Robot control method, apparatus, device, medium, and product

Abstract

This application discloses a robot control method, apparatus, device, medium, and product, belonging to the field of computer technology. The robot control method in this application predicts the initial action by acquiring the robot's current motion state characteristics, and acquires the robot's actual state during execution. Based on the deviation between the predicted state and the actual state, it determines the error correction amount, and then corrects the initial action to obtain the target action, thus achieving an organic combination of open-loop action prediction and closed-loop execution feedback. This method can autonomously and online correct and compensate for execution deviations during long sequence executions, effectively suppressing the gradual accumulation of errors over time and significantly improving the stability and accuracy of robot control. Furthermore, because the correction mechanism is independent of the initial action prediction model and does not rely on the prediction accuracy of a single model, it possesses stronger generalization ability and robustness in dynamic or unstructured environments.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to robot control methods, devices, equipment, media and products. Background Technology

[0002] In robot imitation learning based on the LeRobot framework and Action Block Transformation (ACT) model, end-to-end action generation strategies are typically trained by collecting expert demonstration data, enabling the robot to reproduce taught behaviors in specific tasks. However, under this open-loop control paradigm, when the model generates long-term action sequences, the prediction error accumulates gradually over time, and the system lacks closed-loop feedback and correction capabilities for actual state deviations during execution, resulting in a significant decrease in the stability of action execution in long-sequence tasks. At the same time, existing methods have a relatively shallow fusion of multimodal perception data such as vision, force, and joint states, making it difficult to fully explore the temporal correlations and complementary characteristics between different modalities, further restricting the robot's control accuracy and generalization performance in dynamic or unstructured environments. Summary of the Invention

[0003] To address the limitations of existing imitation learning methods in terms of error accumulation, control accuracy, and generalization ability, this paper presents a method, apparatus, equipment, medium, and product that improves robot control stability, accuracy, and generalization ability by constructing an independent closed-loop feedback correction mechanism to achieve online autonomous correction of action execution deviations.

[0004] This application provides a robot control method, including:

[0005] Acquire state features that indicate the robot's current motion state;

[0006] Based on the state characteristics, the robot is predicted to perform an initial action to obtain the robot's initial action. The initial action is used to indicate the robot's predicted action and corresponding predicted state.

[0007] The actual state of the robot is obtained, and the actual state is used to indicate the actual state of the robot during the execution of an action;

[0008] Based on the deviation between the predicted state and the actual state, an error correction amount is determined, which is used to correct and compensate for the initial action.

[0009] The initial action is corrected based on the error correction amount to obtain the target action;

[0010] Based on the target action, the robot is controlled to perform the action.

[0011] Optionally, determining the error correction amount based on the deviation between the predicted state and the actual state includes:

[0012] An explicit error model is constructed, and a systematic error is determined based on the deviation between the predicted state and the actual state. The systematic error is used to quantitatively characterize the actual deviation during the robot's action execution.

[0013] Based on the system error, the error correction amount is determined.

[0014] Optionally, the step of constructing an explicit error model, based on the deviation between the predicted state and the actual state, to determine the system error includes:

[0015] ,

[0016] in, Let be the systematic error at time t. Let be the predicted state at time t. The actual state at time t;

[0017] Alternatively, the predicted action can be mapped to the state space using a robot kinematics model, whereby the system error is defined as:

[0018] ,

[0019] in, The predicted action is the action corresponding to the initial action at time t. This represents the robot's forward kinematics or dynamics mapping function.

[0020] Optionally, determining the error correction amount based on the system error includes:

[0021] An error-driven dynamic compensation mechanism is adopted to generate an error correction amount based on the system error.

[0022] Optionally, the step of correcting the initial action based on the error correction amount to obtain the target action includes:

[0023] Obtain the fusion coefficient corresponding to the error correction amount, wherein the value of the fusion coefficient is positively correlated with the value of the deviation;

[0024] Based on the fusion coefficient, the error correction amount is fused with the initial action to obtain the target action.

[0025] Optionally, after controlling the robot to perform an action based on the target action, the method further includes:

[0026] Obtain the updated state of the robot after it performs the target action;

[0027] The updated state is used for the initial action prediction of subsequent time steps to form a closed-loop optimization process consisting of prediction, execution, feedback, and correction, thereby suppressing the accumulation of errors during the execution of long action sequences.

