific positions arranged in sequence from the starting position are determined as sorting ports, which constitute a forward-sorted exit sequence. By determining this forward-sorted exit sequence, a clear path and guidance are provided for the accurate sorting and efficient outflow of subsequent items on the forward circular cross belt.

[0039] The structural characteristics, operation mode and flow characteristics of the reverse roundabout are also analyzed in depth. Based on these analysis results, a series of exit locations for the outflow of items are identified, thereby establishing a reverse sorting exit sequence. This sequence covers N reverse roundabout sorting ports. The purpose of determining the reverse sorting exit sequence is to ensure that the items on the reverse roundabout can flow out of the corresponding sorting ports accurately and efficiently, and realize an orderly sorting and transportation process.

[0040] Connected with the logistics center, information and resources are shared and coordinated. At the same time, a temporary storage area is set up to temporarily store items in special situations or traffic fluctuations. For the previously determined forward sorting exit sequence and reverse sorting exit sequence, the flow efficiency is set as the optimization goal. This means that the operation of the system is designed to maximize the speed and accuracy of items flowing out of these exits and reduce congestion and delays. To achieve this goal, edge computing servers are used for dynamic management. The edge computing servers can obtain various data about item flow, exit status, storage area conditions, etc. in real time, and dynamically adjust the flow direction and allocation of items based on these real-time data. For example, if a forward roundabout sorting port is congested, the edge computing server will temporarily reduce the number of items allocated to the sorting port, guide it to other relatively idle forward roundabout sorting ports, or store it in a temporary storage area first, waiting for congestion to ease before processing. Similarly, if some sorting ports in the reverse sorting exit sequence are lightly loaded while other sorting ports are under heavy pressure, the edge computing server will reallocate the outflow paths of the items to balance the workload of each sorting port and improve the overall flow efficiency. By connecting the logistics center, setting up a temporary storage area, and relying on the dynamic management of the edge computing server, the use of the positive sorting exit sequence and the reverse sorting exit sequence can be continuously optimized to achieve the best flow efficiency.

[0041] In a possible implementation, step S1313 further includes: Step S13131: Match the destination information in the basic information of the item with the temporary storage area where the location mark exists. After the match is successful, the item circulation task is synchronously updated in the logistics center.

[0042] Step S13132: Based on the edge computing server, the urgency of the updated item circulation tasks is evaluated to determine the priority of each item circulation task.

[0043] Step S13133: Based on the priority of each item circulation task, parallel scheduling management is performed with the positive sorting exit sequence and the reverse sorting exit sequence.

[0044] Specifically, the destination information in the basic information of the item is normalized and standardized, and possible redundant characters, spaces or incorrect formats are removed to make it clear and comparable data. At the same time, the location tags of the temporary storage area are also processed in the same way to ensure the consistency of the data format. According to the preset classification system, specific classification codes are assigned to the destination information and location tags. For example, hierarchical coding is performed according to regions, provinces, and cities, or classified coding is performed according to logistics routes, distribution centers, etc. Try to make an accurate coding comparison. If the coding of the item destination is exactly the same as the coding of a temporary storage area location tag, it is directly determined to be a successful match. If the exact match is not successful, fuzzy matching is performed, and a string similarity algorithm such as edit distance, cosine similarity, etc. is used to calculate the similarity between the destination information and the location tag. A similarity threshold is set. When the calculated similarity exceeds the threshold, the match is considered successful. Matching is performed according to the hierarchy of the destination. For example, if the location tag is "A region" and the item destination is "Province A", since "Province A" belongs to "A region", it can also be considered a successful match. In addition to the destination information, other attributes of the item are also considered, such as the item's size, weight, priority, etc. Only when these comprehensive conditions are consistent with the conditions set by the location tag, the match is considered successful. Multiple rounds of matching and verification. If a match is unsuccessful, multiple rounds of matching attempts can be made, or manual review and intervention can be performed to ensure the accuracy of the match. Once the match is successful, the item flow task is immediately updated in the logistics center, including the current location of the item, the estimated outflow time, the subsequent processing steps and other key information, so that the entire logistics system can grasp the status and flow progress of the item in real time, and make corresponding adjustments and optimizations.

