RPA distributed system and concurrent execution method of RPA distributed system

Abstract

The present disclosure belongs to the technical field of automation, and particularly relates to an RPA distributed system and an RPA distributed system concurrent execution method. The system comprises a central control module and a plurality of execution nodes. The execution nodes are configured to receive a task environment baseline of the RPA distributed system, wherein the task environment baseline is generated by the central control module based on login information of a target account in a target operating system; receive an atomic task in a task queue of the RPA distributed system; after obtaining a globally unique distributed session lock in the RPA distributed system through competition, update a local environment according to the task environment baseline, execute the held atomic task in the updated local environment, and release the distributed session lock after the task ends. The present disclosure not only realizes session sharing, but also forcibly guarantees the consistency of different execution node environments, effectively avoiding the risk control detection of the target system.

Description

Technical Field

[0001] This disclosure belongs to the field of automation technology, and specifically relates to an RPA distributed system and a method for concurrent execution of an RPA distributed system. Background Technology

[0002] With the advancement of enterprise digital transformation, Robotic Process Automation (RPA) is widely used to handle repetitive business processes. In many actual RPA scenarios in enterprises (especially in finance and government), the target business system often only provides a single operating account due to security policies or license authorization restrictions, and does not support multiple simultaneous logins of the same account (i.e., the later login will kick out the previous login, or the system will forcibly block concurrent sessions).

[0003] When dealing with this type of "single account" scenario, existing technologies mainly adopt the following two solutions: 1. Single-instance serial execution: Only one RPA robot is deployed, and tasks are executed sequentially. This approach is inefficient when dealing with massive amounts of tasks and cannot meet business timeliness requirements.

[0004] 2. Simple Session Sharing: Multiple bot instances share the same login cookie or token. However, this method has serious drawbacks: (1) Concurrency conflict: Since the target system usually maintains a stateful session context, when multiple robots operate on different pages concurrently, it will cause page jump conflict, data overwriting and process interruption.

[0005] (2) Environmental fingerprint risk control vulnerability: Existing technologies often overlook the consistency of client-side environmental fingerprints. When different RPA execution nodes reuse the same session, if the browser kernel version, User-Agent, screen resolution, and other characteristics of each node are inconsistent, the existing technology, although it achieves cookie sharing, ignores the consistency of deep environmental characteristics such as Canvas fingerprints and WebGL rendering features. This can lead to the account being easily marked as abnormal by the target system after long-term operation, resulting in frequent CAPTCHA triggering, or even being added to the blacklist and forcibly interrupting the session, which seriously affects the automation efficiency.

[0006] (3) Insufficient fault tolerance: When a session expires and needs to be re-logged in, if the target system has enabled multi-factor authentication (MFA) or CAPTCHA pop-up, the traditional automated script will freeze directly, causing the entire system to stop, lacking a human-machine collaborative recovery mechanism.

[0007] Therefore, there is an urgent need for an RPA execution solution that can achieve distributed concurrent processing of multiple robots while strictly limiting single account login, ensuring consistent environmental fingerprints, completely avoiding state conflicts, and having automatic fault tolerance capabilities in MFA scenarios. Summary of the Invention

[0008] To address the aforementioned issues, this disclosure provides an RPA distributed system with the advantage of conflict-free concurrent execution by a single account. The system includes: The central control module and several execution nodes; among them, The execution node is configured as follows: Receive the task environment baseline of the RPA distributed system; the task environment baseline is generated by the central control module based on the login information of the target account on the target operating system; Receive atomic tasks from the task queue of the RPA distributed system; After acquiring the globally unique distributed session lock in the RPA distributed system through competition, the local environment is updated according to the task environment baseline. The held atomic task is executed in the updated local environment, and the distributed session lock is released after the task is completed.

[0009] Furthermore, the task environment baseline includes: session state data generated by the target account logging into the target operating system, and the network fingerprint configuration of the target operating system.

[0010] Furthermore, the execution node is configured to initialize the operation page corresponding to the current session state data before executing the held atomic task in the updated local environment to clear the preceding operations of other execution nodes.

[0011] Furthermore, the execution node is configured to: when executing the held atomic task in the updated local environment, if the session state data becomes invalid, configure a global system pause lock and send an invalidation signal to the central control module.

[0012] Furthermore, the RPA distributed system also includes: The self-healing collaboration module is configured as follows: After the global system is paused, the central control module takes over control of the target account and the target operating system, and attempts to log in again with AI assistance. If the AI-assisted re-login fails, send a re-login request with human assistance. After the target account logs back in, update the task environment baseline and remove the global system pause lock.