[0028] This application provides a robot control device, including:

[0029] The state acquisition module is used to acquire state features that indicate the robot's current motion state;

[0030] An initial action prediction module is used to predict the initial action of the robot based on the state features, thereby obtaining the initial action of the robot. The initial action is used to indicate the predicted action and the corresponding predicted state of the robot.

[0031] The actual state acquisition module is used to acquire the actual state of the robot, and the actual state is used to indicate the actual state of the robot during the execution of an action.

[0032] An error correction module is used to determine an error correction amount based on the deviation between the predicted state and the actual state, and the error correction amount is used to correct and compensate the initial action.

[0033] An action correction module is used to correct the initial action based on the error correction amount to obtain the target action;

[0034] The control module is used to control the robot to perform actions based on the target action.

[0035] This application provides an electronic device, the electronic device comprising:

[0036] One or more processors; and a memory storing computer program instructions that, when executed, cause the processors to perform the steps of the method described above.

[0037] This application provides a computer-readable medium having a computer program / instructions stored thereon, which, when executed by a processor, implement the steps of the above-described method.

[0038] This application provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described method.

[0039] The beneficial effects of the above technical solution are as follows:

[0040] In this application, the robot control method predicts the initial action by acquiring the robot's current motion state characteristics and obtains the robot's actual state during execution. Based on the deviation between the predicted and actual states, it determines the error correction amount and then corrects the initial action to obtain the target action, thus achieving an organic combination of open-loop action prediction and closed-loop execution feedback. This method can autonomously and online correct and compensate for execution deviations during long sequence executions, effectively suppressing the gradual accumulation of errors over time and significantly improving the stability and accuracy of robot control. Furthermore, because the correction mechanism is independent of the initial action prediction model and does not rely on the prediction accuracy of a single model, it possesses stronger generalization ability and robustness in dynamic or unstructured environments. Attached Figure Description

[0041] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0042] Figure 1 This is a flowchart of one embodiment of the robot control method described in this application;

[0043] Figure 2 A flowchart illustrating an embodiment of a method for determining the error correction amount for this application;

[0044] Figure 3 This is a flowchart illustrating another embodiment of the robot control method described in this application.

[0045] Figure 4 This is a block diagram of one embodiment of the robot control device described in this application;

[0046] Figure 5 This is an exemplary structural diagram of the electronic device of this application. Detailed Implementation

[0047] The advantages of this application are further illustrated below with reference to the accompanying drawings and specific embodiments.

[0048] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0049] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0050] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0051] In the description of this application, it should be understood that the numerical labels before the steps do not indicate the order of the steps, but are only used to facilitate the description of this application and to distinguish each step, and therefore should not be construed as a limitation of this application.

[0052] The following terms are used in this application:

[0053] LeRobot: An open-source framework for robot learning and control, supporting imitation learning, reinforcement learning, and the deployment of multimodal control strategies.

[0054] ACT (Action Chunking with Transformers): A Transformer model for robot behavior generation that improves control stability by dividing continuous actions into action blocks for prediction.

[0055] like Figure 1 As shown, this embodiment provides a robot control method, including the following steps:

[0056] S1. Obtain the state features used to indicate the robot's current motion state;

[0057] S2. Based on the state characteristics, perform initial action prediction on the robot to obtain the initial action of the robot, wherein the initial action is used to indicate the predicted action of the robot and the corresponding predicted state;

[0058] S3. Obtain the actual state of the robot, wherein the actual state is used to indicate the actual state of the robot during the execution of an action;

[0059] S4. Determine an error correction amount based on the deviation between the predicted state and the actual state, the error correction amount being used to correct and compensate for the initial action;

[0060] S5. Correct the initial action based on the error correction amount to obtain the target action;

[0061] S6. Based on the target action, control the robot to perform the action.

[0062] In this embodiment, the robot control method performs initial motion prediction by acquiring the robot's current motion state characteristics and acquires the robot's actual state during execution. It determines the error correction amount based on the deviation between the predicted and actual states, and then corrects the initial motion to obtain the target motion, thus achieving an organic combination of open-loop motion prediction and closed-loop execution feedback. This method can autonomously and online correct and compensate for execution deviations during long sequence executions, effectively suppressing the gradual accumulation of errors over time and significantly improving the stability and accuracy of robot control. Furthermore, because the correction mechanism is independent of the initial motion prediction model and does not rely on the prediction accuracy of a single model, it possesses stronger generalization ability and robustness in dynamic or unstructured environments.