[0045] The edge computing server obtains the updated information about the item flow tasks, including but not limited to the following aspects: delivery deadline, check the item delivery deadline specified in each task, and tasks close to the deadline are generally considered more urgent; customer demand level, some customers have special priority requirements, such as the needs of important customers or long-term partners are given higher urgency. Item nature, considering the characteristics of the items, such as perishable items, urgently needed medical supplies or key production parts, the flow tasks of these items have higher urgency. Logistics cost impact, if delaying the flow of an item will result in significantly increased logistics costs, such as additional warehousing fees or transportation adjustment costs, the task is evaluated as more urgent. Then, the edge computing server uses the preset evaluation algorithm and weights to comprehensively analyze these factors. For example, the delivery deadline is given a higher weight, followed by the weight of the item nature, and the customer demand level and logistics cost impact will also have corresponding weights. By weighted calculation of these factors, the urgency score of each item flow task is obtained. Finally, the priority of each item flow task is determined according to the urgency score. The higher the score, the higher the priority. In this way, through the rapid processing and accurate evaluation of the edge computing server, a reasonable priority can be determined for each item flow task, thereby optimizing the logistics process and improving service quality and efficiency.

[0046] Based on the priority of each item flow task, parallel scheduling management is carried out with the positive sorting exit sequence and the reverse sorting exit sequence to clearly determine the priority of each item flow task. High-priority tasks mean that they need to be processed more quickly and with priority. When performing scheduling management, high-priority item flow tasks will be assigned first to the sorting ports in the positive sorting exit sequence and the reverse sorting exit sequence, which means that high-priority items will be arranged to flow out of these sorting ports first to complete their circulation process as soon as possible. At the same time, the execution of item flow tasks and the working status of each sorting port will be monitored in real time. If a new high-priority task appears during the processing, or a sorting port fails or is congested, the scheduling strategy will be adjusted in time to reallocate tasks to the appropriate sorting port to ensure that high-priority items can complete the circulation at the fastest speed. In addition, in parallel scheduling management, actual factors such as the quantity, size, and weight of items will also be considered to make full use of the processing capacity of the positive sorting exit sequence and the reverse sorting exit sequence to achieve efficient and accurate item flow scheduling. In this way, the circulation needs of items of different priorities can be effectively balanced, improving operational efficiency and service quality.

[0047] In a possible implementation, step S13133 further includes: Based on the M forward-turning roundabout sorting ports in the forward-sorting exit sequence and the N reverse-turning roundabout sorting ports in the reverse-sorting exit sequence, a sorting path decision formula is used to traverse multiple sorting paths.

[0048] The sorting path decision formula: , where j is the index of the sorting port, j∈{1,…,M+N}, is the priority of the jth sorting port, is the matching degree of the jth sorting port, is the turnover efficiency of the jth sorting port, is the operating cost of the jth sorting port, is the prediction model error, , , , It is the decision factor of each factor, which is used to adjust the importance of different factors in the sorting path decision process.

