A robotic process automation multi-tool integration method and system

Abstract

The present invention "A multi-tool integration method and system for robotic process automation" discloses a multi-tool integration method and system for robotic process automation in the field of enterprise automation, which is suitable for the construction of an RPA platform or system based on multiple RPA tools. The present invention encapsulates each RPA tool into an RPA service program that provides a standard interface, and independently receives and executes delegated RPA programs. Using a special translation module, the RPA service program automatically translates the RPA program to be executed into a local program that can be executed by the local RPA tool. According to the capabilities of multiple RPA service programs, the RPA program is automatically split into a tree-like, master-slave call multi-program structure, in which each RPA program can allow a certain RPA service program to execute independently. The split program is automatically delegated to the appropriate RPA service program instance, collaboratively automates the same business process, and provides comprehensive RPA capabilities.

Description

Technical Field

[0001] The present invention relates to the Robotic Process Automation (RPA) technology in the field of enterprise automation, and is applicable to the construction of RPA platforms and systems using a variety of RPA tools. Specifically, it uses automatic splitting, delegation, and translation mechanisms to enable multiple RPA tools (including real people) to collaborate and automate the same business process, providing enterprises with more comprehensive RPA capabilities. Background Art

[0002] There are some low-skilled information processing tasks in the business process of an enterprise. These tasks have a low input-output ratio, low employee enthusiasm, and are prone to errors. The technical limitations of the human-computer interface are the key to such problems. Although human-computer interface technology has been improving, the collaboration between people and information systems is still cumbersome and unsmooth overall. Simple tasks (such as report downloads, data statistics, image uploads, and document records, etc.) will consume a lot of energy once the number increases. In addition, the increasingly serious situation of "garden walls" inside and outside the enterprise has also accelerated the breeding of these low-value tasks: the external large platforms used by enterprises, such as Tmall, JD.com, Baidu and WeChat, are difficult to communicate with the internal information systems of the enterprise. There are often multiple independent information systems within the enterprise, which cannot be effectively integrated. The traditional solution is usually to arrange manpower to transmit information and data between multiple systems and platforms, which is inefficient and error-prone;

[0003] To address the above problems, Robotic Process Automation (RPA) uses automatically executable scripts and software modules to construct software robots (bots) or virtual workers to simulate real-life operations on the information system interface, automatically completing simple, consistent, and repeatable low-skill tasks in business processes, improving labor efficiency and customer and employee experience.

[0004] RPA is an emerging business process automation technology that has attracted many developers and companies, and has formed a wide variety of RPA tools, libraries, frameworks, and scripting languages ​​(for example: UI PATH, TagUI, SikuliX, Robin, ROBOTframework, Automation Anywhere, and RPA for Python, etc.). They each have their own advantages, but there are often compatibility issues. In actual RPA process development, multiple tools or scripts are often used for different scenarios, and even human participation is indispensable (especially for login interfaces that require real-person verification). The benefit of integrating different RPA tools is that it can combine all technical advantages and provide comprehensive RPA capabilities. How to integrate these different (especially open source) RPA tools, scripts, and real-person interactions, and effectively manage and maintain RPA processes has become an important issue. Summary of the invention

[0005] We will explain the various technical features of the present invention based on a specific implementation example. The technical details therein are only used to clearly and completely explain the features and methods of the present invention, rather than to limit the implementation methods of the invention. Based on the features and methods of the present invention, all other embodiments obtained by those skilled in the art will fall within the scope of protection of the present invention. The scope of protection of the entire invention is fully described in the claims.

[0006] 1. RPA Model

[0007] The RPA model defines the logical entities related to the present invention and the associations between them, and is a conceptual framework followed by the implementation of RPA technology. Figure 1 The RPA model of the present invention is shown. The specific entities are explained as follows:

[0008] Application

[0009] Applications (110) are various information systems, platforms and tools that support enterprise operations. Some applications have a human-computer interface for interaction with people, while others directly provide programmable API interfaces. RPA tools interact with these applications to complete business processes.

