Exoskeleton robot control method with semantic perception function

Abstract

The embodiment of the application discloses a control method of an exoskeleton robot with semantic perception function, which can improve the intelligence, autonomy and generalization of the exoskeleton robot. The method comprises the following steps: receiving language description information input by a user for a target task; performing semantic extraction on the language description information through a large language model to analyze the assistance required to complete the target task and generate target assistance parameter configuration information corresponding to the target task; and controlling the exoskeleton robot to provide assistance to the user according to the target assistance parameter configuration information.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computers, and in particular to a control method for an exoskeleton robot with semantic perception function. BACKGROUND

[0002] An exoskeleton robot is a robot designed by imitating the skeletal structure of a human body and utilizing the load-bearing capacity of the skeleton, which can provide assistance to a person (i.e., exoskeleton assistance) to enhance the human body's ability,

[0003] However, the conventional exoskeleton assistance is difficult to recognize semantic information in the environment, and can only open loop or ignore this part of information to assist the wearer, so that the intelligence, autonomy and generalization of the exoskeleton robot's assistance operation are not good, which greatly limits the application of the exoskeleton robot in the field of rehabilitation, etc. SUMMARY

[0004] The purpose of the embodiments of the present application is to provide a control method for an exoskeleton robot with semantic perception function, which can improve the intelligence, autonomy and generalization of the exoskeleton robot's assistance operation.

[0005] In a first aspect, the embodiments of the present application provide a control method for an exoskeleton robot with semantic perception function, which comprises:

[0006] receiving language description information input by a user for a target task;

[0007] performing semantic extraction on the language description information by a large language model to analyze assistance required for completing the target task and generate target assistance parameter configuration information corresponding to the target task;

[0008] controlling the exoskeleton robot to provide assistance to the user according to the target assistance parameter configuration information.

[0009] As a possible implementation, the performing semantic extraction on the language description information by a large language model to analyze assistance required for completing the target task and generate target assistance parameter configuration information corresponding to the target task comprises:

[0010] performing semantic extraction on the language description information by the large language model to obtain first information, the first information comprising feature information of a task execution object associated with the target task and an action instruction associated with the target task;

[0011] The large language model is used to determine second information according to the first information and a pre-learned association relationship between feature information of a task execution object and feature information of an action instruction and assistance, the second information including feature information of assistance required to complete the target task;

[0012] The large language model is used to generate the target assistance parameter configuration information according to the second information and a pre-learned association relationship between feature information of assistance and assistance parameter configuration information.

[0013] As a possible implementation, the large language model is used to generate the target assistance parameter configuration information according to the second information and a pre-learned association relationship between feature information of assistance and assistance parameter configuration information, including:

[0014] The large language model is used to generate the target assistance parameter configuration information according to the second information and a pre-learned association relationship between feature information of assistance and a type and a configuration value of assistance parameter required to be configured.

[0015] As a possible implementation, the method further includes:

[0016] In response to receiving language description information input by a user for a target task, the working mode of the exoskeleton robot is switched to a first working mode to control the exoskeleton robot to not provide assistance to the user;

[0017] In response to receiving language description information input by a user for a target task, the working mode of the exoskeleton robot is switched to a first working mode to control the exoskeleton robot to not provide assistance to the user;

[0018] As a possible implementation, the method further includes:

[0019] According to a sensor parameter of the exoskeleton robot, an abnormality score of the exoskeleton robot is determined, the abnormality score being used to represent an abnormality degree of human-robot contact;

[0020] In a case where the abnormality score is greater than a set score threshold, the user is prompted to re-input language description information for a target task, and a large language model is used to re-generate target assistance parameter configuration information according to the re-input language description information to adjust assistance provided by the exoskeleton robot to the user.

[0021] As a possible implementation, the large language model is used to generate the target assistance parameter configuration information according to the second information and a pre-learned association relationship between feature information of assistance and assistance parameter configuration information, including:

[0022] inputting the target assist parameter configuration information to a controller of the exoskeleton robot to control the exoskeleton robot to provide assist to the user;

[0023] The controller is preconfigured with a parameter adjustment interface for receiving the assist parameter configuration information generated by the large language model, the assist parameter configuration information is represented by using one or more combinations of bits, and different combinations of the one or more bits correspond to different assist parameter configuration schemes.

