An e-book borrowing management system based on OPAC two-way connection

Abstract

This invention provides an e-book lending management system based on OPAC bidirectional linkage, belonging to the field of library information technology. It includes: a hardware acceleration layer that constructs a dynamic resource distribution matrix and calculates the matrix norm in real time using an FPGA coprocessor, combined with an ASIC anti-collision controller to implement request queue scanning every 100ms and a redundant copy generation algorithm; a system service layer that deploys a dynamic priority scheduling engine, achieving intelligent resource allocation based on cross-campus collaboration coefficients and edge-cloud collaboration strategies, while recording borrowing operations and performing automated copyright revenue sharing audits through blockchain notarization services; and a user interaction layer that provides a bidirectional association search interface and a priority borrowing channel based on credit scoring. The system improves resource scheduling efficiency in high-concurrency scenarios through hardware-accelerated matrix norm calculation, an elastic copy allocation mechanism, and an anti-collision algorithm embedded in ASICs. Combined with geolocation scoring and blockchain technology, it achieves cross-campus resource collaboration and accurate allocation of copyright revenue.

Description

Technical Field

[0001] This invention relates to the field of library information technology, and in particular to an e-book borrowing management system based on OPAC bidirectional connection. Background Technology

[0002] In the field of e-book lending in libraries, cross-campus resource collaborative management and accurate copyright sharing are the core challenges. Traditional systems rely on manual scheduling and paper-based ledgers, which suffer from high response latency and frequent borrowing conflicts. Especially in scenarios with concurrent requests from multiple campuses, resource scheduling efficiency decreases, and the error rate of manually calculated copyright sharing is generally high.

[0003] In existing technologies, some solutions attempt to optimize resource allocation through software algorithms, but are limited by pure CPU computing architectures and cannot handle large-scale matrix operations in real time, resulting in a lag in dynamic block granularity adjustment. Other solutions use a single blockchain for evidence storage, which improves data credibility, but does not coordinate with hardware acceleration modules, making it difficult to meet the timeliness requirements in high-concurrency scenarios. In addition, the fragmented management of metadata for paper and e-books further restricts system performance and user experience. Summary of the Invention

[0004] This invention provides an e-book lending management system based on OPAC bidirectional linkage to solve the problems of cross-campus resource collaborative management and accurate copyright sharing in the prior art.

[0005] This invention provides an e-book lending management system based on OPAC bidirectional linkage, comprising:

[0006] The hardware acceleration layer includes an FPGAOPAC coprocessor for performing the following operations:

[0007] During the resource initialization phase, the OPAC metadata database is scanned to construct an initial resource distribution matrix. The resource distribution matrix has a dimension of m×n, where m is the number of campuses and n is the number of resource blocks. Each matrix element represents the number of replicas of the j-th resource block in the i-th campus, and the matrix norm is calculated in real-time by a matrix norm calculation unit. ;

[0008] An ASIC anti-collision controller is used to: scan the request queue every 100ms and calculate the collision probability during the collision resolution phase. And generate redundant replicas according to the solidified formula. , Resource conflict probability, when the real-time conflict probability When the value is greater than 0.2, the generation of the redundant copy is triggered. Base number of replicas;

[0009] During the post-audit phase, the utilization rate of redundant copies is verified. ;

[0010] The system service layer includes a dynamic priority scheduling engine that quantifies system load into system load metrics based on system load. And based on the system load index Dynamically adjust the priority weight parameters used for request scheduling , , For user historical behavior metrics, , Reinforcement learning coefficient, System load metrics are limited through a reinforcement learning model. ;

[0011] Based on the cross-campus collaboration coefficient Allocate resource copies, Location rating Cross-campus resource request volume , Elasticity adjustment factor A baseline geographic threshold is used, and the traffic splitting ratio is dynamically adjusted through an edge-cloud collaborative strategy. ;

[0012] The blockchain-based evidence storage service generates hash values ​​and writes them to the Hyperledger Fabric consortium blockchain when a user borrows a book, and performs copyright revenue sharing deviation audits every day at midnight.

