Body-aware device control method, apparatus, device, medium, and program product

Abstract

The application provides a body intelligent device control method and device, equipment, medium and program product, relates to the technical field of data processing, and is used for improving the accuracy of body intelligent device performance parameter adjustment. The specific technical scheme is as follows: a test request is sent to the body intelligent device, the test request includes a body intelligent test case; a body intelligent test response sent by the body intelligent device is received; a preset task index set includes a task index corresponding to a sensing capability, an interaction capability, a motion control capability and an intelligent decision capability of the body intelligent device in a physical ontology dimension, an environment perception interaction dimension and an intelligent degree dimension; a test value corresponding to each target task index is compared with a preset reference value to obtain comparison result information, the comparison result information is used to represent the capability of the body intelligent device; and a control instruction is sent to the body intelligent device based on the comparison result information. The application is applied to a body intelligent device evaluation control scene.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, apparatus, device, medium and program product for controlling an embodied intelligent device. Background Technology

[0002] With the development of technology, the direction of artificial intelligence is shifting from virtual environments to embodied intelligence that interacts with the real world. Embodied intelligence can achieve intelligent growth through continuous interaction between a physical entity and its environment, and accurately evaluating its performance is key to driving technological development in this field.

[0003] Existing methods for assessing embodied intelligence revolve around a “brain-cerebellum-system-safety” framework to design embodied safety indicators. Specifically, embodied brain assessment is used to quantitatively evaluate multimodal cognitive abilities, embodied cerebellum assessment is used to quantitatively evaluate physical executive abilities, embodied system assessment is used to quantitatively evaluate system coordination abilities, and embodied safety assessment is used to quantitatively evaluate safety protection capabilities.

[0004] However, existing methods for controlling embodied intelligent devices are essentially still decomposition and evaluation of the preset capabilities of traditional robots. The evaluation metrics are relatively singular and fail to reflect the autonomous learning and strategy optimization capabilities of the intelligent system during continuous interaction with the environment. This ultimately leads to poor accuracy in adjusting the performance parameters of embodied intelligent devices based on the aforementioned evaluation metrics. Summary of the Invention

[0005] This application provides a method, apparatus, device, medium, and program product for controlling embody intelligent devices, which is used to improve the control accuracy of embody intelligent devices.

[0006] In a first aspect, embodiments of this application provide a method for controlling an embodied intelligent device. The method includes: sending a test request to the embodied intelligent device, the test request including an embodied intelligent test example; receiving an embodied intelligent test response sent by the embodied intelligent device; wherein the embodied intelligent test response includes a test value for at least one target task indicator, the test value being a test value output by the embodied intelligent device after executing the embodied intelligent test example; the at least one target task is determined from a preset task indicator set of the embodied intelligent device based on the embodied intelligent test example; the preset task indicator set includes task indicators corresponding to the embodied intelligent device's perception capability, interaction capability, motion control capability, and intelligent decision-making capability under the dimensions of physical ontology, environmental perception and interaction, and intelligence level; comparing the test value corresponding to each target task indicator with a preset benchmark value to obtain comparison result information, the comparison result information being used to characterize the capabilities of the embodied intelligent device; and sending a control command to the embodied intelligent device based on the comparison result information.

[0007] The technical solution provided in this application brings at least the following beneficial effects: It can proactively send test requests for test cases to embodied intelligent devices, and receive test responses containing target task indicator test values ​​after the device executes the test cases. The target task indicators are selected on demand from a preset set of indicators, which comprehensively covers evaluation indicators corresponding to various capabilities of the device, including physical entity, environmental perception and interaction, and intelligence level, based on multi-dimensional measured data to obtain accurate test values. Subsequently, the test values ​​of each indicator are compared with preset benchmark values ​​to form a comparison result that truly and objectively reflects the actual comprehensive capabilities of the device. Based on this accurate comparison result, adaptive control commands are generated and sent to the device. The entire process abandons the traditional single-indicator evaluation mode, relying on multi-dimensional test data to ensure the authenticity and reliability of the evaluation results. Based on this, the device performance parameters are optimized and adjusted, gradually improving the accuracy of performance parameter adjustments for embodied intelligent devices.

[0008] One possible implementation is that the aforementioned capability assessment information includes first capability information of the perception capability of the aforementioned embodied intelligent device, second capability information of the interaction capability of the aforementioned embodied intelligent device, third capability information of the motion control capability of the aforementioned embodied intelligent device, and fourth capability information of the intelligent decision-making capability of the aforementioned embodied intelligent device.

[0009] The above-mentioned control commands are sent to the above-mentioned smart device based on the comparison results, including:

[0010] Based on the aforementioned first capability information, second capability information, third capability information, and fourth capability information, control commands are sent to the aforementioned embodied intelligent device.

[0011] The aforementioned control instructions include at least one of the following: a control strategy for the perception capabilities of the aforementioned embodied intelligent device, a control strategy for the interaction capabilities of the aforementioned embodied intelligent device, a control strategy for the motion control capabilities of the aforementioned embodied intelligent device, and a control strategy for the intelligent decision-making capabilities of the aforementioned embodied intelligent device.

[0012] Another possible implementation, after sending control commands to the embodied intelligent device based on the comparison result information, the method further includes:

[0013] After receiving the control feedback sent by the aforementioned embodied intelligent device, a task execution instruction is sent to the aforementioned embodied intelligent device. The task execution instruction indicates that the target task to be executed corresponds to the same application scenario as the aforementioned embodied intelligent test case.

[0014] Another possible implementation is that the aforementioned preset task indicator set includes perception performance indicators, interaction performance indicators, body motion and control performance indicators, and intelligent decision-making performance indicators.

[0015] The aforementioned perception performance indicators are used to characterize the perception capabilities of the aforementioned embodied intelligent device in at least one of the following dimensions: physical ontology, environmental perception and interaction, and intelligence level. The aforementioned interaction performance indicators are used to characterize the interaction capabilities of the aforementioned embodied intelligent device in at least one of the following dimensions: physical ontology, environmental perception and interaction, and intelligence level. The aforementioned ontology motion and control performance indicators are used to characterize the motion control capabilities of the aforementioned embodied intelligent device in at least one of the following dimensions: physical ontology, environmental perception and interaction, and intelligence level. The aforementioned intelligent decision-making performance indicators are used to characterize the intelligent decision-making capabilities of the aforementioned embodied intelligent device in at least one of the following dimensions: physical ontology, environmental perception and interaction, and intelligence level.

[0016] Another possible implementation method is that the above-mentioned perception performance indicators include at least one of the following: environmental perception indicators, target perception indicators, force / touch perception indicators, and perception fusion and dynamic perception indicators;

[0017] The aforementioned interactive performance metrics include at least one of the following: voice interaction metrics, visual interaction metrics, physical contact interaction metrics, multimodal fusion interaction metrics, and contextualized semantic interaction metrics.

[0018] The aforementioned body motion and control performance indicators include at least one of the following: body pose accuracy, motion stability and smoothness, dynamic characteristics, load and endurance, and adaptability and reliability.

[0019] The aforementioned intelligent decision-making performance indicators include at least one of the following: cognitive decision-making performance indicators, scenario autonomous adaptation performance indicators, learning and evolution performance indicators, intelligent collaboration performance indicators, and global planning and problem-solving performance indicators.