[0024] In a second aspect, an exoskeleton robot control device with semantic perception function is provided, and the device comprises:

[0025] A communication module is configured to receive language description information input by a user for a target task;

[0026] A processing module is configured to perform semantic extraction on the language description information by using a large language model to analyze assist to be provided for completing the target task and generate target assist parameter configuration information corresponding to the target task, and control an exoskeleton robot to provide assist to the user according to the target assist parameter configuration information.

[0027] In a third aspect, a computer program product is provided, which comprises computer programs / instructions that are executed by a processor to implement the steps of the exoskeleton robot control method with semantic perception function according to the first aspect.

[0028] In a fourth aspect, a computer readable storage medium is provided, which stores computer programs / instructions that are executed by a processor to implement the steps of the exoskeleton robot control method with semantic perception function according to the first aspect.

[0029] In a fifth aspect, an electronic device is provided, which comprises a memory, a processor, and computer programs stored in the memory and executable on the processor, and the processor implements the steps of the exoskeleton robot control method with semantic perception function according to the first aspect when executing the programs.

[0030] As can be seen from the above technical solutions, the present application uses a large language model to obtain semantic information from language description of a task to analyze assist to be provided for completing the task, and then generates corresponding assist parameter configuration information to control an exoskeleton robot to provide reasonable assist to the user. Thus, the present application realizes an exoskeleton robot control scheme with semantic perception function, which can improve the intelligence, autonomy and generalization of assist operation of the exoskeleton robot, and thus is conducive to improving the applicability of the exoskeleton robot in the field of rehabilitation and the like. Attached Figure Description

[0031] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 A flowchart illustrating the implementation of a control method for an exoskeleton robot with semantic awareness provided in this application embodiment;

[0033] Figure 2 A schematic diagram of a semantic awareness-assisted framework provided in an embodiment of this application;

[0034] Figure 3 A schematic diagram illustrating a semantic extraction and adjustment process provided in an embodiment of this application;

[0035] Figure 4 A schematic diagram of the structure of an exoskeleton robot control device with semantic awareness function provided in this application embodiment;

[0036] Figure 5 This is a schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0037] The technical solutions of 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. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0038] See Figure 1 The diagram shown is an implementation flowchart of an exoskeleton robot control method with semantic awareness provided in this application embodiment. The method may include the following steps:

[0039] Step S101: Receive language description information for the target task input by the user.

[0040] In practice, the system receives language description information (such as "pour me water") input by the user in the form of voice or text, targeting a specific task (such as getting a water cup). The system can perform preprocessing operations such as noise removal, speech recognition, or normalization to ensure the processing effect of the language description information by the subsequent large language model.

[0041] It should be noted that, unlike the motion control instructions that can be directly understood by the exoskeleton robot (such as how many degrees of joint movement, etc.), the task language description information described in the present application refers to high-level instruction such as "help me pour water" which cannot be directly understood by the exoskeleton robot. The present application introduces the high-level instruction into the control of the exoskeleton robot to improve the generalization of the assist operation, that is, to support the user to express his / her intention in a more diversified description manner (i.e., without using fixed motion control instructions) during the control of the exoskeleton robot. For example, the user can express the task he / she wants to complete in a diversified description manner in the task description information, and the user can further express the task execution related information such as the assist force he / she wants to get or the matters needing attention he / she wants to avoid (for example, avoiding water spilling out) in a diversified description manner in the task description information.

[0042] Step S102: performing semantic extraction on the language description information by a large language model to analyze the assist force required to complete the target task and generate target assist parameter configuration information corresponding to the target task.

[0043] For example, taking "help me pour water" as the language description information of the target task as an example, after inputting the language description information into a large language model (Large Language Model, LLM), the large language model will immediately perform semantic extraction and analysis to identify semantic information therein, such as identifying the semantic information that water may spill out and assist force should be provided gently, and then generating target assist parameter configuration information that conforms to the semantic information (i.e., supporting gentle assist force provision).