[0049] Specifically, first, it is clear that there are M forward roundabout sorting ports in the forward sorting exit sequence, and N reverse roundabout sorting ports in the reverse sorting exit sequence. Then, the sorting path decision formula is used to calculate and evaluate each sorting port combination. This means that the system will consider selecting different sorting port combinations from these M+N sorting ports one by one, thereby forming multiple potential sorting paths. By constantly changing the selection and combination of sorting ports, various sorting paths can be fully explored to find the optimal solution. Such a traversal process can cover all situations and ensure that no potential efficient sorting paths are missed, thereby improving the efficiency and accuracy of the entire sorting. The sorting path decision formula takes into account multiple important factors related to the sorting ports to determine the pros and cons of each sorting path. The index j ranges from 1 to M+N, covering the M sorting ports in the forward sorting exit sequence and the N sorting ports in the reverse sorting exit sequence. is the priority of the jth sorting port. A higher priority usually means that the sorting port has a higher importance or urgency in processing certain items. It is the matching degree of the jth sorting port, which reflects the degree of adaptability between the item and the sorting port. A high matching degree indicates that the sorting port is more suitable for processing the current item. is the flow efficiency of the jth sorting port. A higher flow efficiency means that items can pass through the sorting port faster to complete the sorting process. is the operating cost of the j-th sorting port, including costs in terms of manpower, equipment usage, and energy consumption. is the forecast model error, which is used to account for possible inaccuracies or biases in the model when predicting and evaluating these factors. , , , It is the decision factor of each factor, which is used to adjust the relative weight or importance of different factors in the overall evaluation. Through the calculation of this formula, each possible sorting path (i.e., the combination of different sorting ports) can be quantitatively evaluated, so as to select the sorting path with the best comprehensive score to achieve efficient, accurate and cost-effective sorting operations.

[0050] In a possible implementation, step S13133 further includes: Use the Isolation Forest algorithm to monitor in real time the priority of the positively sorted export sequence. The forward roundabout sorting port is used to locate the first fault source set; the Isolation Forest algorithm is used to monitor the priority of the reverse sorting exit sequence in real time. A reverse roundabout sorting port is used to locate the second set of fault sources; based on the first set of fault sources and the second set of fault sources, the fault response mechanism is dynamically adjusted, and the resources of the forward roundabout and the reverse roundabout are mutually reallocated.

[0051] Specifically, the Isolation Forest algorithm is used to monitor the top priority in the positive order export sequence in real time. The Isolation Forest algorithm continuously obtains the real-time operation data of these sorting ports, including but not limited to the number of items processed, speed, error rate, equipment operation parameters and other information. Then, based on its unique principle of isolated abnormal points, the data are deeply analyzed. If the data of a sorting port shows obvious abnormal characteristics, such as its processing speed is significantly lower than the average level of other sorting ports, or the error rate is far beyond the normal range, and it can be relatively easily isolated in data segmentation, then this sorting port will be identified as the source of fault. All the sorting ports that are determined by the algorithm as possible faults are summarized to form the first fault source set. For example, if there are a total of 10 sorting ports in the positive sorting exit sequence, the top 5 sorting ports with the highest priority will be monitored. Assuming that the processing speed of the third sorting port among these 5 sorting ports has been very slow, and it is easy to be distinguished separately in multiple analyses of the Isolation Forest algorithm, then this sorting port will be included in the first fault source set. Through such real-time monitoring and precise positioning, potential fault points in high-priority sorting ports can be quickly discovered, providing a key basis for timely maintenance, adjustment of strategies, and resource reallocation measures, thereby ensuring the efficient and stable operation of the entire forward rotating circular cross belt.

[0052] Similarly, for the reverse order of the export sequence, the Isolation Forest algorithm is also used to monitor the priority in real time. The reverse roundabout belt sorting ports are selected and the sorting ports that show abnormal performance are located as the second fault source set. After the first fault source set and the second fault source set are determined, the fault response mechanism is dynamically adjusted according to the fault conditions in the two sets. First, the severity, impact range and urgency of the fault will be evaluated. If the fault is serious and has a large impact, the emergency response process will be immediately started, including suspending the work of the relevant sorting ports, issuing an alarm to notify maintenance personnel, etc. While adjusting the fault response mechanism, the resources of the forward roundabout belt and the reverse roundabout belt will be re-allocated. For example, if the processing capacity of a forward roundabout belt sorting port in the first fault source set is greatly reduced due to a fault, and the corresponding reverse roundabout belt sorting port in the second fault source set is in a relatively normal working state and has a certain amount of idle processing capacity, then part of the tasks originally assigned to the faulty forward sorting port will be reallocated to the idle reverse sorting port. The purpose of this dynamic adjustment and mutual resource re-allocation is to minimize the impact of the fault on the entire logistics sorting system to ensure that items can continue to be sorted and circulated efficiently and accurately. For example, when a sensor at a forward sorting port fails and cannot accurately identify items, while the sensor at the nearby reverse sorting port works normally and has less processing pressure, the sorting task of some items will be transferred to this reverse sorting port, and technicians will be deployed to repair the sensor failure at the forward sorting port first. Through such flexible adjustments and resource optimization configuration, the stability and adaptability of the system can be improved, and the losses and delays caused by failures can be reduced.

