Robot task planning method and device, robot and readable storage medium

Abstract

The invention relates to the technical field of robot control, in particular to a robot task planning method and device, a robot and a readable storage medium. The method comprises the following steps: acquiring instruction information received by a robot; calling a voice recognition function through a scheduling service, and recognizing the instruction information to obtain corresponding text information; the text information is input into a large text model, if the text information is an action instruction, a corresponding behavior tree is generated, and the behavior tree comprises a plurality of behavior nodes; if the text information is a language interaction instruction, a corresponding response text is generated through the large text model, and the response text is used for performing language interaction response; and controlling the robot to execute a corresponding behavior according to the behavior tree, or to perform a response operation according to the response text. According to the application, various different functions are scheduled through the scheduling service to process the issued instruction, so that the robot control is more flexible and expansible.

Description

Technical Field

[0001] This application relates to the field of robot control technology, and in particular to a robot task planning method, apparatus, robot, and readable storage medium. Background Technology

[0002] The humanoid robot large-scale grasping solution aims to enable robots to grasp designated items through voice input and large-scale model planning and control. Even if the item's position is scrambled and different colors are used for different commands, the robot can still accurately grasp the item based on semantic analysis. Simultaneously, it leverages a language large-scale model to achieve natural and fluent voice communication with users, covering numerous robot large-scale model integration scenarios. Traditional human-computer interaction is too deeply coupled with business operations, hindering business and machine expansion. Each deployment requires redeployment for new environments, which is time-consuming and labor-intensive. Furthermore, it suffers from system call defects. Traditional system calls rely on business code SDKs for targeted function development and service module calls, performing poorly in terms of business scalability, flexibility, and logic visualization. Expansion becomes difficult when business changes occur, failing to meet the diverse needs of modern complex software systems for system call control. Summary of the Invention

[0003] In view of this, embodiments of this application provide a robot task planning method that can effectively solve the problem of insufficient flexibility.

[0004] In a first aspect, embodiments of this application provide a robot task planning method, applied to a robot, the method comprising: Obtain the instruction information received by the robot; The speech recognition function is invoked through the scheduling service to recognize the instruction information and obtain the corresponding text information. The text information is input into the large text model. If the text information is an action instruction, a corresponding behavior tree is generated. The behavior tree includes multiple behavior nodes. If the text information is a language interaction command, then the corresponding response text is generated through the large text model, and the response text is used for language interaction response; The robot is controlled to perform corresponding actions based on the behavior tree, or to respond based on the response text.

[0005] In some embodiments, the text big model generates corresponding behavioral decisions based on the text information, including: If the text information is an action instruction, a corresponding behavior tree is generated, and the behavior tree includes multiple behavior nodes.

[0006] In some embodiments, controlling the robot to perform corresponding behaviors according to the behavior tree includes: The target object in the action command in the surrounding environment is obtained through visual recognition; The target object is matched using the visual recognition; When the target object is matched, pose data for executing the action command is generated; The pose data is configured into each behavior node in the behavior tree, and the behavior tree is invoked to perform action operations on the target object.

[0007] In some embodiments, the method further includes: If the behavior node of the behavior tree contains the target object keyword, the visual detection module is triggered to generate the target object pose data and transmit the pose data to the motion control module, which then performs the grasping operation.

[0008] In some embodiments, the method further includes: if the language interaction instruction does not contain keywords of the target object, triggering an independently encapsulated text-to-speech service module to convert the dialogue text generated by the large text model into speech for broadcast.

[0009] In some embodiments, the response operation based on the response text includes: When the text information is the language interaction command, the scheduling service calls the text-to-speech broadcasting service to convert the response text into speech data, and then plays the speech data.

[0010] In some embodiments, the method further includes: When the execution of the behavior tree fails, the behavior node corresponding to the failure is identified as the failure node, and the failure node is uploaded to the large text model. The large model replans the behavior tree based on the failed nodes and reissues the new behavior tree.

[0011] In some embodiments, the step of identifying the instruction information to obtain the corresponding text information includes: The instruction information is input into the instruction information conversion model to obtain the corresponding text information.