[0044] Step S103: controlling the exoskeleton robot to provide assist force to the user according to the target assist parameter configuration information.

[0045] In specific implementation, according to the target assist parameter configuration information, the assist parameter (i.e., the related parameter that can affect the motion process of the exoskeleton robot, such as affecting the assist trajectory or the joint torque of the robot) of the exoskeleton robot is automatically adjusted to control the exoskeleton robot to provide task-specific assist force to the user, such as providing task-specific assist force for holding a water cup.

[0046] In this embodiment, the present application implements an exoskeleton robot control scheme with semantic perception function, which can obtain semantic information from the language description of the task and guide the motion of the exoskeleton robot accordingly, thereby ensuring the efficiency and safety of the robot motion and improving the generalization of the control strategy, that is, supporting accurate control of the exoskeleton robot to provide task-specific assist force according to the diversified task language description of the user.

[0047] It can be seen from the above technical solutions that the application utilizes a large language model to obtain semantic information from language description of a task, analyzes assistance required for completing the task, and further generates corresponding assistance parameter configuration information to control an exoskeleton robot to reasonably provide assistance to the user. Thus, the application realizes an exoskeleton robot control scheme with semantic perception function, which can improve the intelligence, autonomy and generalization of assistance operation of the exoskeleton robot, and further improves the applicability of the exoskeleton robot in the field of rehabilitation and the like.

[0048] In an optional embodiment, the semantic extraction of the language description information by the large language model to analyze assistance required for completing the target task and the generation of the target assistance parameter configuration information corresponding to the target task include:

[0049] The semantic extraction of the language description information by the large language model obtains first information, which includes feature information of a task execution object associated with the target task and an action instruction associated with the target task.

[0050] The large language model determines second information according to the first information and a pre-learned association relationship between feature information of a task execution object and feature information of an action instruction, the second information including feature information of assistance required for completing the target task.

[0051] The large language model generates the target assistance parameter configuration information according to the second information and a pre-learned association relationship between feature information of assistance and assistance parameter configuration information.

[0052] In specific implementation, to identify semantic information in the environment (i.e., to obtain semantic information from language description of a task) and accurately generate corresponding assistance parameter configuration information, the large language model can first perform semantic extraction on the input language description information to obtain feature information of a task execution object associated with the target task, which can include but is not limited to what the task execution object is (such as a cup, an apple or a dumbbell, etc.) and / or attributes of the task execution object (such as whether it is heavy or fragile, etc.), and obtain an action instruction (such as grasping, etc.) associated with the target task. Optionally, in the case where a default action instruction is set (such as a default action instruction of grasping is pre-set), the large language model can directly determine the action instruction associated with the target task as the default action instruction to improve processing efficiency.

[0053] After obtaining the first information (i.e., the feature information of the task execution object associated with the target task, and the action instruction associated with the target task), the large language model determines the second information according to the first information, and (in the model fine-tuning stage) the pre-learned association relationship between the feature information of the task execution object, the action instruction, and the feature information of the assistance.

[0054] After obtaining the second information, the large language model can generate the target assistance parameter configuration information according to the second information and (in the model fine-tuning stage) the pre-learned association relationship between the feature information of the assistance and the assistance parameter configuration information.

[0055] Optionally, the large language model generates the target assistance parameter configuration information according to the second information and the pre-learned association relationship between the feature information of the assistance and the assistance parameter configuration information.

[0056] The large language model generates the target assistance parameter configuration information according to the second information and the pre-learned association relationship between the feature information of the assistance and the type and configuration value of the assistance parameter to be configured.

[0057] In specific implementation, the large language model determines which assistance parameters need to be adjusted and how to adjust according to the second information and (in the model fine-tuning stage) the pre-learned association relationship between the feature information of the assistance and the type and configuration value of the assistance parameter to be configured, and generates the target assistance parameter configuration information accordingly.

