Logistics park complex operation scene nonlinear intelligent scheduling system and method
By using a combination of master-slave tracking panoramic cameras and artificial intelligence servers in the logistics park, track operators and operation status in real time and optimize nonlinear scheduling, the problem of low scheduling efficiency in complex operation scenarios in the logistics park is solved, and management and Execution efficiency.
Patent Information
- Application Number
- CN202411488748.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-24
- Publication Date
- 2024-11-22
AI Technical Summary
The prior art is difficult to track the operator's operating status, operation progress and operating site equipment occupation in complex operation scenarios in logistics parks in real time, resulting in low scheduling efficiency and management difficulties.
采用主从跟踪式全景摄像机、人工智能服务器和非线性调度服务器的组合,通过实时视频数据采集和分析,优化非线性调度决策,实现操作员的多目标跟踪和作业状态监控。
The digitalization and intelligence level of complex operation sites in the logistics park has been improved, the overall performance and management performance have been improved, and the problem of operator management and scheduling efficiency has been solved.
Smart Images

Figure CN119005646A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to an intelligent dispatching system, and more particularly to a nonlinear intelligent dispatching system for complex operation scenarios in a logistics park. Background Art
[0002] Logistics park is an economic functional area with industrial development nature, which relies on relevant logistics service facilities to reduce logistics costs, improve logistics operation efficiency, improve logistics service-related circulation processing, raw material procurement, and facilitate production activities directly connected with material sales places; as an urban logistics functional area, logistics park includes a variety of formats, such as logistics center, distribution center, transportation hub facilities, transportation organization and management center and logistics information center, to meet the logistics infrastructure needed for urban logistics management and operation; there are various materials to be transferred in the logistics park, so the packaging transfer form, loading and unloading method, transportation form and storage capacity also change with the different means of production, which is comprehensive; in addition, the comprehensiveness of the logistics park also determines that the logistics park has the characteristics of complexity, which is specifically reflected in the complexity of the internal structure and the complexity of the scheduling management strategy. The logistics park is also the connection point of various logistics operation modes, and various logistics operation modes require the use of a variety of logistics operation facilities. In order to meet the needs of logistics enterprises and third-party logistics service enterprises within different logistics parks, logistics operation facilities have the problem of efficient scheduling and safety control in time and space.
[0003] Although these digital, intelligent, and smart logistics parks have integrated advanced information and digital technologies such as big data, cloud computing, and artificial intelligence, the existing technologies are still relatively lacking in processing capabilities for complex operation scenarios in logistics parks in the real world, especially nonlinear and complex operation scenarios involving a large number of operators. These scenarios often have long operation execution times, many operation links, incomplete operation site data, poor real-time performance, and a large number of situations that require mechanical or manual handling. Although personal digital devices such as photoelectric scanners have become popular, they can maintain links with remote cloud servers through wireless communication networks at the operation site, and update the actual time, end time, and status of operations in real time, providing a data basis for optimizing the scheduling of complex operations. However, for complex operation scenarios, the actual operation status, operation progress, and occupation of equipment at the operation site of the operators at the operation site are often in a vague state, and the large number of nonlinear links such as nesting, interlacing, and waiting between complex operations create more complex technical problems. Intelligent scheduling systems often need to balance the efficiency of complex operation execution and operator satisfaction.
[0004] Existing digital technologies are still unable to solve such technical problems. The existing solution is to train employees and establish a multi-level management team on the job site to maintain tacit coordination in handling complex operations. It can be seen that the management of operators is the key to improving efficiency, but the existing intelligent scheduling system cannot achieve multi-target tracking of operators, and thus cannot grasp the operation status, operation progress, and equipment occupancy of complex operations in real time. Summary of the invention
[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a system composed of a master-slave tracking system, a panoramic camera and other equipment, and a multi-target tracking technology to track operators in real time, and to grasp the working status, working progress and equipment occupancy of complex operations in real time; at the same time, to provide an algorithm for nonlinear intelligent scheduling of complex operation scenarios based on these data.