[0012] Secondly, this application also provides a robot task planning device for use in a robot, the device comprising: The information acquisition module is used to acquire instruction information received by the robot; The speech recognition module is used to invoke the speech recognition function through the scheduling service to recognize the instruction information and obtain the corresponding text information; The planning module is used to input the text information into the large text model. If the text information is an action instruction, a corresponding behavior tree is generated, and the behavior tree includes multiple behavior nodes. The first execution module is used to generate a corresponding response text through the text model if the text information is a language interaction instruction, and the response text is used for language interaction response. The second execution module controls the robot to perform corresponding actions based on the behavior tree, or to perform response operations based on the response text.

[0013] Thirdly, this application also provides a robot, wherein the terminal device includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the robot task planning method.

[0014] Fourthly, this application also provides a readable storage medium storing a computer program that, when executed on a processor, implements the robot task planning method described above.

[0015] The embodiments of this application have the following beneficial effects: This application utilizes a scheduling service to input text information into a corresponding large model. The adapted large model processes the text information, provides corresponding behavioral decisions, and then controls the robot to execute those decisions, thereby completing the operation instructed by the command information. This improves the scalability and flexibility of robot operations. When new models need to be added, they can be added through the scheduling service, fulfilling diverse system call control needs. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic flowchart of a robot task planning method according to an embodiment of this application is shown; Figure 2 This paper illustrates a scheduling scenario for a scheduling service according to an embodiment of this application. Figure 3 A schematic diagram of the task failure handling process according to an embodiment of this application is shown; Figure 4 A schematic diagram of a robot task planning device according to an embodiment of this application is shown. Detailed Implementation

[0018] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0019] The components of the embodiments of this application described and illustrated in the accompanying drawings can be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of this application provided in the drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0020] In the following text, the terms "comprising," "having," and their cognates, which may be used in various embodiments of this application, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more combinations thereof. Furthermore, the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0021] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of this application pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be construed as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of this application.

[0022] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0023] This application provides a robot task planning method. When receiving input instructions from the robot, the method processes the information through a large model and generates a corresponding behavior tree. Then, it triggers the execution nodes of the behavior tree based on keywords in the speech, thereby optimizing the robot's task control process, making the robot control process more flexible, highly scalable, and reducing the coupling of the program.

[0024] The robot task planning method will be explained below with reference to some specific embodiments.

[0025] Figure 1 A flowchart of a robot task planning method according to an embodiment of this application is shown. Exemplarily, the robot task planning method includes the following steps: Step S100: Obtain the instruction information received by the robot.

[0026] The method in this embodiment is applied to a robot and the corresponding control scenario. The exemplary scenario is a scenario where a robot is controlled through instruction information.

[0027] Commands can be either voice or text. Users can communicate directly with the robot using natural language, issuing commands that are received by the robot's microphone. Text commands can be issued by the user through a control panel, and these commands can be written in natural language.

[0028] Therefore, the robot in this embodiment can receive and process both voice and text commands.

[0029] Step S200: The speech recognition function is invoked through the scheduling service to recognize the instruction information and obtain the corresponding text information.

[0030] Regardless of the type of instruction information, it needs to be converted into corresponding text information. For example, speech-to-text information requires a speech-to-text recognition model to convert it into text. Similarly, text-based instructions require a text recognition model for conversion. For robots, the processed keywords are used as features for semantic recognition. Therefore, the text information is derived from the converted instruction information, extracting keywords and then formatting them into text composed of these keywords.

[0031] The text composed of these keywords can be in various predefined formats, such as JSON. The keywords are extracted through model processing, and corresponding prefixes are added to each keyword. Then, JSON format text is formed, which makes it easier for subsequent programs to process and recognize the text.

[0032] Step S300: Input the text information into the large text model. If the text information is an action instruction, generate the corresponding behavior tree, which includes multiple behavior nodes.

[0033] After obtaining the text information, the scheduling service will send the text information to the corresponding text big data model for processing. The text big data model will generate corresponding behavioral decisions based on the text information.

[0034] Behavioral decisions are corresponding operations made based on instruction information. The type of text information is limited by its specific content, and there are at least two types.

[0035] One type is where text information conveys language interaction commands, such as asking a question to a robot. For the robot, its body does not need to move; it only needs to provide language feedback. In this case, the behavioral decision-making only requires text recognition and then providing the corresponding response.

[0036] In another scenario, the text message conveys specific action instructions, such as asking the robot to pick up a target object. For this type of instruction, the robot will need to perform a series of movements, requiring control operations on the robot.