[0020] Secondly, embodiments of this application provide a control device for an embodied intelligent device, comprising: a first sending module, configured to send a test request to an embodied intelligent device, wherein the test request includes embodied intelligent test cases;

[0021] A receiving module is configured to receive an embodied intelligence test response sent by the embodied intelligence device; wherein the embodied intelligence test response includes a test value of at least one target task indicator, the test value being the test value output by the embodied intelligence device after executing the embodied intelligence test example; the at least one target task is determined from a preset task indicator set of the embodied intelligence device based on the embodied intelligence test example; the preset task indicator set includes task indicators corresponding to the perception capability, interaction capability, motion control capability, and intelligent decision-making capability of the embodied intelligence device under the dimensions of physical ontology, environmental perception and interaction, and intelligence level, respectively.

[0022] The comparison module is used to compare the test value corresponding to each target task indicator with the preset benchmark value to obtain comparison result information. The comparison result information is used to characterize the capabilities of the aforementioned embodied intelligent device.

[0023] The second sending module is used to send control commands to the aforementioned embodied intelligent device based on the comparison result information.

[0024] One possible implementation is that the aforementioned capability assessment information includes first capability information of the perception capability of the aforementioned embodied intelligent device, second capability information of the interaction capability of the aforementioned embodied intelligent device, third capability information of the motion control capability of the aforementioned embodied intelligent device, and fourth capability information of the intelligent decision-making capability of the aforementioned embodied intelligent device.

[0025] The second sending module is specifically used to send control commands to the embodied intelligent device based on the first capability information, the second capability information, the third capability information, and the fourth capability information.

[0026] The aforementioned control instructions include at least one of the following: a control strategy for the perception capabilities of the aforementioned embodied intelligent device, a control strategy for the interaction capabilities of the aforementioned embodied intelligent device, a control strategy for the motion control capabilities of the aforementioned embodied intelligent device, and a control strategy for the intelligent decision-making capabilities of the aforementioned embodied intelligent device.

[0027] Another possible implementation, the aforementioned embodied intelligent device control device also includes:

[0028] The third sending module is used to send a task execution instruction to the aforementioned embodied intelligent device after receiving the control feedback sent by the aforementioned embodied intelligent device. The task execution instruction indicates that the target task to be executed corresponds to the same application scenario as the aforementioned embodied intelligent test case.

[0029] Another possible implementation is that the aforementioned preset task indicator set includes perception performance indicators, interaction performance indicators, body motion and control performance indicators, and intelligent decision-making performance indicators.

[0030] The aforementioned perception performance indicators are used to characterize the perception capabilities of the aforementioned embodied intelligent device in at least one of the following dimensions: physical ontology, environmental perception and interaction, and intelligence level. The aforementioned interaction performance indicators are used to characterize the interaction capabilities of the aforementioned embodied intelligent device in at least one of the following dimensions: physical ontology, environmental perception and interaction, and intelligence level. The aforementioned ontology motion and control performance indicators are used to characterize the motion control capabilities of the aforementioned embodied intelligent device in at least one of the following dimensions: physical ontology, environmental perception and interaction, and intelligence level. The aforementioned intelligent decision-making performance indicators are used to characterize the intelligent decision-making capabilities of the aforementioned embodied intelligent device in at least one of the following dimensions: physical ontology, environmental perception and interaction, and intelligence level.

[0031] Another possible implementation method is that the above-mentioned perception performance indicators include at least one of the following: environmental perception indicators, target perception indicators, force / touch perception indicators, and perception fusion and dynamic perception indicators;

[0032] The aforementioned interactive performance metrics include at least one of the following: voice interaction metrics, visual interaction metrics, physical contact interaction metrics, multimodal fusion interaction metrics, and contextualized semantic interaction metrics.

[0033] The aforementioned body motion and control performance indicators include at least one of the following: body pose accuracy, motion stability and smoothness, dynamic characteristics, load and endurance, and adaptability and reliability.

[0034] The aforementioned intelligent decision-making performance indicators include at least one of the following: cognitive decision-making performance indicators, scenario autonomous adaptation performance indicators, learning and evolution performance indicators, intelligent collaboration performance indicators, and global planning and problem-solving performance indicators.

[0035] Thirdly, this application provides an electronic device comprising: a processor and a memory; the memory stores a program or instructions executable on the processor, wherein the program or instructions, when executed by the processor, implement the method of the first aspect described above.

[0036] Fourthly, this application provides a readable storage medium on which a program or instructions are stored, which, when executed by a computer, implement the method of the first aspect described above.

[0037] Fifthly, this application provides a computer program product stored in a storage medium, which, when executed by a computer, implements the method described in the first aspect.

[0038] In a sixth aspect, embodiments of this application provide a chip including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the method described in the first aspect.

[0039] The beneficial effects of the second to sixth aspects mentioned above are described in the corresponding description of the first aspect and will not be repeated here. Attached Figure Description

[0040] Figure 1 A schematic diagram of the network architecture for an application of an embodied intelligent device control method provided in this application embodiment;

[0041] Figure 2 A flowchart illustrating a method for controlling an embodied intelligent device provided in an embodiment of this application;

[0042] Figure 3 A flowchart illustrating another embodied intelligent device control method provided in this application embodiment;

[0043] Figure 4 A flowchart illustrating another embodied intelligent device control method provided in this application embodiment;

[0044] Figure 5 A schematic diagram illustrating an embodied intelligence performance evaluation system and its interaction with an embodied intelligence system, provided as an embodiment of this application;

[0045] Figure 6 A schematic diagram of the structure of a holographic intelligent device control device provided in an embodiment of this application;

[0046] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0047] The following is a detailed description of the embodied intelligent device control method, apparatus, equipment, medium, and program products provided in this application, with reference to the accompanying drawings.

[0048] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0049] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0050] The terms "at least one," "at least one of," etc., used in the specification and claims of this application refer to any one, any two, or a combination of two or more of the included items. For example, at least one of a, b, and c can mean: "a," "b," "c," "a and b," "a and c," "b and c," and "a, b, and c," where a, b, and c can be single or multiple. Similarly, "at least two" refers to two or more items, and its meaning is similar to that of "at least one."

[0051] In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0052] The embodiments of this application provide a method, apparatus, device, medium, and program product for controlling embodied intelligent devices, which can be applied to scenarios involving the control and evaluation of embodied intelligent devices.

[0053] In existing technologies, the control methods for embodied intelligent devices are essentially still decomposition and evaluation of the preset capabilities of traditional robots. The evaluation indicators are relatively singular and cannot reflect the autonomous learning and strategy optimization capabilities of intelligent systems during continuous interaction with the environment. Ultimately, this leads to the problem of poor accuracy in controlling embodied intelligent devices based on the aforementioned evaluation indicators.

[0054] To address the aforementioned technical issues, this application provides a method, apparatus, device, medium, and program product for controlling embodied intelligent devices. This method proactively sends test requests for test cases to the embodied intelligent device and receives test responses containing target task indicator test values ​​after the device executes the test cases. The target task indicators are selected on demand from a preset set of indicators, which comprehensively covers evaluation indicators corresponding to various capabilities of the device, including physical entity, environmental perception and interaction, and intelligence level, based on multi-dimensional measured data to obtain accurate test values. Subsequently, the test values ​​of each indicator are compared with preset benchmark values ​​to form a comparison result that truly and objectively reflects the actual comprehensive capabilities of the device. Based on this accurate comparison result, adaptive control commands are generated and sent to the device. The entire process abandons the traditional single-indicator evaluation mode, relying on multi-dimensional test data to ensure the authenticity and reliability of the evaluation results. Based on this, the device performance parameters are optimized and adjusted, gradually improving the accuracy of performance parameter adjustments for embodied intelligent devices.