[0010] Operation

[0011] Operation (115) is to complete a certain function on the human-computer interface of the application (110) or through the API. For example, "open the Firefox browser on the desktop", "open baidu.com in the browser", "enter the search keyword: RPA" and "click the search button on the page". In the present invention, a single operation (115) can only be performed on a single device (virtual or physical computer or other intelligent device, etc.). Generally, a single operation (115) is only for a single application (110).

[0012] People

[0013] Real people (105) are employees of the enterprise. In the absence of automation, they are the executors of operations (115). In the automation process, real people are no longer the direct executors of operations (115). In some scenarios where automation is impossible or very difficult, real people (105) participate in the execution of operation commands (120). For example Figure 3 For operations such as “login with QR code” (325), “rotate the image to the correct position” (310), “select the image of the specified object” (315), and “enter the SMS verification code” (320) listed in the example, it is necessary to give a prompt to a real person (105) so that he / she can manually feedback the correct result to the platform.

[0014] Command

[0015] A command (130) is a reusable computer program module. A command usually has one or more command parameters (131) to increase the reuse rate of the code. For example, the command: "Enter text Y in input box X", if X is "search input box" and Y is the keyword "RPA", the operation of "enter keyword RPA" can be automated (115). Commands are divided into three categories:

[0016] 1. Operation command (120): a program module for implementing the operation (115);

[0017] 2. Process call command (135): a program module that implements calling a subprocess (145);

[0018] 3. Control Commands (125): Control Commands (130) execute sequential commands to implement control logic such as IF, WHILE and FOR.

[0019] Data Item

[0020] Data items (170) are variables that are stored permanently. They are stored in<Key, Value> The command (130) uses the key to read or write the data item to realize the reading and saving of the operation status and results. For example, the data item "Current screen image" stores the image file captured by the screenshot command. Subsequent commands can read this file to search for the location of a certain application icon on the screen.

[0021] Step

[0022] Inputting the command parameter (131) value, calling or executing a command (130) forms a step (141). If the control command (125) is called, the step (141) may include a sequence consisting of multiple steps, such as a WHILE control logic, which includes a loop body consisting of steps.

[0023] Process and Subprocess

[0024] A process (140) is a program that automates a business process and consists of multiple steps (141) that follow a certain execution logic. In order to achieve reuse, the process itself will have one or more process parameters (155). A process has a unique process ID (142). A single run of a process is called a process instance (150) of the process. A process (140) can use the process call command (135) to call other processes, and these called processes are called its subprocesses (145).

[0025] RPA Tool

[0026] Different RPA tools have different usage methods. Some use special scripts, such as Sikulix, Robin, Robot Framework, and TagUI; others provide class library calls for some general programming languages, such as RPA for Python, robotjs, and Selenium. In the present invention, we collectively refer to them as RPA tools (185). Different RPA tools (185) have different technical advantages. By integrating multiple tools, we can combine multiple advantages and build comprehensive RPA capabilities.

[0027] Native RPA program:NPG

[0028] A program executed by an RPA tool (185) on a single device is called a native RPA program (190, Native RPA Program, hereinafter referred to as NPG). The device can be a general-purpose PC, tablet, smartphone, or commercial server. The device can be virtual or physical;

[0029] Operation (115) is finally implemented by NPG. When NPG is executed, a series of automated operations are automatically carried out to achieve the business goal. In order to support different open source technologies and tools, the present invention does not limit the implementation technology of NPG, and various scripting languages ​​(Python, Shell script or JavaScript) can be used, and languages ​​such as Java or C# can also be directly compiled into executable programs.

[0030] Robot

[0031] The robot (180) is an RPA service program that runs on a single device and uses a single RPA tool (185) to execute the process (140). The device can be a general-purpose PC, tablet, smart phone, or commercial server. The device can be virtual or physical.

[0032] Robot Type

[0033] In the present invention, the robot (180) is implemented based on a single RPA tool (185). Different robots are implemented by different RPA tools and have different operation (115) capabilities. The capability here refers to the set of supported commands (130). All robots with the same capability belong to one robot model (175).