[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a nonlinear intelligent scheduling system for complex operation scenarios in a logistics park, characterized in that it includes: A master-slave tracking panoramic camera is provided with multiple master-slave tracking panoramic cameras, which are connected via a local area network and are used to monitor complex operation sites in the logistics park and collect real-time video data of complex operation scenes before uploading; The artificial intelligence server is connected to the master-slave tracking panoramic camera to receive the video data uploaded by the master-slave tracking panoramic camera, identify and analyze the video data, and output the analysis results; A nonlinear scheduling server is connected to the artificial intelligence server to receive the analysis results output by the artificial intelligence server and optimize the nonlinear scheduling decision according to the analysis results; A personal digital terminal, which is provided with multiple personal digital terminals, corresponding to each operator one by one, and each personal digital terminal is connected to a nonlinear scheduling server to provide the operator with a portable operation data interaction function, and then receive operation assignment data or return cargo scan code and operation completion status data.
[0007] As a further improvement of the present invention, the master-slave tracking panoramic camera is a Hikvision starlight-level master-slave tracking panoramic camera, the artificial intelligence server uses the Hikvision AI open platform super brain, and the personal digital terminal uses the Xingtong XT8003ER model.
[0008] Another aspect of the present invention further provides a scheduling method, comprising the following steps: Step 1: The master-slave tracking panoramic camera tracks the operator's trajectory in real time; Step 2: The AI server identifies complex operation areas and operation progress; Step 3: The nonlinear scheduling server optimizes the nonlinear scheduling decision; Step 4: The operator receives and executes the nonlinear scheduling task through the personal digital terminal; Step 5: After completing the nonlinear scheduling task, the operator returns to the original operation process.
[0009] As a further improvement of the scheduling method, the specific steps of the master-slave tracking panoramic camera in step 1 to track the operator's trajectory in real time are as follows: Step 11, collecting video data through a master-slave tracking panoramic camera; Step 12, generating a training data set based on the video data collected in step 11; Step 13: Build a convolutional neural network model, perform parallel training using the training data set generated in step 12, and input the generated model parameters into the artificial intelligence server.
[0010] As a further improvement of the scheduling method, the specific steps of the artificial intelligence server identifying the complex operation area and operation progress in step 2 are as follows: Step 21, collecting video data through a master-slave tracking panoramic camera; Step 22, generating a real-time video data stream based on the data collected in step 21; Step 2 and 3, locate the operator's work area, and perform operator object segmentation and recognition; Step 24: Update and save the position data tracked by the operator in real time into the database.
[0011] As a further improvement of the scheduling method, the specific method of locating the operator's work area in steps two and three is: firstly, model training is performed to obtain a positioning model, and then complex work areas are identified based on the obtained model.
[0012] As a further improvement of the scheduling method, the specific steps of performing model training in steps 2 and 3 to obtain the positioning model are as follows: Step 1: Model training starts; Step 2: The master-slave tracking panoramic camera collects video data; Step 3: Perform image pixel operations on the video image; Step 4: Perform morphological opening operation; Step 5: Find contour edge detection; Step 6: Identify the minimum area circumscribed matrix; Step 7: Perform data expansion on the recognition data in the video image; Step 8: Generate the data file after marking the complex operation area; Step 9: Convert to TensorFlow format; Step 10: Configure hyperparameters for TensorFlow mode; Step 11: Input the model network for training; Step 12: Generate the pre-trained model file after training is completed; Step 13: Convert to protocol buffer file; Step 14: Called by TensorFlow inference classification algorithm; Step 15: Training is completed; Among them, steps 1 to 15 are trained 1000 times in a loop, and the recognition effect of the model is tested.
[0013] As a further improvement of the scheduling method, the specific steps for identifying the complex operation area in steps 2 and 3 are as follows: Step 16, the master-slave tracking panoramic camera collects video data; Step 17, inputting personal digital terminal interaction data; Step 18, input the data from steps 16 and 17 into the trained TensorFlow model; Step 19, output complex operation area data, complex operation progress and complex operation progress.