[0063] In an alternative embodiment, such as Figure 2 As shown, step S4 includes:

[0064] S41. Construct an explicit error model, and determine the system error based on the deviation between the predicted state and the actual state. The system error is used to quantitatively characterize the actual deviation during the robot's action execution.

[0065] S42. Based on the system error, determine the error correction amount.

[0066] In this embodiment, by constructing an explicit error model, a quantitative comparison is made directly between the actual state collected during robot execution and the expected state predicted by the model. This allows for the acquisition of accurate system error values, providing a reliable quantitative basis for subsequent action correction. Compared to implicit correction methods in existing technologies that rely on internal model inference or external human intervention, this explicit error modeling mechanism is independent of the initial action prediction process. It can autonomously perceive execution deviations without changing the original model parameters, making error assessment more objective and response more timely.

[0067] Furthermore, the systematic error in step S41 is:

[0068] ,

[0069] in, Let be the systematic error at time t. Let be the predicted state at time t. The actual state at time t;

[0070] Alternatively, the predicted action can be mapped to the state space using a robot kinematics model, whereby the system error is defined as:

[0071] ,

[0072] in, The predicted action is the action corresponding to the initial action at time t. This represents the robot's forward kinematics or dynamics mapping function.

[0073] In this embodiment, the error model supports two complementary computational paths: directly solving for the deviation based on the state space, or mapping the predicted action to the state space through a kinematic / dynamic model before calculating the error. This allows for flexible adaptation to different robot platforms and sensor configurations. This error quantification method with clear physical meaning lays the technical foundation for subsequent error-driven adaptive compensation mechanisms, making the closed-loop correction process more interpretable and controllable, and effectively improving the system's execution accuracy and stability margin in long-sequence tasks.

[0074] In an optional embodiment, step S42 may include:

[0075] An error-driven dynamic compensation mechanism is adopted to generate an error correction amount based on the system error. This allows for online correction of the initial action;

[0076] Wherein, the error correction amount Represented as:

[0077] ,

[0078] in, The adaptive feedback gain coefficient at time t is used to adjust the degree of influence of the system error on the action correction.

[0079] In this embodiment, by employing an error-driven dynamic compensation mechanism, the system can automatically generate corresponding error correction amounts based on the real-time calculated system error, thereby correcting the initial predicted actions online. This mechanism enables the robot to autonomously address execution deviations caused by factors such as model prediction bias, environmental disturbances, or uncertainties in the actuator during execution, without relying on external human intervention or offline retraining. This effectively suppresses the gradual accumulation of errors in long sequences of actions, significantly improving the stability and continuity of task execution.

[0080] Furthermore, the adaptive feedback gain coefficient A dynamic adjustment strategy is adopted, and the expression of the dynamic adjustment strategy is:

[0081] ,

[0082] in, The norm representing the systematic error. and All parameters are preset. The control intensity is adaptively adjusted based on the error amplitude of the system error, so as to enhance the correction capability when the error amplitude is large and avoid over-adjustment when the error amplitude is small.

[0083] In this embodiment, the introduction of an adaptive feedback gain coefficient enables the correction intensity to be dynamically adjusted according to the error amplitude: when the deviation is large, the system automatically increases the correction intensity to quickly correct the deviation and restore execution accuracy; when the deviation is small, the correction intensity is reduced accordingly to avoid unnecessary motion jitter or overshoot due to over-adjustment. This adaptive response characteristic to the error amplitude balances the sensitivity and smoothness of control, enabling the robot to exhibit stronger robustness and adaptability in dynamic or unstructured environments.

[0084] In another alternative embodiment, step S42 may include:

[0085] A learning-based error compensation model is introduced, and an error compensation function is constructed. The error compensation function is used to generate an error correction amount based on the current system error and the actual state of the robot:

[0086] ,

[0087] The corrected target action Represented as:

[0088] ,

[0089] Wherein, the error compensation function It is obtained by training with historical execution data and is used to model complex nonlinear errors to improve the system's adaptability in complex environments.