[0034] Single / Multi-RT Process

[0035] If all commands (130) of a process (140) can be executed by a robot (180) of a certain model (175), we call this process a single-model process (165), otherwise we call it a multi-model process (160). Before execution, the multi-model process (160) must be split into multiple single-model processes (165) according to the rules.

[0036] Translator

[0037] The translator (195) translates the single model process (165) into NPG (190) that can be executed by the local RPA tool (185). When the robot (180) executes the single model process (165), it will first use the translator (195) to translate it into NPG (190), and then call the RPA tool (185) to execute it. Each robot model (175) has a specific translator (195).

[0038] 2. Process splitting

[0039] In order to facilitate understanding of the present invention, we explain the definition and splitting process of the process (140) based on an implementation example. The present invention does not limit how to define the process (140), but only stipulates that the process must be structured and can be split into single-model processes (165) according to certain rules during operation.

[0040] Process example

[0041] This implementation example uses XML to describe the business process and provides visual editing, display and management. Assume there is a business process for downloading and converting advertising data reports. The specific steps are as follows:

[0042] a) Log in to the advertising system and download yesterday’s report data

[0043] b) Use the tool R to generate multiple statistical charts (in the form of pictures)

[0044] c) Add statistical charts to PPT files

[0045] d) Upload the PPT file to the specified FTP directory

[0046] Figure 4 415-490 in the figure is a visual diagram of the process of this implementation example, and 412 is a partial example of the corresponding XML structure. In the diagram, each box represents a step (141). The gray background marks the data items read and written by the process (170). In this implementation example, the entire process is a sequence of steps executed from top to bottom. Unlike other commands, the call step of the FOR-EACH control command (125) contains a loop body consisting of two steps (141) 475 and 480. 410 is the icon of the robot model (175). The current implementation example has 4 models of robots related to this process, as follows:

[0047] 1) Web Automation: supports various operations for Web applications, but does not support real-person verification login;

[0048] 2) R service: provides various R-based statistical analysis services;

[0049] 3) Office Automation: Provides various file editing operations of MS Office suite;

[0050] 4) Human verification: providing a variety of human verification operations, including rotating the image to the correct position (310);

[0051] For the convenience of explanation, we mark the model icon that can execute the command on the right side of step (141). It is easy to see that this is a multi-model process (160) that must be decomposed into single-model processes (165) before it can be executed.

[0052] Split method

[0053] In order to achieve transparency of RPA technology, it is necessary to use an algorithm to automatically split the multi-model process (160) into multiple single-model processes (165), forming a tree-like master-slave call relationship to complete the functions of the original large process. The present invention does not specify the specific splitting method, and the implementer shall implement it according to the specific situation. This implementation example adopts Figure 5 The algorithm realizes automatic splitting. The properties and related functions related to the algorithm are explained as follows:

[0054] 1) Attribute CommType: command type of step call, divided into three types: "control (logic)", "operation (operation)" and "process call (call-proc)";

[0055] 2) Attribute RTList: A list of robot models that can execute the commands in the current step. For example, in this example, there are: ["Web Automation", "R Service", "Office Automation", "Human Verification"];

[0056] 3) Attribute Block: represents the sequence of steps included in the steps of executing the control command;

[0057] 4)Method AddStep: Add a step to the current process;

[0058] 5) Function CreateCallProc: creates a process call step according to ID;

[0059] 6) Function GetDominatedRT: Input a step sequence, a starting and ending step range, scan the RTList property of each step in this range, and count the number of times the robot model (175) appears. Return a list (which we call the dominant model list) containing one or more robot models with the highest number of appearances (if the number of appearances is the same, multiple models will appear); if the command type is a non-logical command, the RTList property will be returned directly;

[0060] like Figure 5It is shown that the whole algorithm is a recursive process of layer-by-layer decomposition. The input is the ID of the parent process and the range of the step sequence to be decomposed. 515 first scans the step sequence within the specified range, counts a dominant robot model CurrRT that supports most commands, and then creates a sub-process based on the parent process ID and the robot model CurrRT. The parent process ID of the process before decomposition is a fixed ID, such as 0. Part 520 is to scan the steps one by one, merge the steps whose adjacent dominant model lists do not contain CurrRT into a sub-process, and continue to decompose. 525 is to find the steps whose adjacent dominant model lists do not contain CurrRT. The first step of 535 determines whether there is one or more adjacent steps of this type. If there is (step 2), split it again to obtain a sub-process ID, and then add a sub-process call step to the current process (step 3). If not (steps 4-5), directly add the current step to the current process. Finally, the calculation process returns the ID of the current process (540).

