Dynamic node reserved A*path optimization method under wafer scheduling system

Abstract

The invention provides an A * path optimization method for dynamic node reservation under a wafer scheduling system, which comprises the following steps of: selecting a current node from an open table according to a manufacturing requirement, sequentially expanding a new node to the last node for the current node based on a preset A * algorithm, and forming an optimal scheduling path by all nodes according to an expansion sequence; and judging whether the evaluation function value of the expanded new node reaches a dynamic threshold value or not, and if so, storing the evaluation function value into an open table. According to the dynamic evaluation and node reservation method provided by the invention, the evaluation function values of all the nodes in the open table are recorded when each current node is explored, and the dynamic threshold value is set to determine node reservation, so that the heap overflow risk during large-scale scheduling processing is solved, the system stability is ensured, and the system calculation complexity is reduced. The method is more suitable for large-scale wafer manufacturing scheduling, efficient and stable path optimization and resource scheduling can be achieved in a high-complexity environment, and the application effect in an actual production scene is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of semiconductor wafer manufacturing, and in particular relates to an A* path optimization method for dynamic node reservation in a wafer scheduling system. Background Art

[0002] Semiconductor chip manufacturing is at the upstream of the semiconductor industry chain, and improving wafer processing efficiency is of great strategic significance. Wafer manufacturing is one of the most complex and advanced manufacturing industries. In the process of wafer manufacturing, a large number of semiconductor assembly equipment are used. A typical semiconductor assembly equipment usually consists of a wafer loading module (LoadPort, LP), a vacuum lock module (LoadLock, LL), several processing modules (Processing Modules, PMs), an alignment module (Aligner, AL), a cooling module (CoolingBuffer, CL), and a single / double-arm manipulator (Transfer Robot, TR). It is a highly automated equipment that is reconfigurable, highly flexible, has a short processing cycle, and high space utilization. It has been proven that assembly equipment scheduling control belongs to NP-Hard problems.

[0003] The production scheduling problem of wafer manufacturing assembly equipment is more complex than that of flexible manufacturing systems (FMS). It is due to its highly sophisticated manufacturing process, diversified products and processes, equipment and resource constraints, immediacy and real-time requirements, uncertainty and complex optimization problems, which require the comprehensive use of interdisciplinary engineering and management methods. In a typical scenario of wafer manufacturing, Petri nets intuitively display concurrency, deadlock and resource usage through graphics and mathematical language. By calculating and deducing the state of the system at different times, system scheduling, simulation and performance estimation are achieved. In recent years, the application of Petri nets to actual production scheduling has been one of the hot research issues internationally.

[0004] The A* algorithm is a classic heuristic search algorithm proposed by Peter Hart, Nils Nielsen and Bert Rafal in 1968. It is widely used in path planning and optimal path search. Its basic principle is to effectively guide the search process and quickly find the optimal path by evaluating the total cost of each node, including the actual cost from the starting point to the current node and the estimated cost from the current node to the target node.

[0005] In view of the above background, in the process of studying the production scheduling problem based on the Petri net model of wafer manufacturing, the introduction of the A* algorithm can optimize the operation control of the semiconductor wafer manufacturing system. The scheduling of the wafer manufacturing system needs to consider multiple factors, among which the residence time (Residence Time) and Q-time (Queue Time) are two key factors. These factors reflect the usage and queuing of system resources during the wafer processing process, so as to simulate the actual production environment more realistically. However, the consideration of residence time and Q-time makes the system more complicated, because it is necessary to accurately track the time limit and queue time of each process to avoid any potential production bottlenecks. This not only increases the computational burden of the A* algorithm when searching for paths, but also leads to an increase in the state search space and an increase in the complexity of the algorithm, making the application of the A* algorithm more difficult, and requiring higher efficiency in dealing with concurrency, time factors and system dynamics. Summary of the invention

[0006] In order to solve the above problems existing in the prior art, the present invention provides an A* path optimization method for dynamic node reservation in a wafer scheduling system. The technical problem to be solved by the present invention is achieved by the following technical solutions:

[0007] In a first aspect, the present invention provides an A* path optimization method for dynamic node reservation in a wafer scheduling system, comprising:

[0008] S1, receiving a plurality of manufacturing requirements for manufacturing a plurality of target wafers;

[0009] S2, select the current node from the open table according to the manufacturing demand, and expand the new nodes to the current node in sequence based on the preset A* algorithm until the last node, and form the optimal scheduling path according to the expansion order of all nodes;

[0010] Wherein, the open table stores a plurality of current nodes; in the process of expanding new nodes, the nodes are expanded in the order of priority, and the order of priority is the order from large to small of the evaluation function value;

[0011] S3, determining whether the evaluation function value of the expanded new node reaches the dynamic threshold, and if so, storing it as the current node in the open table.

[0012] In a second aspect, the present invention provides an A* path optimization device for dynamic node reservation in a wafer scheduling system, comprising:

[0013] A receiving module is configured to receive a plurality of manufacturing requirements for manufacturing a plurality of target wafers;

[0014] The optimization module is configured to select a current node from the open table according to the manufacturing demand, and expand new nodes to the current node in sequence until the last node based on the preset A* algorithm, and form an optimal scheduling path for all nodes in the expansion order;

[0015] Wherein, the open table stores a plurality of current nodes; in the process of expanding new nodes, the nodes are expanded in the order of priority, and the order of priority is the order from large to small of the evaluation function value;

[0016] The judgment module is configured to judge whether the evaluation function value of the expanded new node reaches the dynamic threshold, and if so, store it as the current node in the open table.

[0017] Beneficial effects:

[0018] The present invention provides an A* path optimization method for dynamic node reservation under a wafer scheduling system, including: receiving multiple manufacturing requirements for multiple target wafers; selecting a current node from an open table according to the manufacturing requirements, and sequentially expanding new nodes to the current node based on a preset A* algorithm until the last node, and forming an optimal scheduling path for all nodes in the order of expansion; expanding in priority order during the expansion of new nodes, the priority order being the order from large to small of the evaluation function value; judging whether the evaluation function value of the expanded new node reaches a dynamic threshold, and if so, storing it in the open table as the current node. The dynamic evaluation and node reservation method proposed by the present invention records the evaluation function values ​​of all nodes in the open table when exploring each current node, sets a dynamic threshold to determine the node reservation, thereby avoiding the problem of excessive memory usage, and further solves the risk of heap overflow when processing large-scale scheduling problems relative to the traditional A* algorithm, ensures system stability, and reduces the system calculation complexity. Compared with the scheduling method using a fixed evaluation function, the present invention is more suitable for large-scale wafer manufacturing scheduling, can achieve efficient and stable path optimization and resource scheduling in a highly complex environment, and improves the application effect in actual production scenarios.

[0019] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flow chart of an A* path optimization method for dynamic node reservation in a wafer scheduling system provided by the present invention;

[0021] Figure 2 It is a schematic diagram of the established Petri net model provided by the present invention;

[0022] Figure 3 It is a schematic diagram of the basic scheduling process provided by the present invention;

[0023] Figure 4 It is a schematic diagram of the node retention result provided by the present invention. DETAILED DESCRIPTION

[0024] The present invention is further described in detail below with reference to specific embodiments, but the embodiments of the present invention are not limited thereto.

[0025] The following is a brief introduction to the technical background, technical terms and technical concepts involved in the present invention.

[0026] New scenarios: The production scheduling problem of wafer manufacturing assembly equipment is more complex than that of flexible manufacturing systems (FMS). It is due to its highly sophisticated manufacturing process, diverse products and processes, equipment and resource constraints, immediacy and real-time requirements, uncertainty, and complex optimization problems. It requires the comprehensive use of interdisciplinary engineering and management methods. In a typical scenario of wafer manufacturing, Petri nets intuitively display concurrency, deadlock, and resource usage through graphics and mathematical language. By calculating and deducing the state of the system at different times, system scheduling, simulation, and performance estimation are achieved. In recent years, the application of Petri nets to actual production scheduling has been one of the hot research issues internationally.