[0013] This disclosure also proposes a concurrent execution method for RPA distributed systems based on the above-mentioned RPA distributed system, including: Log in to the target operating system using the target account to generate a task environment baseline; Based on the task environment baseline, update the local environment of the target execution node; wherein, the target execution node is an execution node that has obtained an atomic task from the task queue and has acquired a distributed session lock; the distributed session lock is globally unique in the distributed system; The target execution node, after updating the local environment, executes the atomic task held by the target execution node and releases the distributed session lock after the task is completed.

[0014] Furthermore, when executing the atomic task held by the target execution node after updating the local environment, if the session state data becomes invalid, the login status of the target account on the target operating system is updated based on AI takeover and / or manual collaboration, and the task environment baseline is updated.

[0015] This disclosure also proposes an electronic device, including a memory and a processor, wherein the memory stores a computer program or instructions, and when the computer program or instructions are executed by the processor, they are used to at least implement the above-described RPA distributed system concurrent execution method.

[0016] This disclosure also proposes a computer-readable storage medium storing a computer program or instructions, which, when executed by a processor, are at least used to implement the above-described RPA distributed system concurrent execution method.

[0017] This disclosure also proposes a computer program product stored in a computer-readable storage medium, which, when executed by a processor, is used to at least implement the above-described RPA distributed system concurrent execution method.

[0018] Compared with the prior art, this disclosure has the following advantages: it not only realizes session sharing, but also forces the consistency of different execution node environments, effectively avoids the risk control detection of the target system, reduces the risk of account banning, and ensures that each task is executed in a clean and predictable page state by using session lock in conjunction with task pre-state reset, thus guaranteeing the idempotency of the operation.

[0019] Other features and advantages of this disclosure will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the disclosure. The objects and other advantages of this disclosure may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 An RPA distributed system according to an embodiment is shown; Figure 2 The concurrent execution flow of an RPA distributed system according to an embodiment is illustrated; Figure 3 This illustrates a concurrent execution method based on RPA distributed systems; Figure 4 An electronic device according to an embodiment is shown; Figure 5 A computer-readable storage medium according to an embodiment is shown. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0023] Figure 1 The RPA distributed system proposed in this disclosure includes: The central control module and several execution nodes; among them, The execution node is configured as follows: Receive the task environment baseline of the RPA distributed system; the task environment baseline is generated by the central control module based on the login information of the target account on the target operating system; Receive atomic tasks from the task queue of the RPA distributed system; After acquiring the globally unique distributed session lock in the RPA distributed system through competition, the local environment is updated according to the task environment baseline. The held atomic task is executed in the updated local environment, and the distributed session lock is released after the task is completed.

[0024] Furthermore, the task environment baseline includes: The session state data generated by the target account logging into the target operating system, and the network fingerprint configuration of the target operating system.

[0025] According to some embodiments of this disclosure, the task environment baseline can be "dynamically rotated." The central control module can connect to a cloud fingerprint database and periodically (e.g., daily) pull a set of the latest fingerprint configurations (including updated Chrome version numbers, different graphics card rendering features, etc.) and update them to the distributed shared storage area. The RPA execution node automatically applies the new fingerprints in the next round of tasks. This alternative can simulate the software upgrade behavior of real users, further reducing the risk of being identified by risk control measures during long-term operation.

[0026] According to some embodiments of this disclosure, the RPA distributed system proposed in this disclosure includes a central control module (Controller), a distributed coordination module (Coordinator), and RPA execution modules (Workers). The central control module is responsible for the overall scheduling and signaling routing of the system, and preferably adopts a high availability (HA) deployment. The distributed coordination module is used to store session data, fingerprint configurations, task queues, and provide distributed lock services. The RPA execution modules, i.e., execution nodes, are used to execute the following schemes: Task Acquisition: Retrieve an atomic task from the task queue; if the queue is empty, enter a waiting state. Resource contention: Operations compete for a globally unique distributed session lock by acquiring an atomic lock; State synchronization and environment masquerading: Load session state data and standard browser fingerprint configuration from the shared storage area, and reshape the environment characteristics of the local web automation engine based on the fingerprint configuration; the session state data and standard browser fingerprint configuration are generated by the central control module using a single account login to the target system. Atomic task execution: The specific business logic of executing atomic tasks during the lock holding period; Lock release: Release the distributed session lock after the task is completed or after exception handling.