[0014] As a further improvement of the scheduling method, the specific steps of the nonlinear scheduling server optimizing the nonlinear scheduling decision in step 3 are as follows: Step 31, executing the original operation plan; Step 32, evaluate the cost of inserting a new job. If the cost is high, return to step 31 to execute the original job plan. If the cost is low, determine the order of the affected jobs and then proceed to the next step. Step 33, updating the equipment set required for the new job; Step 34: Encode the complex operation and operation area to obtain the configuration operation sorting code and the configuration area sorting code; Step 35: Start the ant colony genetic nonlinear optimization process. First, start the two-stage encoding of the population, update the job sorting code and the regional sorting code, and then enter the ant colony iteration loop. If the loop is not ended, perform the selection operation and execute two operations: crossover operation and mutation operation; The crossover operation includes single-point crossover of the operation code and adjustment according to the constraints; the mutation operation includes mutation of the operation code and mutation of the area code; the updated result is returned to the judgment of the end of the loop. If the judgment is that the loop is ended, the optimization solution process is entered and the next step is continued; Step 36: Input the genetic algorithm's optimal solution, convert the optimal solution into the ant colony algorithm to initialize the pheromone distribution matrix, initialize the ant colony algorithm, set NC=0, and apply the ant colony algorithm for optimization; update the ant colony algorithm pheromone distribution matrix; update the counter NC=NC+1; determine whether the solution has no obvious improvement? If the optimal solution has obvious improvement, return to the ant colony algorithm to optimize the next iteration; otherwise, the optimal solution for nonlinear scheduling has been obtained, and a new job scheduling plan is generated.
[0015] The beneficial effects of the present invention are as follows: The present invention aims to solve the problems of long operation cycle, high task complexity and low flexibility in dealing with nonlinear scheduling requirements such as job insertion in complex operation sites of logistics parks, and provides a solution using multiple master-slave tracking panoramic cameras. With the help of the AI open platform provided by a dedicated artificial intelligence server, the present invention integrates real-time tracking algorithms for operator trajectories, algorithms for identifying complex operation areas, operation progress and other status information, and nonlinear optimization scheduling decision algorithms, thereby solving the real-time tracking and status update of operators, equipment, areas and goods during the execution of complex operations in logistics parks, and providing technical means for nonlinear optimization decision-making, solving the problem that manual scheduling can only be carried out by the on-site management team under existing technical conditions, further improving the digitization and intelligence level of complex operation sites in logistics parks, and improving the overall performance and management performance of complex operations in logistics parks. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 A structural diagram of a nonlinear intelligent scheduling system for complex operation scenarios in a logistics park provided by the present invention; Figure 2 A flowchart of a nonlinear intelligent scheduling method for complex operation scenarios in a logistics park provided by the present invention; Figure 3 This is a schematic diagram of the linear scheduling process for complex operations in a logistics park; Figure 4 This is a schematic diagram of the nonlinear scheduling process for complex operations in a logistics park; Figure 5 for Figure 2 The flowchart of the algorithm for real-time tracking of operator trajectory by a master-slave tracking panoramic camera in the method flowchart shown; Figure 6 for Figure 2 The algorithm flow chart of the artificial intelligence dedicated server in the method flow chart shown in the figure identifies the complex operation area, operation progress and other status information; Figure 7 for Figure 2 The nonlinear scheduling server in the method flowchart shown is a flowchart of a nonlinear scheduling decision algorithm for optimization. DETAILED DESCRIPTION
[0017] The present invention will be further described below in detail with reference to the embodiments shown in the accompanying drawings.
[0018] Reference Figures 1 to 7 As shown, a nonlinear intelligent scheduling system for complex operation scenes in a logistics park of this embodiment includes: a plurality of master-slave tracking panoramic cameras 101, 102, 103, 104, 105, 106, a wireless network device 107 at the operation site, a nonlinear scheduling server 111, an artificial intelligence dedicated server 112, and personal digital terminals 114, 116, 118 with wireless communication functions carried by a plurality of operators 113, 115, 118; wherein the plurality of master-slave tracking panoramic cameras 101, 102, 103, 104, 105, 106 are connected together through a high-speed local area network, and are connected to the nonlinear scheduling server 111 and the artificial intelligence dedicated server 112, The monitoring area covers the complex operation site 100 in the logistics park, collects real-time video data of complex operation scenes 108, 109, and 110, and uploads them to the artificial intelligence dedicated server 112; personal digital terminals 114, 116, and 118 are connected to the wireless network equipment 107 at the operation site through a wireless LAN, and then connected to the nonlinear scheduling server 111, providing operators 113, 115, and 118 with portable operation data interaction functions, receiving operation assignment data, returned goods scanning, operation completion and other status data; complex operation scenes 108, 109, and 110 are respectively arranged in different areas of the complex operation site 100, and operators 113, 115, and 118 are respectively responsible for completing the operations.