[0053] Embodiment 2 Based on the same inventive concept as the intelligent sorting method for an endless cross belt in the aforementioned embodiment, Figure 2 As shown, the present application provides an intelligent sorting platform for a circular cross belt, and the platform and method embodiments in the present application embodiments are based on the same inventive concept. The platform includes: The item feature extraction module 10 uses the Internet of Things technology to scan the RFID tag to identify the basic information of the item, and combines the deep learning technology to use the image processing module to extract the item features, wherein the basic information of the item includes the destination information, and the item features include the shape features of the item and the size features of the item.

[0054] The grasping mode adjustment module 20 uses a 3D vision sensor to perform three-dimensional modeling of the object based on the characteristics of the object, and connects to a robotic arm to adaptively adjust the grasping mode, and the grasping mode is determined by the stability screening of the robotic arm.

[0055] A bidirectional distribution path matching module 30 is used to adjust the item distribution strategy in real time through a central control center, and adopt a reinforcement learning algorithm to dynamically match the bidirectional distribution path in combination with the basic information of the item. The bidirectional distribution path includes a forward distribution path connected to the forward circular cross belt feed port and a reverse distribution path connected to the reverse circular cross belt feed port.

[0056] An annular cross belt connection module 40, wherein the path starting point of the bidirectional distribution path is a switching guide platform, and the switching guide platform includes a first guide and a second guide, wherein the first slide groove corresponding to the first guide is connected to the forward rotating annular cross belt, and the second slide groove corresponding to the second guide is connected to the reverse rotating annular cross belt.

[0057] The intelligent item sorting module 50 implements a flow optimization strategy based on the bidirectional distribution path and integrates multimodal sensor data to perform intelligent item sorting.

[0058] Furthermore, the object feature extraction module 10 further includes: The item weight measurement unit is used to deploy a weight sensor to measure the item weight information in real time and associate and bind it with the RFID tag.

[0059] The object color feature extraction unit is used to extract the object color feature by using convolutional neural network technology.

[0060] An item sorting configuration unit is used to combine the basic information of the item with the color feature of the item and use a support vector machine classifier to initialize the item sorting configuration.

[0061] Furthermore, the item sorting configuration unit also includes: A key node positioning unit is used to locate key nodes on the forward circular cross belt and the reverse circular cross belt in combination with historical sorting instances, and the key nodes are deployed with edge computing servers.

[0062] A sorting decision unit is used to use a streaming data processing framework to interact with the support vector machine classifier and make sorting decisions.

[0063] Furthermore, the key node positioning unit also includes: A positive sorting exit sequence determining unit is used to determine a positive sorting exit sequence based on the positive rotating circular cross belt, wherein the positive sorting exit sequence includes M positive rotating circular cross belt sorting ports.

[0064] A reverse sorting exit sequence determining unit is used to determine a reverse sorting exit sequence based on the reverse roundabout belt, wherein the reverse sorting exit sequence includes N reverse roundabout belt sorting ports.

[0065] A dynamic management unit is used to connect to the logistics center, set up a temporary storage area, and dynamically manage the positive sorting export sequence and the reverse sorting export sequence with circulation efficiency as the optimization goal using the edge computing server.

[0066] Furthermore, the dynamic management unit also includes: The destination information matching unit is used to match the destination information in the basic information of the item through a temporary storage area with a location mark, and after the match is passed, the item circulation task is synchronously updated in the logistics center.