[0027] More factors to consider: The scheduling of wafer manufacturing systems requires comprehensive consideration of multiple factors, among which residence time (Residence Time) and Q-time (Queue Time) are two key factors. These factors reflect the usage and queuing of system resources during wafer processing, so as to simulate the actual production environment more realistically. This not only increases the computational burden of the A* algorithm when searching for paths, but also requires the algorithm to be more efficient in handling concurrency, time factors and system dynamics. Therefore, in the process of designing the algorithm, the optimization strategy of path search must be comprehensively considered, and an effective mechanism must be introduced to handle concurrent events and time constraints in the system.

[0028] Previous studies usually focus on how to improve production efficiency and system response speed by optimizing Petri net structure and parameters. For example, Chen et al. proposed an energy-efficient scheduling framework based on Petri nets, combined with an improved scheduling algorithm and a practical case study, and finally verified the effectiveness of the proposed method by using a case study of an FMS producing different types of products; Zhao et al. proposed a method based on iterative greedy algorithm (IGA) to obtain a near-optimal solution. These methods use different neighborhood structures and combine variable neighborhood descent methods to find the optimal or near-optimal scheduling solution within an acceptable time.

[0029] Petri net: The Petri net model theory was first proposed by Carl Adam Petri in the 1960s. It is suitable for describing asynchronous and concurrent computer system models. The research field tends to believe that Petri net is the mother of all process definition languages. The classic Petri net is a simple process model composed of elements such as places, transitions, and directed arcs.

[0030] POPN: Process-Oriented Petri Net (POPN), this modeling method has strong modeling capabilities for automated manufacturing systems, can intuitively describe the flow of workpieces and the resource allocation and recycling process in production and processing, and is suitable for various manufacturing systems.

[0031] ROPN: (Resource-Oriented Petri Net, ROPN) can be used to describe the dynamic resource allocation process of various parts in the automated manufacturing system. Compared with ordinary Petri nets, resource-oriented Petri nets hide certain information about the processing process, and the network scale is smaller and the structure is compact. The ROPN model is a colored Petri net. There is a one-to-one mapping between the library in the ROPN model and the resources in the system. It can be used by different processes at different times. Therefore, it is necessary to distinguish different jobs and their processing processes in the system by coloring the labels. As the model evolves, the color of the label changes as the library moves, and the system is well modeled.

[0032] PTPN: Place Delay Petri Net; TTPN: Transition Delay Petri Net; Token: The small black dot in the place after Petri net modeling represents a token, which indicates the number of resources in the place. Pjob: Wafers of the same processing procedure are called a group of Pjobs. WPF: Wafer processing flow WPF (Wafer processing flow), for a group of Pjobs, its WPF = {m, p, s}, where m refers to the number of wafers processed by a group of Pjobs, p refers to the number of parallel PM processing chambers in the processing path, and s refers to the number of PM processing chambers in a parallel path. Introduction to each module of the combined equipment: The combined equipment includes a wafer loading module (LoadPort, LP), a vacuum lock module (LoadLock, LL), several processing modules (Processing Modules, PMs), an alignment module (Aligner, AL), a cooling module (CoolingBuffer, CL), and a single / double-arm manipulator (Transfer Robot, TR).

[0033] The main purpose of the present invention is to provide an A* path optimization method with dynamic node reservation under a wafer scheduling system. The research idea adopted is based on the integrated application research of Petri nets and other methods. This technical solution not only improves the production efficiency and equipment utilization of the wafer manufacturing system, but also optimizes the overall efficiency of the production process, ensures stable operation of the system and realizes efficient scheduling and path optimization in complex environments.

[0034] The specific implementation details of the solution of the present invention are described in detail below.