[0019] As a specific implementation method of the improvement, the master-slave tracking panoramic cameras 101, 102, 103, 104, 105, and 106 are Hikvision's starlight-level master-slave tracking panoramic camera DS-2SC402P. The device consists of a network camera and an intelligent dome camera, and has image detection, recognition, and tracking functions. It has its own video analysis algorithm module and multi-target tracking algorithm module. In conjunction with the pan-tilt control system, it can realize automatic (or manual), fast, continuous, and smooth capture and tracking of 30 moving targets or designated targets within the panoramic screen area of the network camera; although the device provides functions such as outputting alarm signals for regional intrusion, entry into the area, and leaving the area of the moving target, which are designed for security monitoring needs, the complex operation scenes provided by the present invention are very The linear intelligent scheduling method can utilize these basic recognition and tracking algorithm modules for combined application, and combine with the artificial intelligence algorithm module provided by the artificial intelligence dedicated server 112 to perform real-time recognition and tracking of operators 113, 115, and 118 to complete the update of complex job status. The artificial intelligence dedicated server 113 of the present invention uses Hikvision AI open platform super brain, with 4 built-in GPU chips, supporting detection algorithms, classification algorithms, and supporting SDK development of mainstream artificial intelligence network models such as TensorFlow. The personal digital terminals 114, 116, and 118 of the present invention use the Xingtong XT8003ER model, which has a photoelectric scanning component, a back-clip structure design for easy carrying, an LCD display, and wireless communication modules such as Bluetooth, and has a long-distance communication capability of more than 50 meters.
[0020] like Figure 2 As shown, this embodiment provides a nonlinear intelligent scheduling method for complex operation scenarios, which specifically includes the following steps: Step S1: The master-slave tracking panoramic camera tracks the operator's trajectory 200 in real time; Step S2: The artificial intelligence dedicated server identifies the complex operation area, operation progress and other status information 201; Step S3: the nonlinear scheduling server optimizes the nonlinear scheduling decision 202; Step S4: The operator receives and executes the nonlinear scheduling task 203; Step S5: After completing the nonlinear scheduling task, the operator returns to the original operation process 204.
[0021] Figure 3 and Figure 4 Further explain the difference between linear scheduling and nonlinear scheduling of complex jobs; e.g. Figure 3As shown in the schematic diagram of the linear scheduling process of complex operations in the logistics park, there are three complex operations 301, 302, and 303 in the complex operation site 300, all of which use logistics pallets as operation units; the operator 304 uses the personal digital terminal 305 carried with him to connect to the nonlinear scheduling server 307 through the wireless device 306 at the operation site to receive the assignment of the operation task, and feedback the execution status of the operation task and confirm the completion of the operation and other information exchanges; the linear scheduling process of complex operations uses a single complex operation as the smallest operation unit. Only after completing the assigned complex operation, the operator 304 will perform the next complex operation, which has the advantage of easy management; therefore, the linear scheduling process includes the following steps: Step 1: The operator accepts the assignment of job 1 308; Step 2: The operator confirms that the job 1 309 is completed; Step 3: The operator accepts the assignment of job 2 310; Step 4: The operator confirms that the job 2 311 is completed; Step 5: The operator accepts the assignment 312 of job 3; Step 6: The operator confirms that job 3 313 is completed.