[0058] For example, in the case of needing to provide assistance gently, the large language model can determine that the related assistance parameters (such as impedance coefficient and speed) that affect the strength of assistance need to be adjusted according to the pre-learned knowledge (i.e., the association between the feature information of assistance and the type and configuration value of the required configuration assistance parameter), and further determine the configuration value of the related assistance parameter (which is used to represent binary configuration such as high or low impedance and fast or slow speed), and then take the configuration value of the related assistance parameter (such as low impedance and low speed) as the target assistance parameter configuration information to achieve providing assistance to the user gently. The correspondence between the configuration value of the assistance parameter and the specific value can be flexibly set according to actual needs, for example, fast can be set to represent that the maximum speed can reach 70 degrees / s, and slow can be set to represent that the maximum speed does not exceed 30 degrees / s.

[0059] In this embodiment, the present application designs a large language model to generate target assistance parameter configuration information according to the association between the feature information of assistance and the type and configuration value of the required configuration assistance parameter, to further ensure the accuracy of the provided assistance.

[0060] In an optional embodiment, the method further comprises:

[0061] In response to receiving the language description information input by the user for the target task, switching the working mode of the exoskeleton robot to the first working mode to control the exoskeleton robot to not provide assistance to the user;

[0062] In response to receiving the language description information input by the user for representing the action instruction associated with the user completing the target task (such as "grabbing a water cup", etc.), switching the working mode of the exoskeleton robot to the second working mode to control the exoskeleton robot to start providing assistance to the user.

[0063] In this embodiment, the present application realizes the working mode switching function based on semantic perception, which can provide assistance to the user in time at the appropriate time, thereby further improving the intelligent degree of the exoskeleton robot.

[0064] In an optional embodiment, the method further comprises:

[0065] According to the sensor parameters of the exoskeleton robot, determining an abnormal score of the exoskeleton robot, the abnormal score being used to represent the abnormal degree of human-machine contact;

[0066] In the case where the abnormal score is greater than a set score threshold, prompting the user to re-input the language description information for the target task, and re-generating target assistance parameter configuration information by a large language model according to the re-input language description information to adjust the assistance provided by the exoskeleton robot to the user.

[0067] For specific implementation, see Figure 2 The schematic diagram of the semantic perception-assisted framework illustrates that in the anomaly detection stage, the sensor parameters (such as encoder and force sensor parameters) X of the exoskeleton robot are input into a pre-trained diffusion model to obtain an anomaly score S. A higher S value indicates a more abnormal human-computer interaction. If S exceeds a set score threshold (e.g., 0.5), an anomaly in the human-computer interaction is determined. At this point, the process returns to the first step, which involves calling the Application Programming Interface (API) of a large language model (which can be a commercial large model) to regenerate the target parameter configuration information. This improves the reliability of the exoskeleton robot control.

[0068] In an optional embodiment, controlling the exoskeleton robot to provide assistance to the user based on the target assistance parameter configuration information includes:

[0069] The target assistance parameter configuration information is input into the controller of the exoskeleton robot to control the exoskeleton robot to provide assistance to the user;

[0070] The controller is pre-configured with a parameter adjustment interface for receiving auxiliary parameter configuration information generated by the large language model. The auxiliary parameter configuration information is represented by a combination of one or more bits, and different combinations of the one or more bits correspond to different auxiliary parameter configuration schemes.

[0071] In practical implementation, a parameter adjustment interface is reserved for the exoskeleton robot's controller. This interface receives the assist parameter configuration information (i.e., 0, 1 commands) output by the large language model, thereby simplifying the control process. The large language model output (0, 0) represents low impedance and low speed; (0, 1) represents low impedance and high speed, and so on. Different parameter configuration schemes for impedance coefficients and speeds can be represented by combinations of two bits, thus reducing the amount of information transmitted between the large language model and the controller, thereby ensuring control efficiency. It is understandable that... Figure 2 As shown, upon receiving the assist parameter configuration information P through the parameter adjustment interface, the controller enters the task execution phase. Then, based on the assist parameter configuration information P, it performs trajectory refinement to obtain the exoskeleton robot trajectory q. d Based on this, interactive control is then performed, which means outputting the motor torque u to control the exoskeleton robot to move and thus provide corresponding assistance to the user.