[0055] The following description, in conjunction with the accompanying drawings, details the embodied intelligent device control method, apparatus, equipment, medium, and program products provided in the embodiments of this application.

[0056] Figure 1 This illustration shows the network architecture of an application of an embodied intelligent device control method provided in an embodiment of this application. For example... Figure 1 As shown, the network architecture includes multiple data processing devices 101. The embodied intelligent device control device is interconnected with the embodied intelligent devices, and the carriers of the embodied intelligent device control device and the embodied intelligent devices include, but are not limited to, the data processing devices 101.

[0057] In some embodiments, the multiple data processing devices 101 may be servers, computers, or processors or processing units within servers or computers. The server may be a single server or a server cluster consisting of multiple servers. It should be noted that the embodiments of this application do not limit the specific device form of the data processing devices 101. Figure 1 The example shown is a data processing device 101, which is a single server.

[0058] In some embodiments, the embodied intelligent device control device 101 is used to send a test request to the embodied intelligent device, the test request including embodied intelligent test examples, and to receive an embodied intelligent test response sent by the embodied intelligent device. The embodied intelligent test response includes test values ​​for at least one target task indicator. The test value corresponding to each target task indicator is compared with a preset benchmark value to obtain comparison result information. The comparison result information is used to characterize the capabilities of the embodied intelligent device. Based on the comparison result information, a control command is sent to the embodied intelligent device. The embodied intelligent device is used to receive embodied intelligent test examples, send embodied intelligent test responses to the embodied intelligent device control device, and receive control commands.

[0059] It should be noted that the network architecture described in the embodiments of this application is for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and does not constitute a limitation on the technical solutions provided in the embodiments of this application. As network architectures evolve, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0060] See Figure 2 This is a flowchart illustrating a method for controlling an embodied intelligent device according to an embodiment of this application. Figure 2 As shown, the embodied intelligent device control method provided in this application embodiment can be implemented by the above-mentioned embodied intelligent device control device, specifically including the following steps 201 to 204.

[0061] Step 201: The embodied intelligent device control device sends a test request to the embodied intelligent device, the test request including embodied intelligent test cases.

[0062] In some embodiments, embodied intelligence test cases characterize the comprehensive ability of an intelligent agent to complete a specific task through closed-loop interaction of perception, decision-making, and action in a physical or simulated environment. They serve as a standardized vehicle for quantitatively evaluating an agent's environmental adaptability, task execution efficiency, and generalization transfer level.

[0063] Step 202: The embodied intelligent device control device receives the embodied intelligent test response sent by the aforementioned embodied intelligent device.

[0064] In some embodiments, the aforementioned embodied intelligence test response includes a test value of at least one target task indicator, the test value being the test value output by the aforementioned embodied intelligence device after executing the aforementioned embodied intelligence test example.

[0065] In some embodiments, the at least one target task is determined from a preset task index set of the embodied intelligence device based on the embodied intelligence test cases.

[0066] In some embodiments, the aforementioned preset task indicator set includes task indicators corresponding to the perception capability, interaction capability, motion control capability, and intelligent decision-making capability of the aforementioned embodied intelligent device under the dimensions of physical ontology, environmental perception and interaction, and intelligence level.

[0067] Step 203: The embodied intelligent device control device compares the test value corresponding to each target task indicator with the preset benchmark value to obtain the comparison result information.

[0068] In some embodiments, the comparison result information is used to characterize the capabilities of the aforementioned embodied smart device.

[0069] Step 204: The embodied intelligent device control device sends a control command to the embodied intelligent device based on the comparison result information.

[0070] In some embodiments, the embodied intelligent device control device sends real-time action correction, strategy adjustment, or task execution instructions to the embodied intelligent device based on the comparison results of the difference between the perception feedback and the task target or baseline state, thereby realizing closed-loop control of perception, decision-making, and execution.

[0071] In some embodiments, combined with Figure 2 ,like Figure 3 As shown, the aforementioned capability assessment information includes first capability information of the perception capability of the aforementioned embodied intelligent device, second capability information of the interaction capability of the aforementioned embodied intelligent device, third capability information of the motion control capability of the aforementioned embodied intelligent device, and fourth capability information of the intelligent decision-making capability of the aforementioned embodied intelligent device. The aforementioned step 204 can be specifically implemented through the following step 204a.

[0072] Step 204a: The embodied intelligent device control device generates control commands for the embodied intelligent device based on the first capability information, the second capability information, the third capability information, and the fourth capability information.

[0073] In some embodiments, the control instructions include at least one of the following: a control strategy for the sensing capabilities of the aforementioned embody intelligent device, a control strategy for the interaction capabilities of the aforementioned embody intelligent device, a control strategy for the motion control capabilities of the aforementioned embody intelligent device, and a control strategy for the intelligent decision-making capabilities of the aforementioned embody intelligent device.

[0074] In some embodiments, to more intuitively illustrate the types of control strategies, exemplary descriptions are provided below, including: For example, if the assessment finds weak perception and poor low-light recognition, information can be issued to increase the infrared perception weight, decrease the visual recognition confidence threshold, and automatically activate the low-light mode. If the assessment finds large deviations in the executed actions, information can be issued to correct the joint trajectory parameters, reduce operational tolerances, and decrease the movement speed. If the assessment finds insufficient autonomous intelligence and susceptibility to erroneous decisions, information can be issued to lower the autonomous operation level, add a manual pre-review step, and restrict autonomous actions in unknown scenarios. If the assessment finds slow human-computer interaction response, information can be issued to simplify the interaction parsing model, shorten command response latency, and disable redundant interaction functions.

[0075] In this way, by integrating capability information from four dimensions—sensing, interaction, motion control, and intelligent decision-making—targeted and differentiated control strategies can be generated for embodied intelligent devices, enabling precise reinforcement of capability shortcomings and significantly improving the device's task execution accuracy, reliability, and adaptability to complex environments.

[0076] The embodied intelligent device control method provided in this application can proactively send test requests for test cases to the embodied intelligent device and receive test responses containing target task indicator test values ​​after the device executes the test cases. The target task indicators are selected as needed from a preset set of task indicators. This set of indicators comprehensively covers three dimensions: physical entity, environmental perception and interaction, and level of intelligence, fully encompassing evaluation indicators corresponding to various capabilities of the device, including perception, interaction, motion control, and intelligent decision-making. Accurate test values ​​are obtained based on multi-dimensional measured data. Subsequently, the test values ​​of each indicator are compared with preset benchmark values ​​to form a comparison result that truly and objectively reflects the actual comprehensive capabilities of the device. Adaptive control commands are then generated and sent to the device based on this accurate comparison result. The entire process abandons the traditional single-indicator evaluation mode, relying on multi-dimensional test data to ensure the authenticity and reliability of the evaluation results. Based on this, the device performance parameters are optimized and adjusted, gradually improving the accuracy of performance parameter adjustments for the embodied intelligent device.

[0077] In some embodiments, combined with Figure 2 ,like Figure 4 As shown, following step 204 above, the embodiment of this application provides the following step: The embodied intelligent device control method may further include step 301.