[0022] like Figure 4 As shown in the flowchart of nonlinear scheduling of complex operations in a logistics park, there are three complex operations 401, 402, and 403 in a complex operation site 400, all of which are based on logistics pallets as operation units; the operator 404 uses a personal digital terminal 405 that he carries with him to connect to a nonlinear scheduling server 408 through a wireless device 406 at the operation site to receive the assignment of the operation task, and to feedback the execution status of the operation task and confirm the completion of the operation and other information interactions; since the goods on the pallet may have different specifications such as shape, size, and weight, as well as multiple operations such as handling, disassembly, assembly, packaging, and film wrapping, the operation process is relatively complex and the execution cycle of a single complex operation is relatively long; during this period, if an emergency inserted operation task is encountered, the existing operation sequence cannot be adjusted; in order to improve the flexibility of complex operation scheduling, the traditional method is to strengthen operator training and improve management standards and levels; the present invention provides a method of identifying and tracking the operator 404 through a master-slave tracking panoramic camera 407 and an artificial intelligence dedicated server 409, and performing nonlinear scheduling, and the specific process includes the following steps: Step 1: The operator accepts the assignment of job 1 410; Step 2: The operator performs job 1 411; Step 3: If no new job is inserted, the operator continues to execute job 1 416; Step 4: The master-slave tracking panoramic camera 407 and the artificial intelligence dedicated server 409 perform target tracking to obtain the operation progress 412; Step 5: When receiving an urgent job insertion task, evaluate the cost of inserting a new job 413; Step 6: If the cost is high, the operator continues to perform job 1 416; otherwise, if the cost is low, the operator accepts the assignment of job 2 414, and after the operator confirms the completion of job 2 415, it returns to the operator to continue to perform job 1 416; Step 7: The operator confirms completion of job 1 417; Step 8: The operator accepts the assignment 418 of job 3; Step 9: The operator confirms that job 3 419 is completed.
[0023] like Figure 5 As shown, Figure 2 The flowchart of the algorithm for real-time tracking of operator trajectories by a master-slave tracking panoramic camera in the flowchart of the nonlinear intelligent scheduling method for complex operation scenarios in a logistics park provided by the present invention is shown. The algorithm of one embodiment of the present invention includes the following steps: It is divided into two modules: the big data platform parallel training convolutional neural network module and the Hikvision artificial intelligence AI open platform real-time operator recognition module; 1. Parallel training of convolutional neural network modules on the big data platform Step 1: The master-slave tracking panoramic camera collects video data 500; Step 2: Generate a training data set 506; Step 3: Design a convolutional neural network model 507; Step 4: parallel training 508; Step 5: Train and generate model parameters 509, which are provided to the real-time operator recognition module of the Hikvision artificial intelligence AI open platform; 2. Hikvision AI Open Platform Real-time Operator Recognition Module Step 1: The master-slave tracking panoramic camera collects video data 500; Step 2: Generate real-time video data stream 501; Step 3: Operator work area positioning 502; Step 4: Operator object segmentation 503; Step 5: Operator object identification 504; Step 6: The result is persisted to the database 505, that is, the position data tracked by the operator in real time is updated and saved in the database.
[0024] like Figure 6 As shown, Figure 2The flowchart of the algorithm for identifying complex operation areas, operation progress and other status information of the artificial intelligence dedicated server in the flowchart of the nonlinear intelligent scheduling method for complex operation scenarios in the logistics park provided by the present invention is shown. The algorithm of one embodiment of the present invention includes the following steps: The whole algorithm consists of two stages: model training stage and complex operation area identification stage; 1. Model training stage Step 1: Model training starts 600; Step 2: The master-slave tracking panoramic camera collects video data 601; Step 3: Perform image pixel operations 602 on the video image; Step 4: Perform morphological opening operation 603; Step 5: Find contour edge detection 604; Step 6: Identify the minimum area circumscribed matrix 605; Step 7: Perform data expansion 606 on the recognition data in the video image; Step 8: Generate a data file 607 of the complex operation area after annotation; Step 9: Perform TensorFlow format conversion 608; Step 10: Configure hyperparameters 609 for the TensorFlow mode; Step 11: input the model network for training 610; Step 12: Generate a pre-trained model file 611 after training is completed; Step 13: Convert to protocol buffer file 612; Step 14: The TensorFlow inference classification algorithm is called 613; Step 15: Training ends 614; Preferably, the model training phase is repeated 1000 times, and the recognition effect of the model is tested. The model obtained after training will be used in the complex operation area recognition phase; 2. Complex Operation Area Identification Stage Step 1: The master-slave tracking panoramic camera collects video data 615; Step 2: Input personal digital terminal interaction data 616; Step 3: Two data are input into the TensorFlow model 617; Step 4: Output complex operation area data 618; Step 5: Output complex operation progress 619; Step 6: Output complex job status 620.