[0072] For example, see Figure 3 The flowchart shown illustrates the semantic extraction and assistance process. After receiving the user's input of language description information for the target task (such as "Help me get the water"), the system enters a thinking process to simultaneously execute the relevant steps in the Intent-integrative Grasping and Semantic-aware assistance stages.

[0073] Specifically, in the Intent-integrative Grasping stage, the system switches the exoskeleton robot's operating mode to the first operating mode (i.e., transparent mode, such as setting the mode parameter Mode=0) to control the exoskeleton robot to not provide assistance to the user, allowing the wearer (i.e., the user) to reach the task execution object associated with the target task with the exoskeleton robot, and to execute the action commands associated with the target task (such as grasping an object) independently. Optionally, the system can also perform anomaly monitoring simultaneously in this stage to initiate a replanning of the control strategy when an anomaly occurs (such as when the anomaly score exceeds a set score threshold).

[0074] In the Semantic-aware assistance phase, the system utilizes a large language model to extract and analyze the currently acquired language description information based on pre-learned knowledge (such as semantic differences between different task description information, predefined movements required by the exoskeleton robot to complete different tasks, etc.) to understand and infer human intentions. Based on this, it plans (or replans) control strategies for adjusting assistance, such as generating bit values ​​(i.e. configuration values) corresponding to the impedance parameter Imp and the assistance speed Sped, where a bit value of 1 indicates high and a bit value of 0 indicates low.

[0075] Subsequently, upon receiving language description information from the user that represents the action instructions associated with the user's completion of the target task, the working mode of the exoskeleton robot can be switched to the second working mode (e.g., setting the mode parameter Mode=1) so that the exoskeleton robot can start providing corresponding assistance to the user based on the target assistance parameter configuration information generated by Think process.

[0076] It should be noted that for the method embodiments, the series of acts performed are presented in sequential order for simplicity and clarity. However, it should be appreciated that some of the acts could be performed in other order, or performed concurrently. Moreover, it is to be understood that described acts can be carried out in sequence or in parallel by, for example, separate circuits, separate modules, or separate processors of a single digital circuitry. Further, some of the described acts can be stored in computer-readable media, which would result in execution of the acts by computer or digital processing device. Additionally, the scope of methods disclosed should not be limited by the language to "comprising," "containing," or "including" acts, but should be understood clearly to mean "comprising," "containing," or "including" at least the stated acts, but also including additional acts in various embodiments.

[0077] The application further provides a kind of exoskeleton robot control device with semantic perception function, as shown in Figure 4 The device comprises:

[0078] Communication module, for receiving the language description information of target task input by user;

[0079] Processing module, for extracting the semantic of the language description information by large language model, to analyze the help required to complete the target task and generate the target help parameter configuration information corresponding to the target task;And for controlling exoskeleton robot to provide help to the user according to the target help parameter configuration information.

[0080] Optionally, the processing module further comprises the following steps:

[0081] Extract the semantic of the language description information by the large language model, obtain first information, and the first information includes: the characteristic information of the task execution object associated with the target task, and the action instruction associated with the target task;

[0082] According to the first information, the characteristic information of the task execution object, the action instruction and the characteristic information of the help pre-learned by the large language model, determine second information, and the second information includes: the characteristic information of the help required to complete the target task;

[0083] According to the second information, the characteristic information of the help and the association relationship between the help parameter configuration information pre-learned by the large language model, generate the target help parameter configuration information.

[0084] Optionally, the processing module further comprises the following steps:

[0085] Optionally, the processing module further comprises the following steps:

[0086] in response to receiving the language description information input by the user for the target task, switching the working mode of the exoskeleton robot to a first working mode to control the exoskeleton robot to not provide assistance to the user;

[0087] in response to receiving the language description information input by the user for the action instruction associated with the user completing the target task, switching the working mode of the exoskeleton robot to a second working mode to control the exoskeleton robot to start providing assistance to the user.