[0035] like Figure 1 As shown, the present invention provides an A* path optimization method for dynamic node reservation in a wafer scheduling system, including:

[0036] S1, receiving a plurality of manufacturing requirements for manufacturing a plurality of target wafers;

[0037] S2, select the current node from the open table according to the manufacturing demand, and expand the new nodes to the current node in sequence based on the preset A* algorithm until the last node, and form the optimal scheduling path according to the expansion order of all nodes;

[0038] Wherein, the open table stores a plurality of current nodes; in the process of expanding new nodes, the nodes are expanded in the order of priority, and the order of priority is the order from large to small of the evaluation function value;

[0039] When a new node is generated, not only its heuristic function value and actual cost are calculated, but also the evaluation value of the current node is compared to calculate the evaluation function difference. This difference reflects the priority of the new node relative to the current node in path optimization.

[0040] When calculating the scheduling under the target number of wafers, each time a new node is expanded, the difference in the evaluation function between the current node and the newly generated node needs to be calculated to evaluate its relative priority. Specifically, when a new node is generated, not only its heuristic function value and actual cost are calculated, but also compared with the evaluation value of the current node to calculate the difference in the evaluation function. In this way, the priority of the new node in the open table can be dynamically adjusted, thereby guiding the search process more effectively and ensuring that high-priority nodes are explored and processed earlier. This difference evaluation mechanism can help the algorithm select key nodes more accurately in complex scheduling problems, improve the efficiency and quality of the overall scheduling, and avoid irrelevant nodes occupying valuable computing resources.

[0041] S3, determining whether the evaluation function value of the expanded new node reaches the dynamic threshold, and if so, storing it as the current node in the open table.

[0042] The dynamic threshold is dynamically adjusted according to the distance between the current node and the end point. When the current node is close to the end point, the dynamic threshold is smaller, and vice versa.

[0043] The dynamic threshold is adjusted according to the distance of the node from the end point. The closer to the end point, the smaller the threshold, thereby reducing the number of retained nodes. The core idea of ​​this mechanism is that when the node is close to the end point, the path selection space has been greatly reduced, and there is no need to retain too many nodes for redundant search, thereby effectively saving memory resources. Through this dynamic threshold adjustment method, the algorithm can manage memory usage more intelligently, reduce unnecessary node retention, ensure that high-priority nodes are processed in a timely manner, and improve the overall scheduling efficiency and path optimization quality. The core idea of ​​this mechanism is that when the node is close to the end point, the path selection space has been greatly reduced, and there is no need to retain too many nodes for redundant search, thereby effectively saving memory resources. Through this dynamic threshold adjustment method, the algorithm can manage memory usage more intelligently, reduce unnecessary node retention, ensure that high-priority nodes are processed in a timely manner, and improve the overall scheduling efficiency and path optimization quality. This dynamic evaluation mechanism can not only effectively solve the heap overflow problem, but also optimize the performance and stability of the algorithm while ensuring the solution quality.

[0044] Of course, the present invention can also dynamically adjust the threshold based on statistical methods. By recording the distribution of node evaluation values, a suitable threshold is determined using statistical methods to ensure that the number of nodes retained at different stages achieves an optimized effect. Although the computational complexity of this method may be slightly higher, it can further improve the stability and performance of the algorithm in some cases.

[0045] In an optional implementation manner of the present invention, before S1, the A* path optimization method for dynamic node reservation in the wafer scheduling system further includes:

[0046] S00, obtaining various equipment modules that need to be loaded in the wafer manufacturing process, and establishing a process-oriented Petri net model of the combined equipment according to the scheduling order of the various equipment modules in the wafer manufacturing process;

[0047] The various equipment modules include a wafer loading module, a single-arm or dual-arm robot, a processing module, a vacuum lock module and a cooling module; wherein, in the wafer loading module, the wafers to be processed are represented by a number of tokens in the wafer loading module library; the number of available resources of the single-arm or dual-arm robot is represented by a number of tokens in the robot library; the upper capacity limit of the processing module, the vacuum lock module and the cooling module is determined by the number of tokens in the corresponding capacity library; the transitions are divided into three categories: robot taking wafers, robot putting wafers, and robot exchanging wafers.