[0025] like Figure 7 As shown, Figure 2The nonlinear scheduling server optimization nonlinear scheduling decision algorithm flow chart in the nonlinear intelligent scheduling method flow chart of the complex operation scene of the logistics park provided by the present invention is shown. One embodiment of the present invention adopts an ant colony genetic nonlinear optimization scheduling algorithm, including the following steps: Step 1: Execute the original operation plan 700; Step 2: Evaluate the cost of inserting a new job 701. If the cost is high, return to executing the original job plan 700. Otherwise, determine the order of the affected jobs 702. Step 3: Update the equipment set required for the new job 703; Step 4: Encode the complex operation and the operation area to obtain the configuration operation sort code 704 and the configuration area sort code 705; Step 5: Start the ant colony genetic nonlinear optimization process, first start the population two-stage coding 706, update the job sorting code 707 and the area sorting code 708; Step 6: Enter the ant colony iteration loop. If the loop 709 is not ended, perform a selection operation 717 and execute two operations: a crossover 718 operation and a mutation 721 operation. The crossover 718 operation includes a single-point crossover of the operation code 719 and an adjustment according to the constraint 720; and the mutation 721 operation includes a mutation of the operation code 722 and a mutation of the area code 723. The updated result is returned to the judgment of ending the loop 709. Step 7: If loop 709 is ended, the optimization solution process is entered; Step 8: Input the genetic algorithm optimal solution 710; Step 9: Convert the better solution into the ant colony algorithm to initialize the pheromone distribution matrix 711; Step 10: Initialize the ant colony algorithm and set NC=0 712; Step 11: Apply ant colony algorithm for optimization 713; Step 12: Update the ant colony algorithm pheromone distribution matrix 714; Step 13: Update counter NC=NC+1 715; Step 14: Determine whether the solution has no obvious improvement 716? If the better solution has obvious improvement, return to the next iteration of applying the ant colony algorithm for optimization 713; Step 15: Otherwise, the optimal solution for nonlinear scheduling has been obtained, and a new job scheduling solution 724 is generated.
[0026] To sum up, this embodiment provides a nonlinear intelligent scheduling system for complex operation scenarios in a logistics park and a scheduling method combined therewith. It adopts a master-slave tracking panoramic camera to realize real-time tracking of operators, and then realizes nonlinear scheduling through an artificial intelligence server and a nonlinear scheduling server. Finally, the final scheduling is realized through a personal digital terminal. In this way, multi-target tracking of operators can be realized, and the operation status, operation progress, and equipment occupancy of complex operations can be grasped in real time.
[0027] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.
Claims
1. A nonlinear intelligent scheduling system for complex operation scenarios in a logistics park, characterized by: include: A master-slave tracking panoramic camera is provided with multiple master-slave tracking panoramic cameras, which are connected via a local area network and are used to monitor complex operation sites in the logistics park and collect real-time video data of complex operation scenes before uploading; The artificial intelligence server is connected to the master-slave tracking panoramic camera to receive the video data uploaded by the master-slave tracking panoramic camera, identify and analyze the video data, and output the analysis results; A nonlinear scheduling server is connected to the artificial intelligence server to receive the analysis results output by the artificial intelligence server and optimize the nonlinear scheduling decision according to the analysis results; A personal digital terminal, which is provided with multiple personal digital terminals, corresponding to each operator one by one, and each personal digital terminal is connected to a nonlinear scheduling server to provide the operator with a portable operation data interaction function, and then receive operation assignment data or return cargo scan code and operation completion status data.
2. The nonlinear intelligent scheduling system for complex operation scenarios in logistics parks according to claim 1 is characterized by: The master-slave tracking panoramic camera is Hikvision's starlight-level master-slave tracking panoramic camera, the artificial intelligence server uses Hikvision's AI open platform super brain, and the personal digital terminal uses the Xingtong XT8003ER model.
3. A scheduling method using the nonlinear intelligent scheduling system for complex operation scenarios in a logistics park according to claim 1 or 2, characterized in that: The steps include: Step 1: The master-slave tracking panoramic camera tracks the operator's trajectory in real time; Step 2: The AI server identifies complex operation areas and operation progress; Step 3: The nonlinear scheduling server optimizes the nonlinear scheduling decision; Step 4: The operator receives and executes the nonlinear scheduling task through the personal digital terminal; Step 5: After completing the nonlinear scheduling task, the operator returns to the original operation process.