[0078] Step 301: After receiving the control feedback sent by the aforementioned embodied intelligent device, the embodied intelligent device control device sends a task execution instruction to the aforementioned embodied intelligent device.

[0079] In some embodiments, the target task to be executed as described above corresponds to the same application scenario as the embodied intelligence test case described above.

[0080] In some embodiments, the control feedback received by the embodied intelligent device control device typically refers to the execution status data, capability verification results, and real-time environmental perception information returned by the device after executing the aforementioned differentiated control strategies of perception, interaction, motion control, or intelligent decision-making. The task execution instruction and the corresponding test case belong to the same application scenario, which means that the operating environment, task objectives, and action complexity required by the instruction are completely aligned with the test cases previously used for capability assessment. For example, after the device has been verified for the test case of object grasping in a dark environment, it will issue actual workpiece grasping task instructions in the same scenario to ensure that the performance parameters corrected based on the capability assessment results can be directly mapped to the actual task execution.

[0081] In this way, the core pain point of the disconnect between "testing and execution" in traditional embodied intelligence systems is effectively solved, ensuring that the device capability boundaries evaluated based on test cases are accurately matched with actual task requirements, and greatly reducing the risk of task failure due to overestimation of capabilities or mismatch with the environment. At the same time, through the closed-loop linkage between control feedback and task execution, the effect of the adjustment of the control strategy in the early stage can be quickly verified, providing reliable data support in real scenarios for the continuous iterative optimization of device capabilities.

[0082] In some embodiments, the aforementioned preset task indicator set includes perception performance indicators, interaction performance indicators, body motion and control performance indicators, and intelligent decision-making performance indicators.

[0083] In some embodiments, the perception performance index is used to characterize the perception capability of the embodied intelligent device in at least one of the dimensions of physical ontology, environmental perception and interaction, and intelligence level; the interaction performance index is used to characterize the interaction capability of the embodied intelligent device in at least one of the dimensions of physical ontology, environmental perception and interaction, and intelligence level; the ontology motion and control performance index is used to characterize the motion control capability of the embodied intelligent device in at least one of the dimensions of physical ontology, environmental perception and interaction, and intelligence level; and the intelligent decision-making performance index is used to characterize the intelligent decision-making capability of the embodied intelligent device in at least one of the dimensions of physical ontology, environmental perception and interaction, and intelligence level.

[0084] In some embodiments, the above-mentioned perception performance indicators include at least one of the following: environmental perception indicators, target perception indicators, force / touch perception indicators, and perception fusion and dynamic perception indicators.

[0085] In some embodiments, the above-mentioned interaction performance indicators include at least one of the following: voice interaction indicators, visual interaction indicators, physical contact interaction indicators, multimodal fusion interaction indicators, and contextualized semantic interaction indicators.

[0086] In some embodiments, the above-mentioned body motion and control performance indicators include at least one of the following: body pose accuracy indicator, motion stability and smoothness indicator, dynamic characteristic indicator, load and endurance indicator, and adaptability and reliability indicator.

[0087] In some embodiments, the above-mentioned intelligent decision-making performance indicators include at least one of the following: cognitive decision-making performance indicators, scene autonomous adaptation performance indicators, learning evolution performance indicators, intelligent collaboration performance indicators, and global planning and problem-solving performance indicators.

[0088] Figure 5 This is a schematic diagram illustrating an embodied intelligence performance evaluation system and its interaction with the embodied intelligence system, provided as an embodiment of this application. Figure 5 As shown, the embodied intelligence performance evaluation system includes four types of modules: evaluation modules for the embodied intelligence multimodal perception subsystem (modules 101-104), evaluation modules for the embodied intelligence interaction subsystem (modules 201-205), evaluation modules for the embodied intelligence motion control subsystem (modules 301-304), and evaluation modules for the embodied intelligence decision control subsystem (modules 401-405). Each evaluation and testing module contains the benchmark values ​​of performance indicators in the test cases specified in the test specifications. The test values ​​of the performance indicators after the embodied intelligence system executes the test cases are obtained through the embodied intelligence system data storage and preprocessing subsystem. The benchmark values ​​and test values ​​of the performance indicators are compared and analyzed to obtain the comprehensive evaluation result of the embodied intelligence performance. The embodied intelligence system is the embodied intelligence device.

[0089] Furthermore, Figure 5This demonstrates a layered, multi-dimensional, full-stack performance evaluation system and a two-way data interaction mechanism with the embodied intelligence core system. The embodied intelligence system at the bottom consists of a control bus, an embodied intelligence multimodal perception subsystem, an interaction subsystem, a motion control subsystem, a data bus, a data storage and preprocessing subsystem, and an embodied intelligence decision control subsystem. The control bus enables low-level hardware control, the data bus handles data transmission between subsystems, the data storage and preprocessing subsystem manages and cleans the raw data, and the decision control subsystem coordinates overall scheduling. The performance evaluation system at the top, corresponding to the core capability dimensions of embodied intelligence, constructs a three-layer evaluation module system. The first layer is the perception layer evaluation module, covering force perception... The system comprises three layers: a first layer of multimodal perception, a second layer of target and environment perception, and a third layer of comprehensive capability evaluation module. The first layer consists of a multimodal perception layer, which includes single-modal and multimodal fusion capabilities such as voice, vision, and physical contact, as well as contextualized semantic interaction capabilities. The second layer is an interaction layer evaluation module, which covers basic motion capabilities such as pose accuracy, load endurance, and motion performance, as well as advanced intelligent capabilities such as cognitive decision-making, learning evolution, scene autonomous adaptation, and intelligent collaboration. The evaluation system interacts with the embodied intelligence system through a two-way data channel. It obtains the operating data and execution feedback of each subsystem from the embodied intelligence system to complete a multi-dimensional capability quantitative evaluation. At the same time, it feeds the evaluation results back to the decision control subsystem to support the dynamic adjustment of control strategies and the continuous optimization of system capabilities.

[0090] In some embodiments, to more clearly demonstrate the content of the metrics included in the preset task metric set, this application provides a performance metric system. The metrics of the preset task metric set may include any metric in this performance metric system, which includes:

[0091] 1. Perceived performance indicators

[0092] Embodied intelligence's perception capabilities include decision-making and inference, multimodal perception fusion, dynamic environment understanding, and contextual semantic perception, adapting to autonomous operation needs in various scenarios such as home, commercial, and industrial settings.

[0093] 1.1 Environmental Perception Indicators

[0094] Embodied intelligent bodies utilize multi-dimensional sensors such as vision, lidar, ultrasound, gravimeter, and inertial measurement unit (IMU) to achieve functions such as spatial modeling, obstacle detection, and terrain recognition. This forms the basis for the autonomous movement, obstacle avoidance, and operation of embodied intelligent bodies.

[0095] 3D environment modeling accuracy: Based on multimodal fusion methods such as LiDAR and machine vision, the maximum deviation between the spatial dimensions of the constructed 3D environment model and the real environment. In vertical industry applications, a LiDAR + machine vision fusion mode is typically used, with a 3D environment modeling accuracy of ≤±2cm; in the personal service field, machine vision or low-cost LiDAR is used, with a 3D environment modeling accuracy of ≤±3cm.