[0067] A circulation task priority determination unit, wherein the circulation task priority determination unit evaluates the urgency of the updated item circulation tasks based on the edge computing server and determines the priority of each item circulation task.

[0068] A parallel line scheduling management unit is used to perform parallel line scheduling management with the positive sorting exit sequence and the reverse sorting exit sequence based on the priority of each item flow task.

[0069] Furthermore, the parallel line scheduling management unit also includes: A sorting path traversal unit, wherein the sorting path traversal unit traverses multiple sorting paths based on the M forward-rotating round-robin sorting ports in the forward-sorting exit sequence and the N reverse-rotating round-robin sorting ports in the reverse-sorting exit sequence, using a sorting path decision formula.

[0070] The sorting path decision formula: , where j is the index of the sorting port, j∈{1,…,M+N}, is the priority of the jth sorting port, is the matching degree of the jth sorting port, is the turnover efficiency of the jth sorting port, is the operating cost of the jth sorting port, is the prediction model error, , , , It is the decision factor of each factor, which is used to adjust the importance of different factors in the sorting path decision process.

[0071] Furthermore, the sorting path traversal unit also includes: The first fault source set locating unit is used to use the IsolationForest algorithm to monitor in real time the priority of the first fault source set in the positive order egress sequence. A forward-rotating circular cross belt sorting port is used to locate the first fault source collection.

[0072] The second fault source set locating unit is used to use the IsolationForest algorithm to monitor in real time the priority of the first fault source in the reverse sorting egress sequence. A reverse circular cross belt sorting port is used to locate the second set of fault sources.

[0073] A fault dynamic adjustment unit, which dynamically adjusts the fault response mechanism based on the first fault source set and the second fault source set, and performs mutual resource reallocation of the forward ring cross belt and the reverse ring cross belt.

[0074] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0075] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0076] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. An intelligent sorting method for a circular cross belt, characterized in that: The method comprises: Using the Internet of Things technology, scanning the RFID tag to identify the basic information of the item, and combining with the deep learning technology, using the image processing module to extract the item features, wherein the basic information of the item includes the destination information, and the item features include the shape features and the size features of the item; Based on the features of the object, a 3D visual sensor is used to perform three-dimensional modeling of the object, and a robotic arm is connected to adaptively adjust the grasping mode, where the grasping mode is determined by the stability screening of the robotic arm; The central control center adjusts the item distribution strategy in real time, and adopts a reinforcement learning algorithm to dynamically match the bidirectional distribution path in combination with the basic information of the item, wherein the bidirectional distribution path includes a forward distribution path connected to the forward roundabout feeding port and a reverse distribution path connected to the reverse roundabout feeding port; The starting point of the bidirectional distribution path is a switching guide platform, and the switching guide platform includes a first guide and a second guide. The first chute corresponding to the first guide is connected to the forward annular cross belt, and the second chute corresponding to the second guide is connected to the reverse annular cross belt. Based on the bidirectional distribution path, a flow optimization strategy is implemented, and multimodal sensor data is integrated to perform intelligent sorting of items.

2. An intelligent sorting method for a circular cross belt as claimed in claim 1, characterized in that: The method comprises: Deploy a weight sensor to measure the weight information of the item in real time and associate it with the RFID tag; Use convolutional neural network technology to extract the color features of items; The basic information of the items is combined with the color features of the items, and a support vector machine classifier is used to initialize the item sorting configuration.

3. An intelligent sorting method for a circular cross belt as claimed in claim 2, characterized in that: Using a support vector machine classifier to initialize item sorting configuration, the method includes: On the forward-rotating circular cross belt and the reverse-rotating circular cross belt, key nodes are located in combination with historical sorting instances, and edge computing servers are deployed on the key nodes; A streaming data processing framework is used to interact with the support vector machine classifier and make sorting decisions.