[0013] Preferably, the step of using the cross-campus collaboration coefficient... Allocating resource copies involves the following steps:

[0014] Step 1: Score based on chapter independence and cross-campus collaboration coefficient Calculate the initial block granularity ;

[0015] Step 2, when the resource distribution matrix When the Frobenius norm changes, the tanh function is calculated using the FPGA's DSP array, and the block granularity is dynamically adjusted. , , Matrix Frobenius norm, Rank of a matrix;

[0016] Step 3: Update the resource distribution matrix using an exponential decay strategy, with a decay rate of... It is controlled by the PLL clock divider of the FPGA.

[0017] Preferably, the copyright revenue sharing method implemented by the blockchain-based evidence storage service includes the following steps:

[0018] Based on user reading time and the proportion of copyright holders Dynamically calculate profit sharing ;

[0019] Parameters are stored using the tamper-proof fuse memory of the FPGA. ,implement ,in, The revenue share is based on a pre-set benchmark.

[0020] Daily comparison of actual revenue sharing Compared with theoretical value If the deviation exceeds the preset fixed deviation value, the account will be frozen and the administrator will be notified. ,in, The theoretical value is a preset fixed deviation value. Based on the aforementioned revenue sharing The actual division is calculated by cumulative calculation. This refers to the amount of copyright revenue sharing already executed.

[0021] Preferably, the strategy executed by the ASIC anti-collision controller includes:

[0022] When system load At that time, the elastic coefficient is dynamically adjusted according to the linear formula. Limit the adjustment step size ;

[0023] By combining the elasticity coefficient and the request queue length, a virtual queuing time is generated, with the error controlled by a high-precision clock ±2ppm.

[0024] Preferably, the edge-cloud collaboration method executed by the dynamic priority scheduling engine includes:

[0025] Based on resource popularity and location rating If the score exceeds the preset geographic location score threshold, the edge node will be pre-cached; otherwise, it will be discarded.

[0026] Adjust the cloud traffic splitting ratio by adjusting load parameters. Ensure edge load When it exceeds 75%, the diversion rate rises rapidly.

[0027] Preferably, the hardware acceleration layer includes a dynamic verification mechanism, including:

[0028] Calculate performance degradation metrics every 24 hours. If it exceeds 5%, FPGA recalibration will be triggered;

[0029] Verify a parameter value embedded in the ASIC to determine if it is equal to a preset value ε, where the preset value ε = 10⁻ 5 ±10⁻ 7 The preset value ε is fixed in the address range of 0x3FF0x401 of the ASIC mask ROM to ensure that the parameter cannot be tampered with.

[0030] Preferably, the system further includes a user interaction layer: an OPAC search interface that supports bidirectional association of metadata between paper books and e-books, and indicates the campuses where books can be borrowed and copyright restrictions; the user interaction layer also provides a credit scoring-based... The priority borrowing channel is scored by weighting historical on-time rate (60%), device reliability (30%), and sharing contribution (10%).

[0031] Preferably, the system further includes an anomaly detection and self-healing module, configured as follows:

[0032] The topology mapping latency of the FPGAOPAC coprocessor is monitored in real time. When the latency is detected to be greater than 1.2ms for three consecutive times, the redundancy replica multiplication mechanism of the ASIC anti-collision controller is triggered.

[0033] The generation interval of the blockchain evidence number of the blockchain evidence service is periodically checked. If the interval is detected to be greater than 30 seconds, the system automatically switches to the backup smart contract instance.

[0034] When the reinforcement learning model parameters of the dynamic priority scheduling engine or When the value exceeds the preset range, it will be forcibly reset to the initial value and the model will be retrained.

[0035] This invention significantly improves resource scheduling efficiency in high-concurrency scenarios and effectively reduces the probability of resource conflicts by working in tandem with a hardware-accelerated matrix operation module and a dynamic elastic block partitioning strategy. Based on a geographically sensitive cross-campus collaborative model, it achieves intelligent matching and load balancing of resources across multiple regions, greatly optimizing storage space utilization. Combining blockchain notarization technology and smart contract auditing mechanisms ensures the transparency and immutability of the copyright sharing process. Furthermore, a bidirectional association retrieval architecture breaks down the barriers between print and e-book metadata, providing users with a seamless cross-media retrieval experience. The system as a whole achieves breakthrough improvements in resource management accuracy, real-time response, and operational compliance, building a secure and efficient e-book management ecosystem for multi-campus libraries. Attached Figure Description