[0061] Figure 5 The algorithm is executed once to decompose a multi-model process (160) into multiple single-model processes (165). There is a master-slave calling relationship between these single-model processes. Figure 6 is Figure 4 The result after the process is split. The main process (610) is split into three single-model sub-processes 615, 620 and 625. The main process calls these sub-processes through the process call steps (530, 535 and 540).

[0062] 3. General RPA technology integration platform

[0063] Figure 2 This is an RPA technology integration platform developed based on the RPA model in the previous implementation example. The entire platform is divided into 7 main modules, which are explained as follows:

[0064] Storage Center

[0065] The storage center (240) is a persistent storage system for metadata, operation or status data. Other modules read and write data in the storage center through APIs. The stored data mainly includes:

[0066] a) Process information (250): Use XML structure to save the specific definition of the process (140), including the definition of the sub-process after splitting. Other modules can read the corresponding process definition through the process ID;

[0067] b) Process instance information (245): data related to the instance after the process is executed, including the instance status, execution start time, instance ID, execution progress, etc.;

[0068] c) Data item information (260): stores the key and value information of the data items used by the process instance (150), and supports unstructured data, including text, images, and videos;

[0069] d) Robot information (255): Each robot will register and update relevant information in the storage center, including: service interface, robot model and current status. In addition, the command set supported by each robot model is also registered in the robot information in advance.

[0070] Process Editor

[0071] RPA developers create, edit, save and test processes (140) through a visual "process editor" (235). The process editor can call the RPA console (265) to test, run and observe the operation of the process (140).

[0072] RPA Console

[0073] Generally, administrators or business personnel use the RPA console (265). It provides an interface and API to manage, schedule and monitor processes. It issues process scheduling instructions to the RPA scheduling center (270) to run, pause or close process instances (150).

[0074] RPA Scheduling Center

[0075] The RPA scheduling center (270) is mainly responsible for running, pausing or closing the process instance (150) according to the process scheduling instructions. The three key modules are:

[0076] 1) "Process instance construction" (275): According to the process ID to be started, a process instance (150) is constructed, a unique instance ID is assigned, and the relevant information is saved in the storage center;

[0077] 2) “Process Splitting” (280): For the first run process (140), read the information from the storage center and Figure 5 The process is split into multiple single-model processes (165), and then the process information is saved to a storage center (240);

[0078] 3) "Instance delegation" (285): According to the single model process corresponding to the instance ID, through a certain algorithm, find the robot with the same model and the smallest load in the robot information, and notify it to execute this instance. Figure 2Only the scenario of notifying robot X (210) is shown, but other robots (295) have the same relationship with the RPA scheduling center. During the scheduling process, the RPA scheduling center (270) will also send the instance status (progress) to the instance status message service (290).

[0079] robot

[0080] The platform has multiple robots (180) (such as 210 and 295), which are distributed on multiple devices. When the robot is started and running, it will send the status and other related information to the storage center and save it in the robot information (255). It is provided to other modules for query. After receiving the delegated instance from the RPA scheduling center (270), the robot (180) will first use the translator module (215) to translate the single-model process (165) corresponding to the instance into NPG (190), and then the process execution module (220) will call the RPA tool (185) for execution. Each robot model (175) has a specific translator (180). During the execution process, each process instance (150) will send a status message (running progress) to the instance status message service (290), and other modules can query the running status of the instance. In addition, each process instance will read and write data items (260), which is the main way to share data between single-model processes (165);

[0081] The running process instance (150) will request the RPA scheduling center (270) to start the sub-process and return relevant information through the process call command (135). The process call command (135) will query the (running) status of the sub-process instance on the instance status message service (290).