[0037] Therefore, different behavioral decisions will be made based on different instruction information.

[0038] As an example, when the text message conveys a specific action instruction, the text message will be specifically scheduled by the scheduling service to be processed in the behavior tree planning text model that handles action instructions. This model will output a behavior tree based on the action instruction. The behavior tree is a modular, hierarchical decision-making architecture. The behavior tree organizes behavioral logic through a tree structure, making complex behavioral decisions clear, maintainable, and easy to extend.

[0039] Once the behavior tree is generated, the robot can follow the execution logic of the behavior tree to perform the operation of picking up the target object.

[0040] Step S400: If the text information is a language interaction command, then the corresponding response text is generated through the text big model, and the response text is used for language interaction response.

[0041] For language interaction commands, a behavior tree is not needed because no coherent actions are required. Therefore, when processing language interaction commands, the scheduling service sends the commands to the large interaction processing model. This large interaction processing model and the aforementioned large text model for processing action commands are set up independently and can be considered two different modules. When scheduling is needed, the scheduling service performs the corresponding scheduling processing. In this way, the large interaction processing model can simply provide the corresponding feedback text. This feedback text then schedules the text-to-speech service module through the scheduling service to convert the feedback text into speech, which is then played back, thus realizing the corresponding language interaction process.

[0042] Regarding scheduling services, such as Figure 2As shown, after receiving text information 100, scheduling service 200 transmits it to the large text model 300 for processing. Different instructions correspond to different large models, including the aforementioned behavior tree planning text model 310 and interactive text model 320. Scheduling service 20 dispatches the text information to the corresponding large model for processing according to the aforementioned rules. It can be understood that in this structure, each large model, or other schedulable service, can be independently written, forming independent modules, which can be scheduled through the scheduling service when needed.

[0043] For example, in the language interaction process, the robot needs to convert the generated text into speech. The scheduling service is also used here to transmit the generated text to the text-to-speech service module. Therefore, the scheduling service in this embodiment can not only handle the scheduling between large models, but also handle the scheduling of various functional modules.

[0044] In one feasible embodiment, the various services, models, and modules scheduled by the scheduling service can be configured locally on the robot or reside on a cloud server. For example, if the robot is currently connected to the internet and has a connection to the cloud, the scheduling service can send text information to a large text model on the cloud server to generate the behavior tree, leveraging the cloud server's computing power, in order to obtain better recognition results or a better behavior tree. If the robot is not connected to the internet, it can also schedule a large model mounted locally.

[0045] Step S500: Control the robot to perform corresponding actions according to the behavior tree, or to perform response operations according to the response text.

[0046] Once the behavioral decision is determined, it can be executed directly, that is, the robot can be directly controlled to perform the corresponding operation according to the behavioral decision.

[0047] In the two examples above, when processing motion commands, a visual bounding box is also triggered to acquire the target object in the motion command. It can be understood that issuing motion commands to the robot could be operations such as grasping objects, which requires triggering visual recognition to identify the target object in the selected command.

[0048] Visual recognition involves robots capturing images of their surroundings using cameras, then using image recognition to identify target objects or perform environmental analysis. Visual recognition can also utilize pre-trained visual models. One approach is to trigger a visual bounding box. For example, a rectangular area in the center of the robot's visual image can be designated as the bounding box, allowing the robot to scan and recognize only objects within this area, ignoring objects at the edges and corners to minimize the impact of incomplete or missing images.

[0049] For example, if the instruction is "lift the box", then the box is the target object. At this time, image recognition is started through the visual bounding box to scan the surrounding environment to find boxes. After recognizing the box, the process of lifting the box begins.

[0050] Meanwhile, the box's location, size, and other state data can be identified through visual frame scanning. This data affects how the robot should actually move the box. For example, for a box that is elevated, the robot needs to lift its robotic arm; for a box on the ground, the robot needs to bend down. Therefore, when matching a target object, the robot generates corresponding positional data based on the object's position. This pose data is the robot's pose when moving the target object.

[0051] Meanwhile, in order to move objects, a behavior tree has been pre-generated. For example, the behavior nodes planned by the behavior tree can be: 1. Walk to a distance of 30 centimeters in front of the object, 2. Extend the robotic arm to grab the object, 3. Lift the object to a certain height, 4. Move the object to the target position.