[0088] Optionally, the processing module is further configured to perform the following steps:

[0089] determining an abnormality score of the exoskeleton robot according to the sensor parameters of the exoskeleton robot, the abnormality score being used to represent the degree of abnormality of the human-robot contact;

[0090] in a case where the abnormality score is greater than a set score threshold, prompting the user to re-input the language description information for the target task, and re-generating target assistance parameter configuration information according to the re-input language description information by using a large language model to adjust the assistance provided by the exoskeleton robot to the user.

[0091] Optionally, the processing module is further configured to input the target assistance parameter configuration information to a controller of the exoskeleton robot to control the exoskeleton robot to provide assistance to the user; wherein the controller is pre-configured with a parameter adjustment interface for receiving assistance parameter configuration information generated by a large language model, the assistance parameter configuration information being represented using one or more combinations of bits, and different combinations of the one or more bits corresponding to different assistance parameter configuration schemes.

[0092] As can be seen from the above technical solutions, the present application uses a large language model to obtain semantic information from the language description of a task to analyze the assistance required to complete the task, and then generates corresponding assistance parameter configuration information to control the exoskeleton robot to reasonably provide assistance to the user. Thus, the present application realizes an exoskeleton robot control scheme with semantic perception function, which can improve the intelligence, autonomy and generalization of the assistance operation of the exoskeleton robot, and thus is conducive to improving the applicability of the exoskeleton robot in the field of rehabilitation and the like.

[0093] The present application also provides an electronic device, which refers to Figure 5 , Figure 5 is a schematic diagram of an electronic device according to an embodiment of the present application. As shown in Figure 5As shown, the electronic device 100 includes a memory 110 and a processor 120, the memory 110 and the processor 120 are connected through a bus in communication, the memory 110 stores a computer program, the computer program can run on the processor 120, and then the steps in the exoskeleton robot control method with semantic perception function disclosed in the embodiment of the application are realized.

[0094] The embodiment of the application further provides a computer readable storage medium, which stores a computer program / instruction, and the computer program / instruction is executed by a processor to realize the exoskeleton robot control method with semantic perception function disclosed in the embodiment of the application.

[0095] The embodiment of the application further provides a computer program product, which includes a computer program / instruction, and the computer program / instruction is executed by a processor to realize the exoskeleton robot control method with semantic perception function disclosed in the embodiment of the application.

[0096] Each embodiment in the specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same and similar parts between each embodiment can be referred to each other.

[0097] Those skilled in the art should understand that the embodiments of the application can be provided as a method, device or computer program product. Therefore, the embodiments of the application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of the application can adopt the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0098] The embodiments of the application are described with reference to flowcharts and / or block diagrams according to the method, system, device, storage medium and program product of the embodiments of the application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be realized by computer program instructions. These computer program instructions can be provided to a general purpose computer, a special purpose computer, an embedded processor or other programmable data processing terminal equipment to produce a machine, so that the instructions executed by the computer or other programmable data processing terminal equipment produce a device for realizing the functions specified in the flowcharts and / or block diagrams. Figure 1 The device for realizing the functions specified in one flow or multiple flows and / or blocks Figure 1 The device for realizing the functions specified in one flow or multiple flows and / or blocks

[0099] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow Figure 1 one or more flows and / or blocks Figure 1 one or more blocks or multiple blocks.

[0100] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 one or more flows and / or blocks Figure 1 one or more blocks or multiple blocks.

[0101] Finally, it should be noted that the terms "first" and "second" and the like are used merely to distinguish one element from another, and do not necessarily indicate a physical or chronological priority of one element over another. Furthermore, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. The terms "includes", "including", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0102] The above provides a control method of an exoskeleton robot with semantic perception function, and the principle and implementation manner of the present application are described by using specific examples. The above description of the embodiments is only used to help understand the method and core idea of the present application. For those skilled in the art, according to the idea of the present application, the specific implementation manner and application range can be changed, and the above description of the present application should not be understood as a limitation.