[0036] Figure 1 This is a system architecture diagram of an e-book borrowing management system based on OPAC bidirectional connection according to the present invention;

[0037] Figure 2This is a flowchart of the dynamic block reassembly process in this invention;

[0038] Figure 3 This is a sequence diagram of blockchain evidence storage in this invention. Detailed Implementation

[0039] This invention relates to an e-book lending management system based on OPAC bidirectional linkage, aiming to solve the problems of cross-campus resource collaborative management and accurate copyright sharing. The overall approach is as follows: employing a hardware acceleration layer, a system service layer, and a user interaction layer, combined with OPAC metadata and blockchain technology, to achieve efficient scheduling and copyright management of e-book resources. The overall architecture is as follows: Figure 1 As shown, the core processes include: Resource topology mapping: FPGA parses OPAC metadata to generate a dynamic resource network; Conflict prediction and redundancy allocation: ASIC calculates and distributes redundant copies in real time; Cross-campus collaborative scheduling: Dynamically adjusts the resource distribution matrix; Blockchain notarization and profit sharing: Records operation logs and executes smart contracts; User interaction and credit management: Provides priority borrowing channels and personalized services.

[0040] The above technical solutions will now be described in detail with reference to the accompanying drawings and specific embodiments to provide a better understanding of them. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments used only to explain the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. Furthermore, it should be noted that, for ease of description, only the parts related to the present invention are shown in the drawings, not all of them.

[0041] Example 1, as Figure 1-3 As shown: The hardware acceleration layer includes: an FPGAOPAC coprocessor, used to perform the following operations:

[0042] During the resource initialization phase, the OPAC metadata database is scanned to construct an initial resource distribution matrix. And calculate in real time through the matrix norm calculation unit. ;

[0043] ASIC anti-collision controller, used for:

[0044] During the conflict resolution phase, the request queue is scanned every 100ms to calculate the conflict probability. And generate redundant replicas according to the solidified formula. , Resource conflict probability, The base number of copies; in the post-audit phase, verify the validity of redundant copies. Otherwise, an alarm will be triggered;

[0045] The strategies implemented by the ASIC anti-collision controller include:

[0046] When system load At that time, the elastic coefficient is dynamically adjusted according to the linear formula. Limit the adjustment step size ;

[0047] By combining the elasticity coefficient and the request queue length, a virtual queuing time is generated, with the error controlled by a high-precision clock ±2ppm.

[0048] The edge-cloud collaboration methods executed by the dynamic priority scheduling engine include:

[0049] Based on system load, the system load is quantified into system load metrics. And based on the system load index Dynamically adjust the priority weight parameters used for request scheduling , , For user historical behavior metrics, , Reinforcement learning coefficient, System load metrics are limited through a reinforcement learning model. ;

[0050] Based on resource popularity and location rating If the score exceeds the preset geographic location score threshold, the edge node will be pre-cached; otherwise, it will be discarded.

[0051] Adjust the cloud traffic splitting ratio by adjusting load parameters. Ensure edge load When it exceeds 75%, the diversion rate rises rapidly.

[0052] The hardware acceleration layer includes dynamic verification mechanisms, including:

[0053] Calculate performance degradation metrics every 24 hours. If it exceeds 5%, FPGA recalibration will be triggered;

[0054] In ASIC verification It is fixed in the address range of 0x3FF0x401 of the ASIC mask ROM to ensure that the parameters cannot be tampered with.

[0055] Specifically, the hardware acceleration layer FPGAOPAC coprocessor:

[0056] Obtain MARC21 format metadata from the Library Integrated Management System (ILS), including: basic book information (ISBN, title, author); chapter structure (XML format, with chapter independence scores). Copyright license status (including available campuses and maximum concurrent users).

[0057] Convert OPAC metadata into a resource distribution matrix The matrix dimension is (m = number of campuses, n = number of resource blocks);

[0058] Matrix elements This represents the number of replicas of the j-th resource block in the i-th campus, with an initial value of [value missing]. =1.

[0059] The FPGA has a built-in matrix norm calculation unit that performs the following:

[0060]

[0061] in The matrix singular values ​​are calculated every 5 ns; the norm values ​​are output to the dynamic priority scheduling engine.