[0096] Obstacle detection and recognition rate and ranging accuracy: Detection and recognition rate for obstacles of different materials, sizes, and motion states; deviation between detected distance and actual distance. Static obstacle detection and recognition rate ≥99%; dynamic low-speed (≤0.8m / s) obstacle detection and recognition rate ≥95%; obstacle detection and recognition rate ≥90% in low light (≤50lux) / backlight / high reflectivity environments; ranging accuracy ≤±1cm when ranging range is 0~5m; ranging accuracy ≤±3cm when ranging range is 5~10m.

[0097] Terrain and ground condition recognition rate: The classification and recognition rate for different ground surfaces (cement / tile / carpet / uneven surface) and simple terrain features (slope / steps / pits). Common ground terrain and ground condition recognition rate ≥98%; ≤5° slope / ≤3cm step terrain and ground condition recognition rate ≥95%;

[0098] Environmental perception update rate: The real-time update frequency of environmental data after sensor fusion, with a threshold of ≥10Hz.

[0099] 1.2 Target Perception Indicators

[0100] Embossed intelligent devices achieve the positioning, recognition, and posture detection of work targets through the fusion of vision, force, and tactile multimodal sensors, which is a prerequisite for core operations such as picking up and placing objects, assembly, and tool operation.

[0101] Target positioning accuracy: The three-dimensional spatial positioning deviation of the target (such as a water cup, part, or material box), including static targets and slightly offset dynamic targets. In industrial manufacturing scenarios, when the static small target is ≤10cm×10cm, the target positioning accuracy is ≤±0.5cm; in general service scenarios, the target positioning accuracy is ≤±1cm; in low-speed movement (≤0.3m / s) scenarios, when slightly offset dynamic targets are allowed, the target positioning accuracy is ≤±1.5cm.

[0102] Target recognition success rate: The success rate of classifying and recognizing a specified target in complex backgrounds (such as piles of debris) and partially occluded (occlusion ≤ 30%). In simple background scenes without occlusion, the target recognition success rate is ≥ 99%; in complex background scenes with occlusion ≤ 30%, the target recognition success rate is ≥ 95%.

[0103] Target attitude detection accuracy: The deviation in attitude (pitch / yaw / roll) detection for non-upright targets (such as tilted cups or inverted parts). In lightweight industrial scenarios, the attitude angle deviation is ≤ ±2°; in general service scenarios such as commercial or household use, the attitude angle deviation is ≤ ±3°.

[0104] Multi-target recognition and sorting accuracy: Simultaneously identifies ≥3 different types of targets and sorts them according to instructions. In commercial sorting or household organization scenarios, the multi-target recognition and sorting accuracy is ≥98%.

[0105] Micro-target perception capability: Success rate in identifying and locating micro-targets ≤2cm×2cm. In lightweight industrial precision operation scenarios, the perception capability for micro-targets such as small parts is ≥90%.

[0106] 1.3 Force / Touch Perception Indicators

[0107] Force detection, contact sensing, and pressure detection are achieved through a six-dimensional force sensor, a tactile sensor, and a fingertip pressure sensor.

[0108] End-effector force detection accuracy: The deviation between the force (tension / compression / torque) detected by the end effector during the grasping / operation process and the actual force value. In common operating force scenarios of 0~50N, the force deviation is ≤±5N; in heavy load operating scenarios of 50~100N, the force deviation is ≤±10N.

[0109] Contact sensing sensitivity: The minimum force required to sense slight contact between the device's end cap and a human body or object, and the contact sensing response time. In safe scenarios involving human contact, the minimum sensed contact force is ≤8N, and the contact sensing response time is ≤50ms.

[0110] Adaptive gripping pressure success rate: The system autonomously adjusts gripping pressure to maintain a stable gripping success rate for targets of varying hardness and fragility (e.g., eggs, metal blocks, sponges). For common hardness targets, the adaptive gripping pressure success rate is ≥98%; for fragile items such as eggs and glass, the adaptive gripping pressure success rate is ≥95%.

[0111] 1.4. Perceptual Fusion and Dynamic Perceptual Indicators

[0112] To avoid the limitations of single-sensor perception (such as visual failure in backlight or radar failure to detect transparent obstacles), embodied intelligent entities will improve the reliability of perception in dynamic and complex scenes based on multimodal fusion perception.

[0113] Multimodal perception fusion accuracy: Based on the fusion of multiple sensors such as vision, LiDAR, force sensing, and voice, the comprehensive deviation between the perception results of the environment / target / human-computer interaction and the real scene. Spatial positioning comprehensive perception deviation ≤ ±1.5cm; target / interaction recognition fusion success rate ≥ 98%.

[0114] Dynamic scene perception latency: The latency of multi-sensor fusion perception and decision-making in dynamic environments such as target movement or sudden obstacle appearance. In scenarios where the target is moving at low speed or obstacles suddenly appear, the dynamic scene perception latency is ≤200ms.

[0115] Self-calibration success rate: When a single sensor experiences slight interference / deviation (such as radar point cloud noise or visual blur), the success rate of the subject's autonomous calibration of sensing parameters through multi-sensor fusion. In scenarios with slight interference from a single sensor, the self-calibration success rate is ≥95%; in scenarios with slight interference from two sensors, the self-calibration success rate is ≥90%.

[0116] Cross-scene perception adaptation success rate: The success rate of the perception system autonomously adapting and completing basic tasks when switching from one scene to another without manual debugging. The success rate of the perception system autonomously adapting and completing basic tasks when switching from a home living room scene to a commercial supermarket scene without manual debugging is ≥90%.

[0117] Long-term sensing stability: The rate of decrease in sensing accuracy after a certain period of continuous sensing operation. In commercial scenarios, the rate of decrease in sensing accuracy is ≤5% after ≥8 hours of continuous sensing operation.

[0118] 2. Interaction performance metrics

[0119] By using voice, vision, microphone arrays, and other technologies, natural language understanding, gesture recognition, and facial expression / human posture perception are achieved. This is the core of embodied intelligent robots' ability to "understand people, listen to people, and follow people." Unlike the fixed command interaction of ordinary service robots, it focuses on understanding accuracy, recognition success rate, and real-time response.

[0120] 2.1. Voice Interaction Metrics

[0121] Voice interaction is the most basic and most frequent form of interaction for embodied intelligent entities. Unlike the fixed command interaction of traditional robots, it focuses on conversational language, anti-interference, multi-turn dialogue, and semantic understanding. The core assessment is "hearing clearly, understanding the meaning, answering accurately, and reacting quickly".

[0122] Voice command recognition accuracy: The accuracy of intent recognition and semantic understanding of Mandarin natural language commands, including simple single commands, complex multi-step commands, and vague colloquial commands. Accuracy for simple commands such as "get the water glass" ≥99%; accuracy for complex multi-step commands such as "go to the kitchen to get the water glass and put it on the living room table" ≥95%; accuracy for vague colloquial commands such as "bring that red thing over" ≥90%.

[0123] Voice anti-interference recognition accuracy: The success rate of voice command recognition in scenarios with background noise, long distance, and non-frontal orientation. ① Background noise: ≤60dB (home use) ≥95%, ≤70dB (commercial use) ≥90%; ② Long distance: ≥98% within 3m, ≥95% within 5m; ③ Non-frontal orientation: ≥45° side recognition ≥98%, ≤90° side recognition ≥93%;

[0124] Semantic intent understanding accuracy (including contextual context): The accuracy of understanding the core intent, action object, and execution scenario of voice commands, including contextual understanding in multi-turn dialogues (such as the reference of "it" or "that"). ≥98% (single command without context); ≥93% (multi-turn dialogues with more than 3 rounds of contextual context).