Claims

1. A control method for an exoskeleton robot with semantic perception function, characterized in that, The method includes: Receive user input of language description information specific to the target task; Using a large language model, semantic extraction is performed on the language description information to analyze the assistance required to complete the target task and generate target assistance parameter configuration information corresponding to the target task. Based on the target assistance parameter configuration information, the exoskeleton robot is controlled to provide assistance to the user. The process involves semantic extraction of the language description information using a large language model to analyze the assistance required to complete the target task and generate target assistance parameter configuration information corresponding to the target task, including: The language description information is semantically extracted using the large language model to obtain first information, which includes: feature information of the task execution object associated with the target task, and action instructions associated with the target task. Based on the first information and the correlation between the pre-learned feature information of the task execution object, the action instructions and the feature information of the assistance, the second information is determined using the large language model. The second information includes the feature information of the assistance required to complete the target task. Based on the second information and the correlation between the pre-learned feature information of the assist and the assist parameter configuration information, the target assist parameter configuration information is generated using the large language model.

2. The method according to claim 1, characterized in that, The step of generating the target assist parameter configuration information through the large language model, based on the second information and the correlation between pre-learned assist feature information and assist parameter configuration information, includes: Based on the second information and the correlation between the pre-learned assist feature information and the type and configuration value of the required assist parameters, the target assist parameter configuration information is generated using the large language model.

3. The method according to claim 1, characterized in that, The method further includes: In response to receiving a language description of the target task input by the user, the working mode of the exoskeleton robot is switched to the first working mode to control the exoskeleton robot not to provide assistance to the user. In response to receiving language description information input by the user that represents the action instructions associated with the user's completion of the target task, the working mode of the exoskeleton robot is switched to a second working mode to control the exoskeleton robot to begin providing assistance to the user.

4. The method according to claim 1, characterized in that, The method further includes: Based on the sensor parameters of the exoskeleton robot, an anomaly score of the exoskeleton robot is determined, and the anomaly score is used to characterize the degree of anomaly in human-robot contact; If the abnormal score exceeds a set score threshold, the user is prompted to re-enter the language description information for the target task. Based on the re-entered language description information, the large language model regenerates the target assistance parameter configuration information to adjust the assistance provided by the exoskeleton robot to the user.

5. The method according to claim 1, characterized in that, The step of controlling the exoskeleton robot to provide assistance to the user based on the target assistance parameter configuration information includes: The target assistance parameter configuration information is input into the controller of the exoskeleton robot to control the exoskeleton robot to provide assistance to the user; The controller is pre-configured with a parameter adjustment interface for receiving auxiliary parameter configuration information generated by the large language model. The auxiliary parameter configuration information is represented by a combination of one or more bits, and different combinations of the one or more bits correspond to different auxiliary parameter configuration schemes.

6. A control device for an exoskeleton robot with semantic perception function, characterized in that, The device includes: The communication module is used to receive user-input language description information for the target task; The processing module is used to perform semantic extraction on the language description information through a large language model, analyze the assistance required to complete the target task, generate target assistance parameter configuration information corresponding to the target task, and control the exoskeleton robot to provide assistance to the user according to the target assistance parameter configuration information. The processing module is further configured to perform the following steps: The language description information is semantically extracted using the large language model to obtain first information, which includes: feature information of the task execution object associated with the target task, and action instructions associated with the target task. Based on the first information and the correlation between the pre-learned feature information of the task execution object, the action instructions and the feature information of the assistance, the second information is determined using the large language model. The second information includes the feature information of the assistance required to complete the target task. Based on the second information and the correlation between the pre-learned feature information of the assist and the assist parameter configuration information, the target assist parameter configuration information is generated using the large language model.

7. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the exoskeleton robot control method with semantic awareness as described in any one of claims 1 to 5.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the exoskeleton robot control method with semantic awareness as described in any one of claims 1 to 5.

9. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the exoskeleton robot control method with semantic awareness as described in any one of claims 1 to 5.

Images