[0090] In this embodiment, through a learned error compensation model, the system can utilize the error compensation function trained on historical execution data to comprehensively model the current system error and the robot's actual state, thereby generating a more accurate error correction. Compared to linear compensation mechanisms, this learning model can capture complex nonlinear error patterns, such as joint friction, load changes, and environmental interaction forces—factors that are difficult to model explicitly. This makes the corrected target action more closely match actual execution requirements, significantly improving the robot's control accuracy and adaptability in complex or unstructured environments. Furthermore, the error compensation function can be continuously optimized as execution data accumulates, enabling the system to learn from experience. The compensation strategy becomes increasingly accurate as the task progresses, thereby continuously enhancing the execution stability and robustness in long-sequence tasks.

[0091] In an optional embodiment, step S5 may include:

[0092] Obtain the fusion coefficient corresponding to the error correction amount, wherein the value of the fusion coefficient is positively correlated with the value of the deviation;

[0093] Based on the fusion coefficient, the error correction amount is fused with the initial action to obtain the target action.

[0094] In this embodiment, by introducing a fusion coefficient positively correlated with the deviation value, the system can adaptively adjust the weight of the error correction in the final target action according to the severity of the actual execution deviation. When the deviation between the predicted state and the actual state is large, the fusion coefficient increases accordingly, strengthening the adjustment of the initial action by the correction, thereby quickly correcting the execution trajectory and restoring control accuracy. When the deviation is small, the fusion coefficient automatically decreases, allowing the initial action to be largely preserved and avoiding unnecessary motion fluctuations caused by over-correction. This adaptive fusion mechanism based on the deviation amplitude effectively balances the speed of correction and the smoothness of execution, enabling the robot's motion output in dynamic environments to possess both stability and flexibility.

[0095] In an alternative embodiment, such as Figure 3 As shown, after step S6, the robot control method may further include the following steps:

[0096] S7. Obtain the updated state of the robot after it performs the target action;

[0097] S8. Use the updated state for the initial action prediction of subsequent time steps to form a closed-loop optimization process consisting of prediction, execution, feedback, and correction, thereby suppressing the accumulation of errors during the execution of long action sequences.

[0098] In this embodiment, after each target action is executed, the system obtains the robot's updated state and feeds it back to the initial action prediction stage of the next time step, thus forming a closed-loop optimization process of "prediction-execution-feedback-correction". By continuously introducing the corrected actual execution results into subsequent predictions, the system can dynamically perceive and compensate for the residual deviation of each action step, breaking the inherent defect of the prediction error monotonically accumulating with the step size in the traditional open-loop method, and making the action execution in long sequence tasks continuously stable.

[0099] In this embodiment, the closed-loop optimization process can be specifically described as follows: First, a predicted sequence of actions for multiple future time steps is generated based on the current observed state; second, the system executes the action at the current moment and acquires the robot's actual state in real time; subsequently, the deviation between the predicted state and the actual state is calculated through the error feedback module; further, the action compensation amount is generated using the adaptive correction module to correct the current action in real time; finally, the corrected state is re-inputted into the model for subsequent action prediction. Through the above-described "prediction-execution-feedback-correction" cyclic process, the accumulation of errors in long-term sequences is effectively suppressed, improving the system's stability and robustness.

[0100] Furthermore, this closed-loop process enables the robot to respond to and self-recover from sudden disturbances in dynamic or unstructured environments. Even if a momentary execution deviation occurs, it can be quickly corrected in subsequent steps, preventing the error from propagating along the timeline. This continuous iterative feedback correction mechanism effectively improves the system's robustness and the success rate of task execution.

[0101] In an optional embodiment, step S2 includes:

[0102] The robot's initial action is predicted based on an imitation learning model, which is trained on expert demonstration data.

[0103] In this embodiment, initial action prediction is achieved through an imitation learning model. This model, trained on expert demonstration data, can directly map predicted actions and corresponding predicted states that conform to the task objectives from state features. Because the imitation learning model learns the inherent rules and action distribution of task execution from expert demonstrations, the robot can quickly obtain high-quality initial action sequences without relying on manually written complex control rules or extensive trial-and-error exploration. This significantly reduces the time cost of deploying new tasks while ensuring the naturalness and rationality of the initial actions, laying a solid foundation for subsequent closed-loop correction. The initial prediction capability based on expert data training enables the robot to output reasonable actions that conform to the task semantics even when facing unseen environmental changes, thereby enhancing the generalization starting point of the entire control framework.