[0125] Voice interaction response latency: The total time from the end of the voice command to the robot's voice response / action execution (excluding the action execution time itself). ≤500ms (voice response); ≤800ms (voice response + initiation action)

[0126] Multilingual / dialect recognition: Mandarin + 1~2 mainstream dialects (such as Cantonese / Sichuanese) recognition ≥85% (suitable for home use by middle-aged and elderly users);

[0127] Speech synthesis naturalness: The naturalness score of the robot's response speech is ≥4.5 / 5 (no mechanical feel, adjustable speech rate);

[0128] Wake-up success rate: Custom wake-up word (2~4 words), wake-up rate ≥99% within 5m / 60dB noise, false wake-up ≤1 time / h.

[0129] Multi-turn dialogue understanding accuracy: The accuracy of intent understanding in multi-turn continuous human-computer dialogue (≥5 turns). In complex interaction scenarios, the multi-turn dialogue understanding accuracy is ≥90%.

[0130] 2.2. Visual Interaction Metrics

[0131] In scenarios where it is difficult to communicate via voice (such as in libraries or supermarkets), the embodied intelligent entity uses visual sensors to capture the user's gestures, body posture, facial expressions / gaze, and achieve natural interaction without voice.

[0132] General interactive gesture recognition success rate: Success rate for recognizing static core gestures and dynamic simple gestures, including non-frontal / slightly occluded scenarios. Core gestures include pointing, raising / waving, nodding / shaking head, and making a heart shape / numbers (1-3). Thresholds: ≥98% (frontal, no occlusion); ≥93% (≤45° side view / ≤20% slight occlusion); ≥90% (≤90° side view).

[0133] Human posture interaction understanding accuracy: The accuracy of understanding the user's core movement posture intent, such as pointing at a target (requiring grasping), raising a hand (requiring approach), waving a hand (requiring stopping), bending over (requiring picking up an item on the ground). ≥95% (front view without obstruction); ≥90% (≤45° side view)

[0134] Visual interaction response latency: The time delay from when the user makes a gesture / pose to when the robot recognizes it and responds / acts. ≤300ms (static gestures); ≤500ms (dynamic gestures / human poses)

[0135] Auxiliary reference indicators: Facial expression / emotion perception accuracy: ≥90% perception of the four basic emotions of happiness, anger, calmness, and confusion (home companionship / commercial shopping guide);

[0136] Eye tracking accuracy: The robot can identify the target area that the user's gaze is pointing at with an accuracy of ≥90% (e.g., pointing at a water glass on a table, the robot recognizes the target).

[0137] Multi-person interaction recognition: In scenarios with 2 to 3 people, the system can recognize the gestures / postures of the person actively interacting with an accuracy rate of ≥95% (avoiding misidentification of irrelevant personnel).

[0138] Facial expression / emotion perception accuracy: The accuracy rate in perceiving basic emotions such as happiness, anger, and calmness. In scenarios such as home companionship and commercial sales guidance, the accuracy rate of facial expression / emotion perception is ≥90%.

[0139] Relative latency: In emotional communication scenarios, micro-expression changes should be strictly controlled within 50 milliseconds before and after the speech stress point; in teaching demonstration scenarios, actions should lead speech by 80-120 milliseconds.

[0140] 2.3. Physical Contact Interaction Indicators

[0141] Physical contact interaction is the core form of human-machine collaborative operation and close-range interaction of embodied intelligent bodies. It realizes contact perception, pressure feedback and flexible response through force / touch sensors. It is different from the "hard contact" of traditional robots. The core assessment is "able to touch, accurate to sense, soft to respond and safe enough". It is the key to lightweight industrial human-machine collaboration and intimate interaction at home.

[0142] Contact sensing sensitivity: The minimum force required to sense slight contact between the embodied intelligent body / end effector and a human body / object, and the response time of contact sensing. Threshold: Minimum sensed contact force ≤ 8N (human contact safety threshold); contact sensing response time ≤ 50ms.

[0143] Contact pressure adaptive response accuracy: The system autonomously adjusts contact pressure and movement speed based on the contact object (human body / hard object / fragile item) to avoid damaging the object / injuring the human body. ≥98% (distinguishing between human body / hard object); ≥95% (distinguishing between fragile items / ordinary items), automatically reducing speed to ≤0.1m / s upon contact with the human body.

[0144] Physical interaction command execution accuracy: The accuracy rate of executing user-issued interaction commands through physical operations, such as gently pushing the robot to move, gently gripping the end effector to grasp, and gently lifting the end effector to raise it. ≥95% (simple physical commands); ≥90% (complex physical commands, such as a combination of gentle push and gentle grip).

[0145] Contact buffer distance: After sensing contact, the robot's buffer braking distance is ≤5mm (to avoid hard collisions);

[0146] Body tactile perception coverage: The tactile perception coverage area of ​​the robot arm / torso is ≥80% (full-area safety protection);

[0147] Interactive force feedback accuracy: The interaction is completed according to the user's preset contact force value (such as 10N / 20N), and the actual force value deviation is ≤±3N (lightweight industrial collaboration).

[0148] 2.4. Multimodal Fusion Interaction Metrics

[0149] Multimodal fusion interaction is the core feature that distinguishes embodied intelligent bodies from traditional service robots. By fusing multimodal information from voice, vision, and physical contact, it achieves "cross-modal intent understanding, multi-command combination interaction, and contextual continuous interaction," adapting to the complex interaction needs of real-world scenarios. The core assessment is "accurate fusion, comprehensive understanding, and seamless connection."

[0150] Multimodal fusion intent understanding accuracy: The accuracy of understanding cross-modal combined interaction intents such as voice + gesture, voice + physical contact, and gesture + physical contact. ≥98% (voice + gesture, e.g., "Take that → point to the water glass"); ≥95% (voice + physical contact, e.g., "Put it here → gently push the robot to the table"); ≥90% (gesture + physical contact, e.g., pointing + gently grasping the end effector).

[0151] Multi-turn continuous interaction context retention rate: In ≥5 rounds of multimodal continuous interaction, the robot's accuracy in retaining the referent, action requirements, and scene context, avoiding "memory gaps". ≥95% (5~8 rounds of continuous interaction); ≥90% (8~10 rounds of continuous interaction)

[0152] Multimodal interaction conflict resolution success rate: When minor conflicts occur between interaction commands from different modalities, the success rate of the robot autonomously resolving the conflict and executing the command based on the scenario, user habits, and action logic. ≥90% (minor conflicts, such as object referencing deviation); ≥85% (action requirement deviation, such as the gesture for "put on the table" pointing to the ground).

[0153] Cross-scene interaction adaptation success rate: When switching from one scene (living room) to another scene (kitchen), multimodal interaction does not require re-debugging, with an accuracy rate of ≥90%;

[0154] Interactive memory accuracy: The accuracy of remembering the user's recent (within 1 hour) interactive habits (such as preferred operation methods and target locations) is ≥95%;

[0155] Multimodal feedback richness: The robot provides feedback on the interaction results through a multimodal approach of voice, light, and action, with a richness score of ≥4.5 / 5 (enhancing user perception).