[0062] Pipeline optimization technology is used to process 256 bits of data stream per clock cycle; measured mapping latency is 0.8ms (mean) ± 0.05ms (standard deviation).

[0063] Furthermore, the ASIC anti-collision controller redundant replica generation process involves scanning the edge node request queue every 100ms to calculate the request density. ;when The time is marked as a high-conflict area.

[0064] Exponential smoothing forecasts based on historical data:

[0065]

[0066] like >0.3, triggers the generation of redundant replicas.

[0067] The formula for embedding in ASIC mask ROM: The basic replica number Nbase can be set to 100, and can be further defined by adjusting the base replica number Nbase.

[0068] The calculation result is rounded up to generate an integer number of replicas; the generated replicas are distributed to edge nodes according to the following distribution strategy: if the campus type ==="teaching area", then the allocation ratio is 70% to prioritize teaching needs; otherwise, the allocation ratio is 30%.

[0069] Redundant replica utilization is calculated every 5 minutes. ;like If the number of replicas is less than 0.7, an alarm will be triggered and the system will roll back to the baseline replica count. .

[0070] Furthermore, the system service layer includes a dynamic priority scheduling engine, which schedules tasks based on cross-campus collaboration coefficients. Allocate resource copies, Location rating Cross-campus resource request volume , Elasticity adjustment factor A baseline geographic threshold is used, and the traffic splitting ratio is dynamically adjusted through an edge-cloud collaborative strategy. ;

[0071] The blockchain-based evidence storage service generates hash values ​​and writes them to the Hyperledger Fabric consortium blockchain when a user borrows a book, and performs copyright revenue sharing deviation audits every day at midnight.

[0072] Cross-campus collaboration coefficient The method for allocating resource replicas includes the following steps:

[0073] Step 1: Score based on chapter independence and cross-campus collaboration coefficient Calculate the initial block granularity ;

[0074] Step 2, when the resource distribution matrix When the Frobenius norm changes, the tanh function is calculated using the FPGA's DSP array, and the block granularity is dynamically adjusted. , , Matrix Frobenius norm, Rank of a matrix;

[0075] Step 3: Update the resource distribution matrix using an exponential decay strategy, with a decay rate of... It is controlled by the PLL clock divider of the FPGA.

[0076] The copyright revenue sharing method implemented by the blockchain-based evidence storage service includes the following process:

[0077] Based on user reading time and the proportion of copyright holders Dynamically calculate profit sharing ;

[0078] Parameters are stored using the tamper-proof fuse memory of the FPGA. ,implement ;

[0079] Daily comparison of actual revenue sharing Compared with theoretical value If the deviation exceeds the preset fixed deviation value, the account will be frozen and the administrator will be notified. ,in, The theoretical value is a preset fixed deviation value. Based on the aforementioned revenue sharing The actual division is calculated by cumulative calculation. This refers to the amount of copyright revenue sharing already executed.

[0080] Specifically, the system service layer dynamic resource scheduling engine's cross-campus collaboration strategy:

[0081] Count the number of requests per hour for each campus ; Calculate the distance attenuation factor:

[0082]

[0083] Example: When the distance difference between Campus A and the main campus is 5km, =0.606.

[0084] Synergy coefficient generation:

[0085]

[0086] like >1.2, marked as a high-demand campus.

[0087] Resource reallocation: from low-demand campuses Allocate at least 20% of the copies to high-demand campuses; update the resource distribution matrix. The norm is then recalculated using an FPGA.

[0088] Edge-cloud collaboration: User-initiated borrowing request: When a user needs to read or obtain resources, they will initiate a borrowing request to the edge node.

[0089] Edge node query for local resource copies: After receiving a request, the edge node will first query whether the relevant resource copy is already stored locally.

[0090] Determine if there are enough replicas: If there are enough replicas, the edge node will notify the FPGA to schedule resource allocation and return a reading link to the user.

[0091] If there are insufficient replicas: the edge node will forward the request to the cloud scheduling center, which will then forward it to the ASIC to calculate and generate redundant replicas. The new replicas will then be returned to the cloud scheduling center, which will then distribute them to the edge nodes. Finally, the edge nodes will return the reading link to the user.

[0092] Key parameter: Edge node response threshold Diversion ratio Calculation error ≤ 3%.