[0052] As can be seen, in the four steps described above, steps 2, 3, and 4 require different pose data for robot control. Therefore, when a target object is detected, the generated pose data is applied to various behavior nodes in the behavior tree. This means that pose data will be generated when grasping, carrying, and lifting the object. It can be understood that this data needs to be generated based on actual real-world conditions, rather than being generated by a large model simply based on instructions.

[0053] After the pose data is generated, it will be configured into each behavior node in the behavior tree. Then, the corresponding operation can be completed by directly calling the behavior tree to perform action operations on the target object.

[0054] For voice interaction commands, the response text obtained in the previous steps can be used directly as a reply. For example, when replying by voice, the scheduling service schedules the text-to-speech service to convert the text and generate the response audio, which is then played by the speaker.

[0055] For interactions using a console, there is no need for the text-to-speech step. The desired text can be displayed directly on the interface, thus completing the language interaction process.

[0056] When a robot executes operations from a behavior tree, failures may occur. For example, if the behavior tree has four behavior nodes (1, 2, 3, and 4), and the robot fails to execute the task at behavior node 3, it cannot execute the operation at behavior node 4. Therefore, if a node fails, the entire instruction will stop, resulting in instruction execution failure.

[0057] like Figure 3 As shown, this embodiment provides a troubleshooting process for such failures, including: Step S410: When the execution of the behavior tree fails, the behavior node corresponding to the failure is determined as the failure node, and the failure node is uploaded to the large text model.

[0058] At this point, the robot in this embodiment can identify the behavior node corresponding to the failure as a failure node and upload the failure node to the large text model. The large model can analyze the failure node to determine the cause of the failure and make corresponding adjustments.

[0059] It is understandable that the behavior tree in this embodiment itself has refined the steps of the task to be performed, dividing it into individual behavior nodes. Compared with the usual operation of integrating solutions in the program or selecting a predetermined solution after model recognition, this embodiment can easily trace the source of abnormal situations through the structural features of the behavior tree. When the task fails, the scope of the abnormality can be greatly narrowed, and the behavior node that caused the task failure can be directly located.

[0060] Furthermore, the large model can have sub-supervision capabilities, and failed nodes can be used as data for self-supervised learning to optimize the model. Therefore, when a task fails, the current failed behavior node can be directly fed back to the large text model.

[0061] In step S420, the large model replans the behavior tree based on the failed nodes and reissues the behavior tree.

[0062] Based on the failed node and previous instructions, the large model re-plans the behavior tree. After the re-planning is completed, the new behavior tree is issued again, and the previous instructions can be executed again.

[0063] In some cases, there may be multiple failures. In such cases, steps S410 and S420 will be repeated for multiple self-corrections. Under this logic of replanning the behavior tree, the behavior tree obtained each time will be different, and it will eventually converge toward the result of successfully executing the behavior. Therefore, the planning scheme after failure in this embodiment can enable the robot to have a certain self-repair capability, so that when the task fails, it does not need human intervention to repair it, nor will it enter an infinite loop. Instead, it will continuously learn and iterate on its own, and eventually successfully execute the action command.

[0064] In summary, this embodiment uses a behavior tree to execute the robot's action commands, making each step independent and clear. When a task fails, the specific steps that failed can be identified, allowing the robot to optimize based on the failed behavior nodes. This allows for the replanning of an optimized behavior tree, which, once issued, is more likely to lead to successful task execution. Furthermore, the system architecture with a scheduling service significantly reduces the coupling within the robot system, making the functions of each module clear, facilitating maintenance and expansion, and improving the overall flexibility of the system. When processing voice interaction-related commands, the scheduling service handles them flexibly, responding directly without generating a behavior tree, avoiding rigid processing behavior. This application, through a behavior tree node failure or success status feedback mechanism, enables the large model to replan action modules based on failure information, effectively improving the success rate of task execution and enhancing the system's adaptability in complex scenarios.