[0104] Specifically, the initial action prediction of the robot based on the imitation learning model to obtain the initial action includes:

[0105] The imitation learning model includes a hierarchical action modeling submodule and a two-layer Transformer submodule;

[0106] The hierarchical action modeling submodule divides a continuous action sequence into action blocks of different granularities. These action blocks include coarse-grained action blocks and fine-grained action blocks. The coarse-grained action blocks are used to represent the global planning of the action sequence, while the fine-grained action blocks are used to represent the local execution details of the action sequence.

[0107] The two-layer Transformer submodule generates the predicted actions and corresponding predicted states based on the action blocks of different granularities. The two-layer Transformer submodule includes an upper-layer global planning network and a lower-layer local generation network. The upper-layer global planning network is used for global action planning, and the lower-layer local generation network is used for generating specific control actions.

[0108] In this embodiment, a hierarchical action modeling submodule decomposes continuous action sequences into coarse-grained and fine-grained action blocks, enabling the model to simultaneously consider global task planning and local execution details, effectively enhancing the consistency of long sequence actions. The two-layer Transformer submodule, through the collaboration of an upper-layer global planning network and a lower-layer local generation network, is responsible for high-level policy generation and low-level action implementation, respectively, allowing the model to better handle the hierarchical characteristics of complex tasks.

[0109] Furthermore, the imitation learning model also includes a temporal position augmentation encoding submodule, which enhances the model's ability to model long-term sequential dependencies. By introducing temporal position augmentation encoding, the model can more accurately capture long-distance temporal dependencies in action sequences, further improving the consistency and coherence of action prediction in long-sequence tasks.

[0110] In an optional embodiment, step S1 includes:

[0111] Acquire multi-source perception data during robot operation, including visual image data, depth information, joint state data, and force feedback data;

[0112] The multi-source sensing data is subjected to time synchronization, noise reduction and normalization.

[0113] The multi-source sensing data of different modalities are encoded into a unified feature representation to obtain the state features.

[0114] In this embodiment, multi-source perception data, including visual images, depth information, joint states, and force feedback, are collected. These heterogeneous data are then time-synchronized, denoised, and normalized before being encoded into a unified state feature representation. This enables the robot to perceive its own motion state and external environmental information from multiple dimensions. This multimodal data fusion method effectively compensates for the limitations of single sensors in terms of perception range, accuracy, or robustness, providing richer and more consistent state inputs for subsequent initial action prediction, thus improving the accuracy of action prediction and the reliability of task execution. Simultaneously, the unified state feature representation eliminates modal differences between different data sources, allowing the model to more fully exploit the temporal correlations and complementary characteristics between multimodal information, further enhancing the robot's perception robustness and control generalization ability in dynamic or unstructured environments.

[0115] The robot control method presented in this application does not merely optimize the model structure; instead, it introduces an independent error feedback and adaptive correction mechanism outside the model to construct a closed-loop control system oriented towards the actual execution process of the robot. During action execution, the robot's current state is acquired in real time and compared with the prediction results to dynamically calculate the execution error. Based on this error, subsequent actions are corrected online, enabling the system to continuously make adaptive adjustments during execution. This mechanism directly affects the robot control process and effectively suppresses the accumulation of errors over time. Compared to open-loop methods that indirectly reduce errors by enhancing the model's expressive power, this application introduces an explicit feedback adjustment mechanism to achieve direct control and correction of execution errors, without relying on the prediction accuracy of a single model, thus improving performance from the control principle perspective.

[0116] Furthermore, this application determines the error correction amount based on the deviation and corrects the initial action based on the error correction amount. This correction mechanism is independent of the initial action prediction model. Therefore, the technical solution of this application is not only applicable to the initial action prediction model based on imitation learning, but can also be used in combination with other types of action prediction models, and has wide applicability.

[0117] like Figure 4 As shown, this embodiment provides a robot control device, including:

[0118] State acquisition module 1 is used to acquire state features that indicate the current motion state of the robot;

[0119] The initial action prediction module 2 is used to predict the initial action of the robot based on the state features, so as to obtain the initial action of the robot. The initial action is used to indicate the predicted action of the robot and the corresponding predicted state.

[0120] The actual state acquisition module 3 is used to acquire the actual state of the robot, and the actual state is used to indicate the actual state of the robot during the execution of an action.

[0121] Error correction module 4 is used to determine the error correction amount based on the deviation between the predicted state and the actual state, and the error correction amount is used to correct and compensate the initial action;

[0122] The motion correction module 5 is used to correct the initial motion based on the error correction amount to obtain the target motion;

[0123] Control module 6 is used to control the robot to perform actions based on the target action.