[0082] Instance status message service

[0083] A message service is provided for the process instance (150), and each process instance (150) has an independent message channel in the service. The purpose of setting up the message service is to support asynchronous calling and distributed computing of the process (140).

[0084] Real People

[0085] Figure 2 Medium Robot y The process instances (150) in the code interact with real people (105,230), allowing the human brain to handle problems that algorithms and tools cannot solve automatically. For example, Figure 6 In the sub-process 615, the small pictures are first captured, and then people are asked to rotate the small pictures to the correct position, and the position of the small pictures and the number of clicks are recorded. The subsequent steps are to reproduce the human click operation on the human-computer interface.

[0086] 4. Process Execution Example

[0087] Figure 7 is Figure 2 Execute the split process sample on the platform ( Figure 6 As shown in the figure, the administrator issues a dispatch instruction to the RPA dispatch center through the RPA console (710) to execute the process SP-198083740. The dispatch center (715) first constructs an instance for this process a In the case of the first call, the dispatch center (715) will also split this process to form Figure 6 The four single-model processes (165) of 610, 615, 620 and 625 are shown in Figure 165. Next, the dispatch center (715) is an example a Find a matching robot x (720) and delegates it to execute. x (720) Translate the process into NPG (190) and execute it. During this process, the sub-process call step (530) will be executed, and the RPA dispatch center will be requested to call the sub-process SP-198083740-1 (615), and notify the "Data Analysis Assistant" to manually adjust the image to the correct direction. The dispatch center (715) first constructs an instance for this process b , delegated to the appropriate robot y (725). Robot y (725) Translate the process into NPG (190) and execute it. Process instance b The human-computer interaction interface will send the small picture list to the "data analysis assistant" and ask him / her to click and place the pictures in the correct direction, and then save the click data to the data item (170). y (725) will return the process execution completion message to the robot x (720). Robot x (720) Continue with the next step in the process. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] Figure 1 It is the RPA model, which defines the logical entities in the RPA field and the relationships between them. It is the conceptual framework followed by this invention.

[0089] Figure 2 This is a structural implementation example of a technology integration platform developed based on the RPA model.

[0090] Figure 3 Lists some human verification interfaces that cannot be automated or are very difficult to automate. In this case, real people must participate and guide the RPA platform to complete the correct operation

[0091] Figure 4It is an XML structure and visual interface of an RPA program or process, which belongs to a certain implementation example.

[0092] Figure 5 An automatic splitting algorithm for an RPA program or process, which belongs to a certain implementation example

[0093] Figure 6 is in accordance with Figure 5 Algorithm split Figure 4 Results after using the RPA program sample

[0094] Figure 7 is Figure 2 Executed on the sample platform Figure 4 An RPA program example, including an interaction diagram of some processes.

Claims

1. In a multi-user environment, based on multiple computers or smart devices, a method that uses automatic splitting, delegation, and translation mechanisms to enable multiple RPA tools, including RPA tools that support real-person interaction, to collaborate to automate the same business process is characterized by the following: a) Define a set of commands that are independent of RPA tools and cover operations on various systems, tools, or platforms within the business scope. These commands are divided into three categories: operation commands, process call commands, and control commands. Developers define structured RPA processes based on this command set and globally accessible permanent storage variables to describe business processes. This is independent of the specific RPA tool used to implement the process. The process itself will have one or more process parameters to support process reuse. b) Each RPA tool is encapsulated into an independent RPA service program that provides a standard interface. Each RPA service program only supports some of the commands in step a. One command can be supported by multiple RPA service programs. Each RPA service program registers its service interface and command set with the storage center. c) If a single RPA service program cannot execute all the commands in the process, the algorithm is used to automatically split the process according to the one or more RPA service programs corresponding to each command. By adding process call commands, a tree-like multi-process structure based on the master-slave call relationship is formed. Each process can be executed independently by a certain RPA service program. The multiple RPA service programs that execute them can collaborate to complete the functions of the original large process through process calls; d) When delegating a process, an algorithm is used to find a running instance of the RPA service program on a certain device based on the RPA service program that can independently execute the process corresponding to the process, and the instance is handed over to the RPA service program for execution; e) When executing a process, each RPA service program running instance first translates the process to be executed into a local program that can be executed by the local RPA tool through a dedicated translation module, and then executes these local programs to complete the functions required by the process; f) During the execution process, the running progress of the process instance is sent to the corresponding message pipeline. The local program code corresponding to each command reads and writes the permanently stored variables according to the name or key value through the API interface. The permanently stored variables are stored in a place accessible to all running instances of the RPA service program; g) During the execution process, the dispatch center is requested to delegate and start other processes through the process call command, and the running progress of the sub-process instance is queried from the message pipeline of the process instance.