[0048] The wafers of the same set of processing steps are called a group of Pjobs, and the processing steps can be represented by the wafer processing flow WPF (Wafer processing flow). For a group of Pjobs, its WPF = {m, p, s}, where m refers to the number of wafers processed by a group of Pjobs, p refers to the number of parallel PM processing chambers in the processing path, and s refers to the number of PM processing chambers in a parallel path. According to the different WPFs of wafer processing paths, a process-oriented modeling Petri model N = (S, T; F, W, M_0) is established, where S and T are finite non-empty and non-intersecting sets, S is a set of places, T is a set of transitions, F is a directed arc from a place to a transition or a transition to a place, W represents the weight of the directed arc, and M represents the number of tokens in the place.

[0049] This step can establish a delay Petri model of the dual-arm manipulator combination device. The delay Petri model can be represented by a triple (N, D, Q), where: N = (S, T; F, W, M_0) is an ordinary Petri net, D = {d_1, d_2, ..., d_n} is the set of all library delay times, and Q = {q_1, q_2, ..., q_n} is the set of all transition delay times.

[0050] See also Figure 2 , taking WPF={8,2,1} as an example, the wafer processing route is:

[0051] CH1→CH5→CH1→CH5

[0052] According to the actual scenario description, the flow of wafers between devices must be transferred by robots. Therefore, the actual processing route should be described as:

[0053] LP→step1→(LA2 / LB2)→step2→CH1_1→step3→CHAB_1→step4→CH5_1→step5→CHAB_2→step6→CH1_2→step7→CHAB_3→step8→CH5_2→step9→CHAB_4→step10→(LA1 / LB1)→step11→LPcomp.

[0054] Among them, LP is the storage device, LA2 / LB2 is the feed port from atmosphere to vacuum end in the vacuum lock module, CHAB is the transfer device for the interaction between vacuum chamber 1 and vacuum chamber 2, and LA1 / LB1 is the discharge port from vacuum to atmosphere end in the vacuum lock module. Wafers are taken out and put in the above equipment, and are transferred by robots in the middle. Step 1-11 means that the wafer is in the corresponding robot. In addition, although some equipment will be reused in the processing route, since the completion of the processing is increased unidirectionally, the processing steps corresponding to the same equipment are different in the processing route, and are named by (equipment name_number of times the equipment has appeared in the processing flow). The processing recipe can be described as a Petri net model of a simple sequential process: the weight of all arcs is 1, and the wafer flows unidirectionally in the library, where the library is called the process library.

[0055] S01, designing an A* algorithm based on the Petri net model, and determining the most basic scheduling path for manufacturing a single wafer by executing a calculation process of the A* algorithm; the most basic scheduling path includes a plurality of nodes and a plurality of transitions corresponding to the nodes;

[0056] In an optional implementation manner of the present invention, the calculation process of the A* algorithm includes:

[0057] a, input open table with initial identification;

[0058] b. When the open list is not empty, select the node with the highest priority in the open list as the current identifier;

[0059] c. If the current identifier is the termination identifier, all identifiers in the current path table are returned;

[0060] d. If the current identifier is not a termination identifier, the current identifier is deleted from the open table and the current identifier is added to the closed table;

[0061] e. Traverse all possible changes triggered by the current node to expand the identification. The identification number increases by 1 for each change.

[0062] f, determine whether the identifier of the current node is in the closed table, and if so, select the identifier corresponding to the next node;

[0063] g, if the identifier of the current node is not in the closed table, set the identifier as the current identifier, and add the current identifier and related information to the current path table and the open table;

[0064] h, add the current identifier to the open table.