[0156] 2.5. Contextualized Semantic Interaction Metrics

[0157] Embodied intelligent human-computer interaction ultimately serves scenario-based operations, integrating "interaction + operation" and comprehensively evaluating the understanding and execution capabilities of scenario-based instructions, environmental constraints, and operational logic, which is the key to the implementation of interaction capabilities.

[0158] Scenario-based instruction execution success rate: The understanding and execution success rate of complex operation instructions in different scenarios. Instructions include actions, scenarios, and constraints (e.g., household instruction: "After boiling water in the kitchen, put the kettle in the living room to avoid hitting children"; commercial instruction: "Replenish supermarket shelves, prioritizing empty upper shelves"). ≥95% (simple constraints in a single scenario); ≥90% (complex constraints in a single scenario); ≥85% (complex constraints across scenarios).

[0159] Environmental constraint perception and interaction adaptation accuracy: This refers to the accuracy with which the robot perceives safety and operational constraints in the environment during interactive execution and autonomously adapts its interaction / action methods. It achieves the fusion of "interaction + environment," avoiding operational failures / safety accidents caused by ignoring environmental constraints (e.g., automatically reducing movement speed when there are children nearby; autonomously adjusting the grasping position when there are obstacles on the worktable). ≥98% (obvious environmental constraints); ≥93% (minor / hidden environmental constraints)

[0160] Human-machine collaborative work interaction efficiency: In a human-machine collaborative work scenario, the total time taken to complete a specified task and the number of human interventions reflect the efficiency of the interaction. ① Home use: For tasks completed by a single person within 5 minutes, human-machine collaboration ≤ 6 minutes, and human intervention ≤ 1 time; ② Commercial / light industrial use: For tasks completed by a single person within 10 minutes, human-machine collaboration ≤ 11 minutes, and human intervention ≤ 2 times.

[0161] Unknown command processing capability: For unfamiliar commands that have not been trained on, it can make reasonable responses / actions through semantic reasoning and scenario analysis, with a success rate of ≥80%;

[0162] Job interaction traceability: Record the job steps executed by the interaction, with a traceability accuracy of ≥99% (commercial / industrial operation and maintenance requirements).

[0163] Low-threshold interaction adaptation: The interaction learning time for middle-aged and elderly / non-professional users is ≤5 minutes, and the operation success rate is ≥95% (core home experience).

[0164] 3. Performance indicators of body motion and control

[0165] The body performance index is the core quantitative standard for measuring its motion accuracy, motion flexibility, dynamic response, load capacity, and environmental adaptability. It is designed around six dimensions: pose accuracy, motion stability, dynamic characteristics, load and endurance, environmental adaptability, and motion range. It also distinguishes between static performance (no motion / low speed) and dynamic performance (high speed / variable working conditions), taking into account both general basic indicators and customized indicators for different robot categories.

[0166] 3.1. Body pose accuracy indicators

[0167] The pose accuracy index of the body is the core performance indicator of the body's motion and directly affects the quality of the body's interactive actions. It is divided into three categories: position accuracy, posture accuracy, and repeatability accuracy, all of which are quantifiable hard indicators.

[0168] (1) Position accuracy

[0169] Absolute position accuracy: The maximum and average deviations between the actual position reached by the end effector and the commanded target position, reflecting the comprehensive positioning capability of the body and the control system;

[0170] Relative position accuracy: The relative deviation between the actual position and the theoretical position of the body at different points on the same motion trajectory, suitable for continuous trajectory operations;

[0171] Repeatability of position: When a robot repeatedly executes the same instruction (same pose / same trajectory) multiple times, the deviation range of the actual position of the end effector is the most valuable accuracy indicator in actual operation (it reflects the real operation capability better than absolute position accuracy). The repeatability of the positioning of industrial robotic arms can reach ±0.01mm, the repeatability of collaborative robotic arms is about ±0.05~0.1mm, and the repeatability of mobile robotic arms is about ±1~5mm.

[0172] Trajectory accuracy: The deviation between the actual trajectory and the theoretical trajectory when the robot's end effector moves along the predetermined trajectory (divided into positional deviation and profile deviation). Profile deviation is the core indicator for continuous trajectory operation.

[0173] (2) Attitude accuracy

[0174] Absolute attitude accuracy: The maximum and average deviations between the actual attitude (pitch / yaw / roll) of the end effector and the attitude of the commanded target;

[0175] Repeatable attitude accuracy: The maximum and average values ​​of attitude deviation when repeatedly reaching the same target attitude. It is closer to the actual operation requirements than absolute attitude accuracy.

[0176] 3.2. Motion stability and smoothness indicators

[0177] Focusing on the smoothness, continuity, and lack of impact during the movement of the body, and avoiding operation failures (such as precision assembly or handling of fragile items) or body damage caused by movement jitter or sudden speed changes, are key indicators for collaborative and service-oriented bodies.

[0178] (1) Motion stability

[0179] End effector jitter: The jitter amplitude (mm / °) of the end effector during body movement (especially during start-up / stop / direction change). The lower the jitter, the more stable the operation.

[0180] Speed ​​fluctuation rate: The percentage deviation between the actual speed and the commanded speed during movement. A fluctuation rate of ≤±5% is considered excellent, avoiding sudden changes in speed.

[0181] Impact intensity: The rate of change in acceleration (m / s³) during the start-up, stop, acceleration, deceleration, and change of direction of the main body. Excessive impact intensity can easily lead to wear of the main body and damage to the workpiece.

[0182] (2) Motion fluency

[0183] Track smoothness: The rate of change of curvature of a track during continuous motion. The higher the smoothness, the smoother the motion, which is suitable for continuous operations such as spraying, polishing, and welding.

[0184] Acceleration / deceleration smoothness: Whether the change in acceleration during acceleration / deceleration is linear, without abrupt changes, and to avoid mechanical shock;

[0185] Multi-joint coordination and smoothness: Synchronization of the movement of each joint in the multi-joint body. A joint movement lag time of ≤10ms is considered excellent, avoiding end-effector displacement caused by uncoordinated joint movements.

[0186] (3) Motion flexibility

[0187] Degrees of freedom (DOF): The number of independent motion joints of a body. The higher the degree of freedom, the more flexible the movement. Industrial bodies are mostly 6-axis (6DOF, enabling arbitrary pose in space), while collaborative bodies can add an end effector axis (7DOF, improving obstacle avoidance capabilities). Mobile bodies are basically 2DOF for forward and backward movement plus turning.

[0188] Obstacle avoidance success rate: In a pre-set multi-obstacle scenario, the percentage of times the ontology plans a path and successfully avoids obstacles (%), ≥95% is considered excellent, and is the core of mobile ontologies;

[0189] Success rate in narrow spaces: the percentage of successful passage / operations in a narrow space less than 1.2 times the width of the main body (%).

[0190] In-situ turning capability: Whether the mobile object supports 360° turning in place, and the turning radius. Zero turning radius is optimal and suitable for narrow spaces.

[0191] 3.3. Dynamic Characteristic Indicators

[0192] It measures the coverage of the body's operations, its ability to move in complex spaces, and its ability to move at high speeds, respond quickly, and adjust to changing working conditions. It is suitable for high-speed operation scenarios such as sorting, palletizing, and handling in warehousing. The core is speed, acceleration, and response latency, and it distinguishes between rated values ​​and peak values.

[0193] (1) Range of motion

[0194] Workspace: All spatial areas that the end effector can reach, divided into reachable workspace (reachable but with no posture restrictions) and flexible workspace (reachable and with any posture). The workspace of industrial-type bodies is mostly quantified by the volume of a cube / sphere (m³), while that of mobile bodies is quantified by the travel radius / operation coverage area.