[0124] In this embodiment, the state acquisition module 1 collects and encodes unified state features from multiple sensors. The initial action prediction module 2 generates an initial action containing the predicted action and corresponding predicted state based on these features. The actual state acquisition module 3 collects the robot's real-time state during execution. The error correction module 4 determines the error correction amount by comparing the deviation between the predicted state and the actual state. The action correction module 5 uses this correction amount to correct the initial action online to obtain the target action. Finally, the control module 6 drives the robot to execute the action. These modules work collaboratively to form a closed-loop control chain of "prediction-execution-feedback-correction," enabling the system to autonomously perceive and compensate for deviations in each execution step. This effectively suppresses error accumulation in long-sequence tasks and significantly improves control stability and accuracy. Furthermore, the error correction module 4 is independent of the initial action prediction module 2, and the correction process does not rely on retraining or parameter adjustment of the prediction model. This gives the device stronger generalization ability and robustness when facing dynamic or unstructured environmental changes.

[0125] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units are absent in this embodiment.

[0126] Furthermore, some embodiments of this application also provide an electronic device. The electronic device can be various forms of digital computer, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, etc. The electronic device can also be various forms of mobile devices, such as cellular phones, smartphones, wearable devices, and other similar computing devices.

[0127] The electronic device includes: one or more processors; and a memory storing computer program instructions that, when executed, cause the processor to perform the steps of the methods provided in any one or more of the above embodiments. Figure 5 An exemplary structural diagram of the electronic device is disclosed. The electronic device includes one or more processors 1101, a memory 1102, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components are interconnected via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the electronic device, including instructions stored in or on memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In some other embodiments, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple electronic devices can be connected, each providing some of the necessary operations. The components, their connections and relationships, and their functions shown herein are merely examples and are not intended to limit the implementation of the present application described and / or claimed herein.

[0128] The electronic device may further include an input device 1103 and an output device 1104. The processor 1101, memory 1102, input device 1103 and output device 1104 may be connected by a bus or other means, as shown in the figure, which is connected by a bus.

[0129] Input device 1103 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the electronic device, such as a touch screen, keypad, mouse, trackpad, touchpad, joystick, one or more mouse buttons, trackball, joystick, etc. Output device 1104 may include a display device, auxiliary lighting device (e.g., LED), and haptic feedback device (e.g., vibration motor). The display device may include, but is not limited to, a liquid crystal display, a light-emitting diode display, and a plasma display. In some embodiments, the display device may be a touch screen.

[0130] To provide interaction with the user, the electronic device can be a computer. The computer has: a display device (e.g., a cathode ray tube or LCD monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback); and input from the user can be received in any form (e.g., voice input or tactile input).

[0131] In this embodiment, a computer-readable medium stores a computer program / instructions that, when executed by a processor, implement the steps of the methods provided in any one or more of the above embodiments. This computer-readable medium may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into that device. The aforementioned computer-readable medium carries one or more computer-readable instructions.

[0132] The memory 1102 can serve as a non-transitory computer-readable storage medium, used to store non-transitory software programs, non-transitory computer-executable programs, and modules. The processor 1101 executes various functional applications and data processing of the server by running the non-transitory software programs, instructions, and modules stored in the memory 1102, thereby implementing the program instructions / modules corresponding to the methods provided in any one or more of the embodiments described above in this application.

[0133] The memory 1102 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, the memory 1102 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 1102 may optionally include memory remotely located relative to the processor 1101, and these remote memories can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0134] It should be noted that the computer-readable medium described in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. Computer-readable media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, electrical connections having one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, optical fibers, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0135] Computer-readable media include permanent and non-permanent, removable and non-removable media, which can store information by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory, static random access memory, dynamic random access memory, other types of random access memory, read-only memory, electrically erasable programmable read-only memory, flash memory or other memory technologies, read-only optical discs, digital versatile optical discs or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0136] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including local area networks (LANs) or wide area networks (WANs), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0137] In the above embodiments, all or part of the implementation can be achieved through software, hardware, firmware, or any combination thereof. For example, it can be implemented using an application-specific integrated circuit (ASIC), a general-purpose computer, or any other similar hardware device. In some embodiments, the software program of this application can be executed by a processor to implement the above steps or functions. Similarly, the software program of this application (including related data structures) can be stored in a computer-readable recording medium, such as RAM memory, magnetic or optical drives, floppy disks, and similar devices. In addition, some steps or functions of this application can be implemented in hardware, for example, as circuitry that cooperates with a processor to perform the various steps or functions.