[0065] See the A* algorithm pseudocode:

[0066]

[0067] refer to Figure 3 , Figure 3 The A* algorithm is used to construct and execute the scheduling process of a single wafer for the scheduling steps under a specific model, by accurately calculating the path cost and evaluation function value, the evaluation function value includes the actual cost from the starting point to the current node and the heuristic function value; the path cost refers to the actual cost from the starting node to the current node n, expressed as:

[0068]

[0069] Where, marking.getPrefix() is the global time of the current node;

[0070] The heuristic function value refers to the indicator used in the A* algorithm to determine the priority of the exploration node, which can be expressed as:

[0071] h = max(h,curPlaceDepth);

[0072] Where max(h,curPlaceDepth) represents the maximum value of the current heuristic function value h and the place depth of the current node; curPlaceDepth is the place depth of the current node, expressed as:

[0073]

[0074] placeDepth is the place depth for evaluating the function value, and maxDepth is the maximum place depth.

[0075] The evaluation function value is expressed as:

[0076] F=g+h.

[0077] It is worth noting that: designing the A* algorithm based on the above Petri net model and developing the most basic scheduling path for a single wafer can ensure the accuracy and effectiveness of the path search;

[0078] When the A* algorithm handles large-scale problems, the heap overflow problem is mainly caused by the need to maintain a large number of open list and closed list nodes. In complex semiconductor wafer manufacturing systems, scheduling and path planning involve a huge search space, which leads to a sharp increase in the number of nodes and excessive memory usage, which in turn causes heap overflow problems.

[0079] S02, calculating the place depth of each node on the most basic scheduling path, and using the place depth to calculate the evaluation function value of the node, and recording the evaluation function value into a character string.

[0080] In response to the heap overflow problem mentioned above, the scheduling problem is divided into three stages. The first stage is to calculate the scheduling under a single wafer, and when exploring each current node, record the evaluation function value of all nodes in the current open table. In addition, the library depth and global time of each current node need to be recorded. This information will be used as an important basis for judging whether a node should be retained in the subsequent evaluation process, ensuring sufficient data support when selecting key nodes, thereby optimizing the overall scheduling efficiency and path quality.

[0081] In an optional embodiment of the present invention, S2 includes:

[0082] S21, determining a manufacturing order of a plurality of target wafers according to the manufacturing requirements;

[0083] S22, selecting a current wafer in sequence according to the manufacturing order of the multiple target wafers, and selecting a current node from the open table for the target wafer;

[0084] S23, using a preset A* algorithm to expand a new node to the current node;

[0085] S24, using the evaluation function value of the new node in the character string to determine the priority of the new node in the current wafer manufacturing process; if the priority of the new node is higher than the priority of other wafers in the manufacturing process, the new node is used as the next current node;

[0086] S25, repeat S23 until the last node, and combine all nodes into the optimal scheduling path for the current wafer.

[0087] In an optional embodiment of the present invention, S23 includes:

[0088] S231, using a preset A* algorithm to expand a new node from the current node, and using the evaluation function value of the new node in the character string, determining a first evaluation function difference between the new node and the current node in the current wafer manufacturing process and a second evaluation function difference between the new node and the current node in other wafer manufacturing processes;

[0089] It is worth noting that when calculating the scheduling under the target number of wafers, each time a new node is expanded, the evaluation function difference between the current node and the newly generated node needs to be calculated to evaluate its relative priority. Specifically, when a new node is generated, not only its heuristic function value and actual cost are calculated, but also compared with the evaluation value of the current node to calculate the evaluation function difference. This difference reflects the priority of the new node relative to the current node in path optimization.

[0090] S232, determine whether the difference of the first evaluation function is greater than the difference of the second evaluation function. If so, determine that the priority of the new node in the current wafer manufacturing process is higher than the priority in other wafer manufacturing processes. If not, determine that the priority of the new node in the current wafer manufacturing process is not higher than the priority in other wafer manufacturing processes.

[0091] In a specific implementation of the present invention, after S23, the A* path optimization method for dynamic node reservation in the wafer scheduling system further includes:

[0092] If the priority of the new node is lower than the priority of other wafers in the manufacturing process, the new node is used as the next current node of other wafers;

[0093] For the current wafer, a new node with a lower priority is selected as the next current node.