[0093] The blockchain-based evidence storage service's evidence storage process is as follows:

[0094] Hash generation: When a user initiates a borrowing request, the system combines the following fields to generate the original data:

[0095] json

[0096] {

[0097] "user_id":"U2023001",

[0098] "book_id":"B9787111636665",

[0099] "timestamp":"20231005T14:23:18Z",

[0100] "location":"Campus X Library 3F"

[0101] }

[0102] Perform SHA3256 hash operation: (Raw data)

[0103] On-chain evidence storage: Call the HyperledgerFabric smart contract `recordBorrow`, with the hash value H as the input parameter; the blockchain returns the evidence storage number (e.g., `DCI20231005142318`) and records it in the local database.

[0104] Copyright revenue sharing trigger: When a user's cumulative reading time reaches 30 minutes, the system automatically executes the revenue sharing formula:

[0105] The revenue sharing results are written to the blockchain and distributed to the author's wallet address.

[0106] Deviation Audit: Automatic audit is performed daily at 2 AM, calculating:

[0107] like Freeze the relevant accounts and notify the administrator. It is a preset fixed deviation value.

[0108] Furthermore, the user interaction layer includes an OPAC search interface that supports bidirectional association of metadata between print and ebooks, indicating available campuses and copyright restrictions; and a credit scoring system. The priority borrowing channel is scored by weighting historical on-time rate (60%), device reliability (30%), and sharing contribution (10%).

[0109] Specifically, the OPAC retrieval at the user interaction layer enhances the association between print and digital resources:

[0110] Search results display: On the traditional OPAC results page, an "Electronic Version" tag is added to entries containing ebooks; clicking the tag will take you to the ebook reading interface, which supports EPUB / PDF format rendering.

[0111] Metadata synchronization: Print books and e-books achieve metadata mapping through ISBN, including: author, publisher, publication year; chapter table of contents, and reference index.

[0112] Copyright Notice: If the e-book has borrowing restrictions (such as only available at specific campuses), the following message will be displayed: "This resource is currently available for borrowing at the following campuses: Campus A and Campus B".

[0113] Credit-based borrowing channels include:

[0114] Credit score calculation: Historical on-time delivery rate, weighted at 60% ;

[0115] Example: User returns on time 15 times / total borrowing 20 times → on-time rate 75% → score 45 points (75% × 60).

[0116] Device trustworthiness, weighted at 30%: Registered devices + IP whitelist: 30 points; Unregistered devices: 9 points (30% × 30).

[0117] Historical contribution, weighted at 10%: ;

[0118] Example: Shared 5 times → Contribution 1.0 → Score 1 point.

[0119] Total score calculation: , Enter the virtual queue.

[0120] In one application case, high-concurrency request handling during exam week: Campus X experiences an explosion of requests during the "Advanced Mathematics" exam week, peaking at 1200 requests / minute, exhausting local replicas.

[0121] Processing procedure:

[0122] 1. Load detection: FPGA detects... Fluctuations exceeding 15% are marked as abnormal areas.

[0123] Real-time request density This triggered a high-conflict warning.

[0124] 2. Redundant copy generation: ASIC computing ,generate:

[0125] 150 copies were transferred from campuses Y and Z to campus X.

[0126] 3. Resource distribution: Edge nodes distribute resources according to the traffic distribution ratio. Some requests are redirected to the cloud; the resource block granularity is dynamically adjusted to 2MB (originally 5MB) to improve parallel processing capabilities.

[0127] Results: Average response time improved; conflict rate decreased by 7%.

[0128] It should be understood that the embodiments disclosed in this invention and the above description enable those skilled in the art to implement this invention. However, this invention is not limited to the embodiments mentioned above. It should be understood that those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this invention, and should all be included within the protection scope of this invention.