[0138] The computer program product provided in this application includes one or more computer programs / instructions. When executed by a processor, these computer programs / instructions generate, in whole or in part, the processes or functions described in this application. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0139] The flowcharts or block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of devices, 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 some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated 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 diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-specific system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0140] The scope of this application is defined by the appended claims rather than the foregoing description, and is therefore intended to encompass all variations falling within the meaning and scope of equivalents of the claims. No reference numerals in the claims should be construed as limiting the scope of the claims. Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a device claim may also be implemented by a single unit or device in software or hardware. Terms such as "first," "second," etc., are used only for distinguishing descriptions and do not indicate any particular order, nor should they be construed as indicating or implying relative importance.

[0141] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily made 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. Therefore, the scope of protection of this application should be determined by the scope of the claims, and the above embodiments should be regarded as exemplary and non-limiting.

Claims

1. A robot control method, characterized in that, The method includes: Acquire state features that indicate the robot's current motion state; Based on the state characteristics, the robot is predicted to perform an initial action to obtain the robot's initial action. The initial action is used to indicate the robot's predicted action and corresponding predicted state. The actual state of the robot is obtained, and the actual state is used to indicate the actual state of the robot during the execution of an action; Based on the deviation between the predicted state and the actual state, an error correction amount is determined, which is used to correct and compensate for the initial action. The initial action is corrected based on the error correction amount to obtain the target action; Based on the target action, the robot is controlled to perform the action.

2. The robot control method according to claim 1, characterized in that, The step of determining the error correction amount based on the deviation between the predicted state and the actual state includes: An explicit error model is constructed, and a systematic error is determined based on the deviation between the predicted state and the actual state. The systematic error is used to quantitatively characterize the actual deviation during the robot's action execution. Based on the system error, the error correction amount is determined.

3. The robot control method according to claim 2, characterized in that, The construction of the explicit error model, based on the deviation between the predicted state and the actual state, determines the system error, including: ， in, Let be the systematic error at time t. Let be the predicted state at time t. The actual state at time t; Alternatively, the predicted action can be mapped to the state space using a robot kinematics model, whereby the system error is defined as: ， in, The predicted action is the action corresponding to the initial action at time t. This represents the robot's forward kinematics or dynamics mapping function.

4. The robot control method according to claim 2, characterized in that, Determining the error correction amount based on the system error includes: An error-driven dynamic compensation mechanism is adopted to generate an error correction amount based on the system error.

5. The robot control method according to any one of claims 1 to 4, characterized in that, The step of correcting the initial action based on the error correction amount to obtain the target action includes: Obtain the fusion coefficient corresponding to the error correction amount, wherein the value of the fusion coefficient is positively correlated with the value of the deviation; Based on the fusion coefficient, the error correction amount is fused with the initial action to obtain the target action.

6. The robot control method according to any one of claims 1 to 4, characterized in that, After controlling the robot to perform an action based on the target action, the method further includes: Obtain the updated state of the robot after it performs the target action; The updated state is used for the initial action prediction of subsequent time steps to form a closed-loop optimization process consisting of prediction, execution, feedback, and correction, thereby suppressing the accumulation of errors during the execution of long action sequences.

7. A robot control device, characterized in that, The device includes: The state acquisition module is used to acquire state features that indicate the robot's current motion state; An initial action prediction module is used to predict the initial action of the robot based on the state features, thereby obtaining the initial action of the robot. The initial action is used to indicate the predicted action and the corresponding predicted state of the robot. The actual state acquisition module is used to acquire the actual state of the robot, and the actual state is used to indicate the actual state of the robot during the execution of an action. An error correction module is used to determine an error correction amount based on the deviation between the predicted state and the actual state, and the error correction amount is used to correct and compensate the initial action. An action correction module is used to correct the initial action based on the error correction amount to obtain the target action; The control module is used to control the robot to perform actions based on the target action.

8. An electronic device, characterized in that, The electronic device includes: One or more processors; and A memory storing computer program instructions, which, when executed, cause the processor to perform the steps of the method as described in any one of claims 1 to 6.

9. A computer-readable medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 6.

10. 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 steps of the method according to any one of claims 1 to 6.

Images