[0094] See also Figure 4 As shown, Figure 4 The figure shows the node retention results of the A* path optimization method for dynamic node retention under the wafer scheduling system provided by the present invention, wherein Nodes to Explore represents the number of nodes to be explored in the open table corresponding to each node, and the overall trend is a curve-like increase; the Nodes Retained curve represents the number of nodes retained by each node after dynamic evaluation, and the overall trend is flat; the Percentage curve represents the ratio of the number of retained nodes corresponding to each node to the total number of nodes to be explored, and the overall trend is downward, which is consistent with what was mentioned before. With the exploration of nodes, the proportion of the number of retained nodes is getting lower and lower.

[0095] The present invention provides an A* path optimization device for dynamic node reservation in a wafer scheduling system, comprising:

[0096] A receiving module is configured to receive a plurality of manufacturing requirements for manufacturing a plurality of target wafers;

[0097] The optimization module is configured to select a current node from the open table according to the manufacturing demand, and expand new nodes to the current node in sequence until the last node based on the preset A* algorithm, and form an optimal scheduling path for all nodes in the expansion order;

[0098] Wherein, the open table stores a plurality of current nodes; in the process of expanding new nodes, the nodes are expanded in the order of priority, and the order of priority is the order from large to small of the evaluation function value;

[0099] The judgment module is configured to judge whether the evaluation function value of the expanded new node reaches the dynamic threshold, and if so, store it as the current node in the open table.

[0100] It is worth noting that the terms "first" and "second" in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0101] Although the present application is described herein in conjunction with various embodiments, in the process of implementing the claimed application, those skilled in the art may understand and implement other variations of the disclosed embodiments by reviewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality of components or steps.

[0102] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A dynamic node reservation A* path optimization method under a wafer scheduling system, characterized in that: include: S1, receiving a plurality of manufacturing requirements for manufacturing a plurality of target wafers; S2, select the current node from the open table according to the manufacturing demand, and expand the new nodes to the current node in sequence based on the preset A* algorithm until the last node, and form the optimal scheduling path according to the expansion order of all nodes; Wherein, the open table stores a plurality of current nodes; in the process of expanding new nodes, the nodes are expanded in the order of priority, and the order of priority is the order from large to small of the evaluation function value; S3, determining whether the evaluation function value of the expanded new node reaches the dynamic threshold, and if so, storing it as the current node in the open table.

2. The A* path optimization method for dynamic node reservation in the wafer scheduling system according to claim 1, characterized in that: Before S1, the A* path optimization method for dynamic node reservation under the wafer scheduling system also includes: S00, obtaining various equipment modules that need to be loaded in the wafer manufacturing process, and establishing a process-oriented Petri net model of the combined equipment according to the scheduling order of the various equipment modules in the wafer manufacturing process; S01, designing an A* algorithm based on the Petri net model, and determining the most basic scheduling path for manufacturing a single wafer by executing a calculation process of the A* algorithm; the most basic scheduling path includes a plurality of nodes and a plurality of transitions corresponding to the nodes; S02, calculating the place depth of each node on the most basic scheduling path, and using the place depth to calculate the evaluation function value of the node, and recording the evaluation function value into a character string.

3. The A* path optimization method for dynamic node reservation in the wafer scheduling system according to claim 2, characterized in that: The equipment modules include a wafer loading module, a single or double-arm robot, a processing module, a vacuum lock module and a cooling module; Among them, in the wafer loading module, the wafers that need to be processed are represented by several tokens in the wafer loading module library; the number of available resources of single-arm or dual-arm robots is represented by several tokens in the robot library; the upper capacity of the processing module, vacuum lock module and cooling module is determined by the number of tokens in the corresponding capacity library; the changes are divided into three categories: robot taking wafers, robot putting wafers, and robot exchanging wafers.