[0065] Figure 4 A schematic diagram of a robot task planning device according to an embodiment of this application is shown. Exemplarily, the robot task planning device includes: The information acquisition module 10 is used to acquire instruction information received by the robot; The speech recognition module 20 is used to call the speech recognition function through the scheduling service to recognize the instruction information and obtain the corresponding text information; The planning module 30 is used to input the text information into the large text model. If the text information is an action instruction, a corresponding behavior tree is generated. The behavior tree includes multiple behavior nodes. The first execution module 40 is used to generate a corresponding response text through the text model if the text information is a language interaction instruction, and the response text is used for language interaction response. The second execution module 50 controls the robot to perform corresponding actions according to the behavior tree, or to perform response operations according to the response text.

[0066] It is understood that the apparatus in this embodiment corresponds to the robot task planning method in the above embodiments, and the options in the above embodiments are also applicable to this embodiment, so they will not be described again here.

[0067] This application also provides a robot, exemplary in that the robot includes a processor and a memory, wherein the memory stores a computer program, and the processor, by running the computer program, causes the robot to perform the functions of the various modules in the robot task planning method or the robot task planning device described above.

[0068] The processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, including at least one of a Central Processing Unit (CPU), Graphics Processing Unit (GPU), Network Processor (NP), Digital Signal Processor (DSP), Application-Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application.

[0069] The memory can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The memory is used to store computer programs, and the processor can execute the computer programs accordingly after receiving execution instructions.

[0070] This application also provides a readable storage medium for storing the computer program used in the robot described above.

[0071] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that, in alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0072] In addition, the functional modules or units in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0073] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a smartphone, personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0074] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A robot task planning method, characterized in that, Applied to robots, the method includes: Obtain the instruction information received by the robot; The speech recognition function is invoked through the scheduling service to recognize the instruction information and obtain the corresponding text information. The text information is input into the large text model. If the text information is an action instruction, a corresponding behavior tree is generated. The behavior tree includes multiple behavior nodes. If the text information is a language interaction command, then the corresponding response text is generated through the large text model, and the response text is used for language interaction response; The robot is controlled to perform corresponding actions based on the behavior tree, or to respond based on the response text.

2. The robot task planning method according to claim 1, characterized in that, The control of the robot to execute corresponding behaviors according to the behavior tree includes: The target object in the action command in the surrounding environment is obtained through visual recognition; The target object is matched using the visual recognition; When the target object is matched, pose data for executing the action command is generated; The pose data is configured into each behavior node in the behavior tree, and the behavior tree is invoked to perform action operations on the target object.

3. The robot task planning method according to claim 2, characterized in that, Also includes: If the behavior node of the behavior tree contains the target object keyword, the visual detection module is triggered to generate the target object pose data and transmit the pose data to the motion control module, which then performs the grasping operation.

4. The robot task planning method according to claim 1, characterized in that, Also includes: If the language interaction command does not contain keywords related to the target object, the independently encapsulated text-to-speech service module is triggered to convert the dialogue text generated by the large text model into speech for broadcast.

5. The robot task planning method according to claim 1, characterized in that, The response operation based on the response text includes: When the text information is the language interaction command, the scheduling service calls the text-to-speech broadcasting service to convert the response text into speech data, and then plays the speech data.

6. The robot task planning method according to claim 1, characterized in that, Also includes: When the execution of the behavior tree fails, the behavior node corresponding to the failure is identified as the failure node, and the failure node is uploaded to the large text model. The large model replans the behavior tree based on the failed nodes and reissues the new behavior tree.

7. The robot task planning method according to claim 1, characterized in that, The process of identifying the instruction information to obtain the corresponding text information includes: The instruction information is input into the instruction information conversion model to obtain the corresponding text information.

8. A robot task planning device, characterized in that, The device, applied to robots, includes: The information acquisition module is used to acquire instruction information received by the robot; The speech recognition module is used to invoke the speech recognition function through the scheduling service to recognize the instruction information and obtain the corresponding text information; The planning module is used to input the text information into the large text model. If the text information is an action instruction, a corresponding behavior tree is generated, and the behavior tree includes multiple behavior nodes. The first execution module is used to generate a corresponding response text through the text model if the text information is a language interaction instruction, and the response text is used for language interaction response. The second execution module controls the robot to perform corresponding actions based on the behavior tree, or to perform response operations based on the response text.

9. A robot, characterized in that, The terminal device includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the robot task planning method according to any one of claims 1-7.

10. A readable storage medium, characterized in that, It stores a computer program that, when executed on a processor, implements the robot task planning method according to any one of claims 1-7.

Images