[0027] Furthermore, the central control module is configured as follows: After acquiring the globally unique distributed session lock in the RPA distributed system through competition, the target account accesses the heartbeat page of the target system.

[0028] According to some embodiments of this disclosure, the central control module (Controller) can also be configured in an "active keep-alive" mode. That is, the central control module maintains a scheduled task (e.g., every 10 minutes) that, after acquiring a distributed session lock, actively controls the browser to access the target system's "heartbeat page" (such as the user center or homepage) to refresh the server-side session's validity period. This approach can significantly reduce the frequency of self-healing processes triggered by natural session expiration, improving the overall continuity of the system.

[0029] Furthermore, the execution node is configured as follows: Before executing the held atomic tasks in the updated local environment, the operation page corresponding to the current session state data is initialized to clear the preceding operations of other execution nodes.

[0030] According to some embodiments of this disclosure, the execution node is also configured to force the Web automation engine to navigate to a preset unified initial page (such as a Dashboard) after reshaping the environmental characteristics of the local Web automation engine according to the fingerprint configuration, so as to clear the page state left over from the previous operation.

[0031] Furthermore, the execution node is configured as follows: Before updating the local environment based on the aforementioned task environment baseline, a target system operation context is created; Before releasing the distributed session lock, destroy the target system operation context.

[0032] According to some embodiments of this disclosure, in addition to the "forced navigation to the initial page (Dashboard)" method, an alternative solution of "context destruction and reconstruction" can be used to implement the resetting of the target operating system state by the execution node. That is, after each lock acquisition, the RPA execution node calls the web automation engine's interface, such as the function `browser.new_context()`, to create a completely new browser context to load the session and fingerprint; before releasing the lock after task completion, it calls, for example, the function `context.close()` to completely destroy the context. Although this solution has slightly higher resource overhead than the page redirection solution, it provides more thorough memory and cache isolation, preventing memory leaks.

[0033] Furthermore, the execution node is configured as follows: When executing the held atomic task in the updated local environment, if the session state data becomes invalid, a global system pause lock is configured, and an invalidation signal is sent to the central control module.

[0034] Furthermore, the RPA distributed system also includes: The self-healing collaboration module is configured as follows: After the global system is paused, the central control module takes over control of the target account and the target operating system, and attempts to log in again with AI assistance. If the AI-assisted re-login fails, send a re-login request with human assistance. After the target account logs back in, update the task environment baseline and remove the global system pause lock.

[0035] According to some embodiments of this disclosure, distributed session locks can be implemented using Redis as the underlying implementation, or they can be implemented using Zookeeper, Etcd, or row-level locks of databases, as long as the middleware technology can meet the requirements of mutual exclusion and atomicity.

[0036] According to some embodiments of this disclosure, the RPA distributed system proposed in this disclosure also includes: Self-Healing Collaboration Module (SCM): Independent of the central control module, it is responsible for handling process control after session failure, AI recognition and processing module calls, and manual collaboration degradation logic. AI Recognition and Processing Module (AI): Integrates OCR and image recognition engines, or integrates object detection models (such as YOLO) to recognize sliders, text selections and digital CAPTCHAs, or is implemented based on the Visual Language Model (VLM) to provide recognition results for unstructured challenges in SCM.

[0037] According to some embodiments of this disclosure, such as Figure 2 As shown, when a session failure exception is detected during the execution of an atomic task, the RPA distributed system performs the following operations: 1. The execution node sets a global system pause lock and sends a failure signal to the central control module, so that the central control module asynchronously routes the failure signal to the self-healing coordination module through the internal message bus (such as Redis Pub / Sub or internal RPC) to decouple the scheduling and self-healing logic; 2. The self-healing collaboration module takes over control and attempts to log in again; if an MFA or CAPTCHA challenge is detected, the AI ​​recognition and processing module is invoked for autonomous identification. 3. If AI recognition is successful, authentication will be completed automatically; if AI recognition fails, the self-healing collaboration module will automatically downgrade to sending a collaboration request to an external human intervention platform via the communication interface. 4. After authentication is complete, the self-healing collaboration module updates the shared storage area data and removes the global system pause lock; 5. Once the execution node detects that the pause lock has been removed, it loads new data and resumes execution.

[0038] According to some embodiments of this disclosure, the self-healing collaboration module (SCM) may also integrate one or more of the following communication interfaces for communicating with an external human intervention platform: (1) Email Protocol (SMTP): Send a screenshot of the verification code to the operations and maintenance team via email; (2) Voice Gateway: Notifies maintenance personnel to intervene via telephone voice broadcast; (3) WebHook callback: Push the collaborative request to the enterprise's unified operation and maintenance management platform (ITSM), and the platform will distribute the work order uniformly.