4. The scheduling method according to claim 3, characterized in that: The specific steps of the master-slave tracking panoramic camera tracking the operator's trajectory in real time in step 1 are as follows: Step 11, collecting video data through a master-slave tracking panoramic camera; Step 12, generating a training data set based on the video data collected in step 11; Step 13: Build a convolutional neural network model, perform parallel training using the training data set generated in step 12, and input the generated model parameters into the artificial intelligence server.
5. The scheduling method according to claim 4, characterized in that: The specific steps of the artificial intelligence server identifying the complex operation area and operation progress in step 2 are as follows: Step 21, collecting video data through a master-slave tracking panoramic camera; Step 22, generating a real-time video data stream based on the data collected in step 21; Step 2 and 3, locate the operator's work area, and perform operator object segmentation and recognition; Step 24: Update and save the position data tracked by the operator in real time into the database.
6. The scheduling method according to claim 5, characterized in that: The specific method of locating the operator's work area in steps two and three is: firstly, model training is performed to obtain a positioning model, and then complex work areas are identified based on the obtained model.
7. The scheduling method according to claim 6, characterized in that: The specific steps of performing model training to obtain the positioning model in steps 2 and 3 are as follows: Step 1: Model training starts; Step 2: The master-slave tracking panoramic camera collects video data; Step 3: Perform image pixel operations on the video image; Step 4: Perform morphological opening operation; Step 5: Find contour edge detection; Step 6: Identify the minimum area circumscribed matrix; Step 7: Perform data expansion on the recognition data in the video image; Step 8: Generate the data file after marking the complex operation area; Step 9: Convert to TensorFlow format; Step 10: Configure hyperparameters for TensorFlow mode; Step 11: Input the model network for training; Step 12: Generate the pre-trained model file after training is completed; Step 13: Convert to protocol buffer file; Step 14: Called by TensorFlow inference classification algorithm; Step 15: Training is completed; Among them, steps 1 to 15 are trained 1000 times in a loop, and the recognition effect of the model is tested.
8. The scheduling method according to claim 7, characterized in that: The specific steps for identifying complex operation areas in steps 2 and 3 are as follows: Step 16, the master-slave tracking panoramic camera collects video data; Step 17, inputting personal digital terminal interaction data; Step 18, input the data from steps 16 and 17 into the trained TensorFlow model; Step 19, output complex operation area data, complex operation progress and complex operation progress.
9. The scheduling method according to claim 8, characterized in that: The specific steps of the nonlinear scheduling server optimizing the nonlinear scheduling decision in step 3 are as follows: Step 31, executing the original operation plan; Step 32, evaluate the cost of inserting a new job. If the cost is high, return to step 31 to execute the original job plan. If the cost is low, determine the order of the affected jobs and then proceed to the next step. Step 33, updating the equipment set required for the new job; Step 34: Encode the complex operation and operation area to obtain the configuration operation sorting code and the configuration area sorting code; Step 35: Start the ant colony genetic nonlinear optimization process. First, start the two-stage encoding of the population, update the job sorting code and the regional sorting code, and then enter the ant colony iteration loop. If the loop is not ended, perform the selection operation and execute two operations: crossover operation and mutation operation; The crossover operation includes single-point crossover of the operation code and adjustment according to the constraints; the mutation operation includes mutation of the operation code and mutation of the area code; the updated result is returned to the judgment of the end of the loop. If the judgment is that the loop is ended, the optimization solution process is entered and the next step is continued; Step 36: Input the genetic algorithm's optimal solution, convert the optimal solution into the ant colony algorithm to initialize the pheromone distribution matrix, initialize the ant colony algorithm, set NC=0, and apply the ant colony algorithm for optimization; update the ant colony algorithm pheromone distribution matrix; update the counter NC=NC+1; determine whether the solution has no obvious improvement. If the optimal solution has obvious improvement, return to the ant colony algorithm to optimize the next iteration; otherwise, the optimal solution for nonlinear scheduling has been obtained, and a new job scheduling plan is generated.
Citation Information
Patent Citations
Distributed replacement flow shop scheduling method based on auxiliary target technology
CN116384639A
Digital twin modeling method and device for logistics park
CN117371208A
Monitoring apparatus and method for performing behavior recognition in complex environment
WO2024124970A1