Claims

1. An e-book lending management system based on OPAC bidirectional connection, characterized in that, include: The hardware acceleration layer includes an FPGAOPAC coprocessor for performing the following operations: During the resource initialization phase, the OPAC metadata database is scanned to construct an initial resource distribution matrix. The resource distribution matrix has a dimension of m×n, where m is the number of campuses and n is the number of resource blocks. Each matrix element represents the number of replicas of the j-th resource block in the i-th campus, and the matrix norm is calculated in real-time by a matrix norm calculation unit. ; An ASIC anti-collision controller is used to: scan the request queue every 100ms and calculate the collision probability during the collision resolution phase. And generate redundant replicas according to the solidified formula. When the real-time conflict probability When the value is greater than 0.2, the generation of the redundant copy is triggered. The base number of replicas; During the post-audit phase, the utilization rate of redundant copies is verified. ; The system service layer includes a dynamic priority scheduling engine that quantifies system load into system load metrics based on system load. And based on the system load index Dynamically adjust the priority weight parameters used for request scheduling , , For user historical behavior metrics, , Reinforcement learning coefficient, To determine system load metrics, a reinforcement learning model is used to constrain [the load]. ; Based on the cross-campus collaboration coefficient Allocate resource copies, , Rate the location For cross-campus resource requests, As an elasticity adjustment factor, The baseline geographic threshold is used, and the traffic splitting ratio is dynamically adjusted through an edge-cloud collaborative strategy. ; The blockchain-based evidence storage service generates hash values ​​and writes them to the Hyperledger Fabric consortium blockchain when a user borrows a book, and performs copyright revenue sharing deviation audits every day at midnight.

2. The e-book lending management system based on OPAC bidirectional connection according to claim 1, characterized in that, The basis for cross-campus collaboration coefficient Allocating resource copies involves the following steps: Step 1: Score based on chapter independence and cross-campus collaboration coefficient Calculate the initial block granularity ; Step 2, when the resource distribution matrix When the Frobenius norm changes, the tanh function is calculated using the FPGA's DSP array, and the block granularity is dynamically adjusted. , , Matrix Frobenius norm, Rank of a matrix; Step 3: Update the resource distribution matrix using an exponential decay strategy, with a decay rate of... It is controlled by the PLL clock divider of the FPGA.

3. The e-book lending management system based on OPAC bidirectional connection according to claim 1, characterized in that, The strategies executed by the ASIC anti-collision controller include: When system load At that time, the elastic adjustment factor is dynamically adjusted according to a linear formula. Limit the adjustment step size ; By combining the elastic adjustment factor and the request queue length, a virtual queuing time is generated, with the error controlled by a high-precision clock of ±2ppm.

4. The e-book lending management system based on OPAC bidirectional connection according to claim 1, characterized in that, The edge-cloud collaboration method executed by the dynamic priority scheduling engine includes: Based on resource popularity The geographical location being in a high-density area necessitates pre-caching of edge nodes; By load parameters Adjust cloud traffic distribution ratio Ensure edge load When it exceeds 75%, the diversion rate rises rapidly.

5. The e-book lending management system based on OPAC bidirectional connection according to claim 1, characterized in that, The hardware acceleration layer includes a dynamic verification mechanism, including: Calculate performance degradation metrics every 24 hours. If it exceeds 5%, FPGA recalibration will be triggered; Verify a parameter value embedded in the ASIC to determine if it is equal to a preset value ε, wherein the preset value... The parameter values ​​are fixed in the address range of 0x3FF0x401 of the ASIC mask ROM to ensure that the parameters cannot be tampered with.

6. The e-book lending management system based on OPAC bidirectional connection according to claim 1, characterized in that, The system further includes a user interaction layer: an OPAC search interface that supports bidirectional association of metadata between paper books and e-books, and indicates available campuses and copyright restrictions; the user interaction layer also provides a credit scoring-based... The priority borrowing channel is scored by weighting historical on-time rate (60%), device reliability (30%), and sharing contribution (10%).

7. The e-book lending management system based on OPAC bidirectional connection according to claim 1, characterized in that, The system further includes an anomaly detection and self-healing module, configured as follows: The topology mapping latency of the FPGAOPAC coprocessor is monitored in real time. When the latency is detected to be greater than 1.2ms for three consecutive times, the redundancy replica multiplication mechanism of the ASIC anti-collision controller is triggered. The generation interval of the blockchain evidence number of the blockchain evidence service is periodically checked. If the interval is detected to be greater than 30 seconds, the system automatically switches to the backup smart contract instance. When the reinforcement learning model parameters of the dynamic priority scheduling engine or When the value exceeds the preset range, it will be forcibly reset to the initial value and the model will be retrained.

Images