4. The A* path optimization method for dynamic node reservation in a wafer scheduling system according to claim 2, characterized in that S2 include: S21, determining a manufacturing order of a plurality of target wafers according to the manufacturing requirements; S22, selecting a current wafer in sequence according to the manufacturing order of the multiple target wafers, and selecting a current node from the open table for the target wafer; S23, using a preset A* algorithm to expand a new node to the current node; S24, using the evaluation function value of the new node in the character string to determine the priority of the new node in the current wafer manufacturing process; if the priority of the new node is higher than the priority of other wafers in the manufacturing process, the new node is used as the next current node; S25, repeat S23 until the last node, and combine all nodes into the optimal scheduling path for the current wafer.

5. The A* path optimization method for dynamic node reservation in a wafer scheduling system according to claim 4, characterized in that: S23 includes: S231, using a preset A* algorithm to expand a new node from the current node, and using the evaluation function value of the new node in the character string, determining a first evaluation function difference between the new node and the current node in the current wafer manufacturing process and a second evaluation function difference between the new node and the current node in other wafer manufacturing processes; S232, determine whether the difference of the first evaluation function is greater than the difference of the second evaluation function. If so, determine that the priority of the new node in the current wafer manufacturing process is higher than the priority in other wafer manufacturing processes. If not, determine that the priority of the new node in the current wafer manufacturing process is not higher than the priority in other wafer manufacturing processes.

6. The A* path optimization method for dynamic node reservation in a wafer scheduling system according to claim 5, characterized in that: The evaluation function value includes the actual cost from the starting point to the current node and the heuristic function value; the actual cost is expressed as: Where, marking.getPrefix() is the global time of the current node; The heuristic function value is expressed as: h = max(h,curPlaceDepth); Where max(h,curPlaceDepth) represents the maximum value of the current heuristic function value h and the place depth of the current node, and curPlaceDepth is the place depth of the current node, which is expressed as: Where placeDepth is the place depth of the evaluation function value, and maxDepth is the maximum place depth; The evaluation function value is expressed as: F=g+h.

7. The A* path optimization method for dynamic node reservation in a wafer scheduling system according to claim 4, characterized in that: After S23, the A* path optimization method for dynamic node reservation in the wafer scheduling system further includes: If the priority of the new node is lower than the priority of other wafers in the manufacturing process, the new node is used as the next current node of other wafers; For the current wafer, a new node with a lower priority is selected as the next current node.

8. The A* path optimization method for dynamic node reservation in a wafer scheduling system according to claim 1, characterized in that: The dynamic threshold is dynamically adjusted according to the distance between the current node and the end point. When the current node is close to the end point, the dynamic threshold is smaller, and vice versa.

9. The A* path optimization method for dynamic node reservation in a wafer scheduling system according to claim 1 or 2, characterized in that: The calculation process of the A* algorithm includes: a, input open table with initial identification; b. When the open list is not empty, select the node with the highest priority in the open list as the current identifier; c. If the current identifier is the termination identifier, all identifiers in the current path table are returned; d. If the current identifier is not a termination identifier, the current identifier is deleted from the open table and the current identifier is added to the closed table; e. Traverse all possible changes triggered by the current node to expand the identification. The identification number increases by 1 for each change. f, determine whether the identifier of the current node is in the closed table, and if so, select the identifier corresponding to the next node; g, if the identifier of the current node is not in the closed table, set the identifier as the current identifier, and add the current identifier and related information to the current path table and the open table; h, add the current identifier to the open table.

10. An A* path optimization device for dynamic node reservation in a wafer scheduling system, characterized in that: include: A receiving module is configured to receive a plurality of manufacturing requirements for manufacturing a plurality of target wafers; The optimization module is configured to select a current node from the open table according to the manufacturing demand, and expand new nodes to the current node in sequence until the last node based on the preset A* algorithm, and form an optimal scheduling path for all nodes in the expansion order; Wherein, the open table stores a plurality of current nodes; in the process of expanding new nodes, the nodes are expanded in the order of priority, and the order of priority is the order from large to small of the evaluation function value; The judgment module is configured to judge whether the evaluation function value of the expanded new node reaches the dynamic threshold, and if so, store it as the current node in the open table.

Images