4. An intelligent sorting method for a circular cross belt as claimed in claim 3, characterized in that: The key node deploys an edge computing server, and the method includes: Based on the forward-rotating circular cross belt, a forward-sorting exit sequence is determined, wherein the forward-sorting exit sequence includes M forward-rotating circular cross belt sorting exits; Based on the reverse roundabout, determine a reverse sorting exit sequence, wherein the reverse sorting exit sequence includes N reverse roundabout sorting exits; The logistics center is connected, a temporary storage area is set up, and the positive sorting export sequence and the reverse sorting export sequence are dynamically managed with the edge computing server with the circulation efficiency as the optimization goal.

5. An intelligent sorting method for a circular cross belt as claimed in claim 4, characterized in that: The forward-sorted export sequence and the reverse-sorted export sequence are dynamically managed using the edge computing server with the flow efficiency as the optimization goal, and the method includes: Matching the destination information in the basic information of the item with the temporary storage area where the location mark exists, and after the match is successful, synchronously updating the item circulation task in the logistics center; Based on the edge computing server, the urgency of the updated item circulation tasks is evaluated to determine the priority of each item circulation task; Based on the priority of each item circulation task, parallel scheduling management is performed with the positive sorting exit sequence and the reverse sorting exit sequence.

6. An intelligent sorting method for a circular cross belt as claimed in claim 5, characterized in that: Performing parallel scheduling management with the positively sorted export sequence and the reversely sorted export sequence, the method comprising: Based on the M forward roundabout sorting ports in the forward sorting exit sequence and the N reverse roundabout sorting ports in the reverse sorting exit sequence, a sorting path decision formula is used to traverse multiple sorting paths; The sorting path decision formula: , where j is the index of the sorting port, j∈{1,…,M+N}, is the priority of the jth sorting port, is the matching degree of the jth sorting port, is the turnover efficiency of the jth sorting port, is the operating cost of the jth sorting port, is the prediction model error, , , , It is the decision factor of each factor, which is used to adjust the importance of different factors in the sorting path decision process.

7. An intelligent sorting method for a circular cross belt as claimed in claim 6, characterized in that: The method comprises: Use the Isolation Forest algorithm to monitor in real time the priority of the positively sorted export sequence. A forward-rotating circular cross belt sorting port is used to locate the first fault source set; Use the Isolation Forest algorithm to monitor in real time the priority of the reverse sorting export sequence. A reverse roundabout with a sorting port to locate the second fault source set; Based on the first fault source set and the second fault source set, the fault response mechanism is dynamically adjusted, and the resources of the forward ring cross belt and the reverse ring cross belt are mutually reallocated.

8. An intelligent sorting platform for a circular cross belt, characterized in that: The intelligent sorting platform for the circular cross belt comprises: An item feature extraction module, which uses the Internet of Things technology to scan RFID tags to identify basic information of items, and combines deep learning technology to use an image processing module to extract item features, wherein the basic information of the item includes destination information, and the item features include item shape features and item size features; A grasping mode adjustment module, which uses a 3D visual sensor to perform three-dimensional modeling of the object based on the object characteristics, and connects to a robotic arm to adaptively adjust the grasping mode, wherein the grasping mode is determined by the stability screening of the robotic arm; A bidirectional distribution path matching module, the bidirectional distribution path matching module is used to adjust the item distribution strategy in real time through the central control center, and adopt a reinforcement learning algorithm to dynamically match the bidirectional distribution path in combination with the basic information of the item, the bidirectional distribution path includes a forward distribution path connected to the forward roundabout feeding port, and a reverse distribution path connected to the reverse roundabout feeding port; An annular cross belt connection module, wherein the path starting point of the bidirectional distribution path is a switching guide platform, the switching guide platform comprises a first guide and a second guide, a first chute corresponding to the first guide is connected to the forward annular cross belt, and a second chute corresponding to the second guide is connected to the reverse annular cross belt; An intelligent item sorting module implements a flow optimization strategy based on the bidirectional distribution path and integrates multimodal sensor data to perform intelligent item sorting.

Images