2. The commands in step a of the method of claim 1 include not only the logical control structures: IF...THEN, WHILE loop and FOR loop, but also operations that require interaction with real people.

3. The commands in step a of the method of claim 1 can read and write permanently stored variables, the effective range of the variables is limited to the entire process, and the permanently stored variables belonging to the same process can be freely read and written by all commands in the process.

4. The feature of step b of the method described in claim 1 is that each RPA service program can be installed and run on one or more computers or other intelligent devices to form multiple RPA service running instances, which are independent of each other and do not depend on each other.

5. A system that is implemented in a multi-user environment based on multiple computers or smart devices, uses standard interfaces to encapsulate multiple RPA tools including RPA tools that support real-person interaction, and uses a mechanism for automatically splitting, delegating, and translating RPA processes to allow multiple RPA tools including RPA tools that support real-person interaction to collaborate and automatically complete the same business process; it manages and runs a structured RPA process that is collaboratively executed by multiple RPA tools including RPA tools that support real-person interaction, and developers do not need to specify which RPA tool to use when editing; the specific features of this system are as follows: a) The system includes an RPA process, a storage center, an RPA console, an RPA scheduling center, an instance status message service, and multiple RPA service programs; b) The storage center stores multiple RPA processes. RPA processes are composed of commands that are independent of RPA tools and cover operations on various systems, tools, or platforms within the business scope. These commands are divided into three categories: operation commands, process call commands, and control commands. They include IF…THEN, WHILE loops, and FOR loops, as well as operations that require interaction with real people. They support globally accessible permanent storage variables as parameter values. The RPA process itself will have one or more process parameters to support process reuse. c) Each RPA tool is encapsulated into an independent RPA service program that provides a standard interface. Each RPA service program can be installed and run on one or more computers or other intelligent devices to form multiple RPA service running instances. They are independent of each other and do not depend on each other. Each RPA service program supports a set of specific commands, and one command can be supported by multiple RPA service programs. Each RPA service program registers its service interface and command set with the storage center. d) The RPA console sends a process scheduling instruction to the RPA scheduling center to run, pause, or close the RPA process; e) After receiving the process run instruction, the RPA dispatch center obtains the RPA process from the storage center. If a single RPA service program cannot execute all the commands in the process, the algorithm is used to automatically split the process according to the one or more RPA service programs corresponding to each command, and the process is saved in the storage center. By adding process call commands, a tree-like master-slave call multi-process structure is formed to work together to complete the functions of the original process. Each process can be executed independently by a certain RPA service program. f) When the RPA dispatch center executes a process, it uses an algorithm to find a running instance of the RPA service program on a certain device based on the RPA service program that can independently execute the process corresponding to the process, and then assigns it to execute; g) Each RPA service program has a dedicated translation module. The running instance will translate the process to be executed into a local program that can be executed by the local RPA tool, and then execute these local programs to complete the functions required by the process; h) During the execution process, the RPA service program requests the dispatch center to delegate and start other processes through process call commands, and queries the running progress of the sub-process instance from the message pipeline of the process instance; i) When the RPA service program instance executes the process, it sends the running progress of the process instance to the corresponding message pipeline in the instance status message service. The local program code corresponding to each command reads and writes permanently stored variables according to the name or key value through the API interface. The permanently stored variables are stored in a storage center that can be accessed by all running instances of the RPA service program.

Images