[0039] Based on the same technological concept, such as Figure 3 As shown, this disclosure also proposes a concurrent execution method for an RPA distributed system based on the above-mentioned RPA distributed system, including: Log in to the target operating system using the target account to generate a task environment baseline; Based on the task environment baseline, update the local environment of the target execution node; wherein, the target execution node is an execution node that has obtained an atomic task from the task queue and has acquired a distributed session lock; the distributed session lock is globally unique in the distributed system; The target execution node, after updating the local environment, executes the atomic task held by the target execution node and releases the distributed session lock after the task is completed.

[0040] Furthermore, when executing the atomic task held by the target execution node after updating the local environment, if the session state data becomes invalid, the login status of the target account on the target operating system is updated based on AI takeover and / or manual collaboration, and the task environment baseline is updated.

[0041] like Figure 4 As shown, this disclosure also proposes an electronic device, including a memory and a processor, wherein the memory stores a computer program or instructions, and when the computer program or instructions are executed by the processor, they are used to at least implement the above-described RPA distributed system concurrent execution method.

[0042] like Figure 5 As shown, this disclosure also proposes a computer-readable storage medium storing a computer program or instructions, which, when executed by a processor, are at least used to implement the above-described RPA distributed system concurrent execution method.

[0043] This disclosure also proposes a computer program product stored in a computer-readable storage medium, which, when executed by a processor, is used to at least implement the above-described RPA distributed system concurrent execution method.

[0044] Although the present disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and such 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 the present disclosure.

Claims

1. An RPA distributed system, characterized in that, include: The central control module and several execution nodes; among them, The execution node is configured as follows: Receive the task environment baseline of the RPA distributed system; the task environment baseline is generated by the central control module based on the login information of the target account on the target operating system; Receive atomic tasks from the task queue of the RPA distributed system; After acquiring the globally unique distributed session lock in the RPA distributed system through competition, the local environment is updated according to the task environment baseline. The held atomic task is executed in the updated local environment, and the distributed session lock is released after the task is completed.

2. The RPA distributed system as described in claim 1, characterized in that, The task environment baseline includes: session state data generated by the target account logging into the target operating system, and the network fingerprint configuration of the target operating system.

3. The RPA distributed system as described in claim 2, characterized in that, The execution node is configured to initialize the operation page corresponding to the current session state data before executing the held atomic task in the updated local environment in order to clear the preceding operations of other execution nodes.

4. The RPA distributed system as described in claim 2 or 3, characterized in that, The execution node is configured to: when executing the held atomic task in the updated local environment, if the session state data becomes invalid, configure a global system pause lock and send an invalidation signal to the central control module.

5. The RPA distributed system as described in claim 4, characterized in that, Also includes: The self-healing collaboration module is configured as follows: After the global system is paused, the central control module takes over control of the target account and the target operating system, and attempts to log in again with AI assistance. If the AI-assisted re-login fails, send a re-login request with human assistance. After the target account logs back in, update the task environment baseline and remove the global system pause lock.

6. A concurrent execution method for an RPA distributed system based on any one of claims 1-5, characterized in that, include: Log in to the target operating system using the target account to generate a task environment baseline; Based on the task environment baseline, update the local environment of the target execution node; wherein, the target execution node is an execution node that has obtained an atomic task from the task queue and has acquired a distributed session lock; the distributed session lock is globally unique in the distributed system; The target execution node, after updating the local environment, executes the atomic task held by the target execution node and releases the distributed session lock after the task is completed.

7. The concurrent execution method for an RPA distributed system as described in claim 6, characterized in that, When executing an atomic task held by a target execution node after updating the local environment, if the session state data becomes invalid, the login status of the target account on the target operating system is updated based on AI takeover and / or manual collaboration, and the task environment baseline is updated.

8. An electronic device comprising a memory and a processor, characterized in that, The memory stores computer programs or instructions, which, when executed by the processor, are used to implement at least the RPA distributed system concurrent execution method as described in claim 7 or 8.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program or instructions, which, when executed by a processor, are used to implement at least the concurrent execution method of the RPA distributed system as described in claim 7 or 8.

10. A computer program product, said computer program product being stored in a computer-readable storage medium, characterized in that, When the computer program product is executed by a processor, it is used to implement at least the concurrent execution method of the RPA distributed system as described in claim 7 or 8.

Images