[0195] Joint range of motion: the range of rotation and swing angles of a single joint (°), such as the rotation range of industrial manufacturing body joints being -180° to +180°;

[0196] Obstacle clearance height / width: The height (mm) and width (mm) of obstacles that mobile units can smoothly pass through are core indicators of ground adaptability. Service units have an obstacle clearance height of ≥20mm, and storage units have an obstacle clearance height of ≥10mm.

[0197] (2) Motion speed and acceleration / deceleration

[0198] Joint speed: The rated rotational speed and peak rotational speed (° / s) of a single joint are the basic indicators of the product body;

[0199] End-effector linear speed: The rated linear speed and peak linear speed (m / s) of the end effector are the core of measuring high-speed operation capability. The end-effector linear speed of warehousing AGVs can reach 2~5m / s, while the travel speed of mobile AGVs is divided into low speed (≤0.5m / s, collaborative), medium speed (0.5~2m / s, general), and high speed (≥2m / s, warehousing AGVs).

[0200] Trajectory motion speed: The average linear velocity (m / s) during continuous trajectory motion, applicable to continuous operations.

[0201] Joint acceleration / deceleration: Rated acceleration / deceleration and peak angular acceleration / deceleration (° / s²) of a single joint;

[0202] End-efficiency acceleration / deceleration: Rated linear acceleration / deceleration and peak linear acceleration / deceleration (m / s²) of the end effector. The higher the acceleration / deceleration, the faster the body's start-up, stop, and direction-changing response, and the higher the work efficiency.

[0203] Maximum deceleration: The maximum deceleration of the vehicle body during emergency braking determines the braking distance and is an important component of safety indicators.

[0204] (3) Response delay

[0205] Command response latency: The time (ms) from when the control system issues a motion command to when the robot joint / endpoint begins to move. The shorter the latency, the faster the response. The response latency of mainstream industrial robots is ≤50ms.

[0206] Variable operating condition response latency: The time (ms) it takes for the body to switch from one motion state (such as constant speed) to another state (such as acceleration / direction change), adapting to dynamic operation scenarios (such as dynamic grasping).

[0207] Emergency stop response delay: The time (ms) after an emergency stop command is issued for the main body to completely stop moving is a core safety indicator, and it is required to be ≤200ms.

[0208] 3.4. Load and Battery Life Indicators

[0209] Load capacity determines the types of tasks that the game can perform. Battery life is a core indicator for mobile games. It is important to balance performance retention under load and avoid excessive load that could lead to a decrease in motion accuracy.

[0210] (1) Load capacity

[0211] Rated load: The maximum weight (kg) that the end effector can stably bear within the rated speed and range of the main body. For example, the rated load of assembly-type main bodies is 2~5kg, and that of warehousing-type main bodies is 50~200kg.

[0212] Peak load: The maximum load that the main body can withstand in a short time (≤30min), which must be lower than the structural design limit of the main body;

[0213] Load torque / inertia: The maximum load torque (N·m) and rotational inertia (kg·m²) that the end effector can withstand. Especially for industrial robotic arms, excessive load inertia can lead to joint vibration and reduced accuracy.

[0214] Load accuracy retention rate: The ratio (%) of body position / repeatability under rated load to that under no load. A retention rate of ≥95% is excellent, avoiding accuracy degradation due to load.

[0215] (2) Battery life and power consumption (specific to mobile devices)

[0216] Battery life: The continuous working time and driving time (h) of the unit under rated working conditions after a single charge / battery swap. For example, the battery life of warehouse-type units is ≥8h, the battery life of service-type units is ≥4h, and the battery life of mobile-type units is ≥6h, which is a general standard.

[0217] Energy consumption indicators: energy consumption per unit of work (kWh / piece, such as sorting pieces, handling tons) or energy consumption per unit of travel distance (kWh / km), reflecting energy efficiency;

[0218] Fast charging efficiency: Time (h) to fast charge to 80% of rated capacity, suitable for high-frequency operation scenarios, mainstream industrial mobile robots can be fully charged to 80% in 1~2 hours;

[0219] Load-based battery life attenuation rate: The ratio (%) of battery life under rated load to battery life under no load. The lower the attenuation rate, the better, and it should be ≥80%.

[0220] 3.5. Adaptability and Reliability Indicators

[0221] Measuring the motion stability and reliability of an entity under different physical environments is crucial for the deployment of industrial and mobile entities in complex scenarios, and can be categorized into two types: environment adaptation and motion reliability.

[0222] (1) Environmental adaptability

[0223] Ground adaptability: Success rate of driving / movement on different ground surfaces (cement / tile / carpet / uneven ground), and the range of adaptability of ground friction coefficient;

[0224] Temperature and humidity adaptability: It can stably operate within the range of ambient temperature (°C) and humidity (RH). Industrial models are adapted to -20°C to 60°C and 0 to 95% RH, while service models are adapted to 0°C to 40°C and 30 to 85% RH.

[0225] Dust and water resistance rating (IP): IP rating of the body motion system. Industrial bodies require IP65 or above, and service bodies require IP54 or above.

[0226] Anti-interference capability: The motion accuracy retention rate (%) ≥90% under electromagnetic interference and vibration environment is considered excellent.

[0227] (2) Motion reliability

[0228] Mean Time Between Failures (MTBF): The time (in hours) during which the machine moves continuously without failure. For industrial machines, the requirement is ≥20,000 hours; for mobile machines, the requirement is ≥10,000 hours.

[0229] Motion performance degradation rate: After the body has been working continuously for a specified period of time (e.g., 1000h), a degradation percentage of position / repetition accuracy and speed of ≤5% is considered excellent;

[0230] Start-stop reliability: A success rate (%) of ≥99.9% after multiple start-stop cycles (≥10,000 times) is considered excellent;

[0231] Motion accuracy after fault recovery: After the body motion system fault is recovered, the ratio (%) of the actual pose accuracy to that before the fault is ≥95% is considered excellent.

[0232] 4. Intelligent performance indicators

[0233] The intelligence of embodied intelligence is its core feature that distinguishes it from traditional robots. It is manifested in its ability to make autonomous decisions, adapt to different scenarios, learn and evolve, and solve problems.

[0234] 4.1. Cognitive Decision-Making Performance Indicators

[0235] Intelligent cognitive decision-making is the underlying capability of embodied intelligence, which is manifested in the comprehensive understanding and logical decision-making of environmental information, task requirements and constraints, focusing on task decomposition, logical judgment and intention reasoning capabilities.

[0236] Success rate of autonomous decomposition of complex tasks: For complex tasks with multiple steps and multiple objectives, the system can autonomously decompose them into executable single-step sub-tasks without manual intervention, and plan a reasonable execution order. ≥95% (for 3-5 step conventional complex tasks, such as "boiling water in the kitchen → getting a water glass → pouring warm water → placing it on the living room table"); ≥90% (for 5-8 step complex tasks, including sequential constraints, such as "first tidying up the table → then getting the plate → placing it on the table → getting the matching cutlery").

[0237] Fuzzy task intent reasoning accuracy: For task instructions without clear objectives or with vague, colloquial language, the accuracy of reasoning the core execution intent by combining the scene environment, user habits, and common sense logic. ≥90% (for scenario-based fuzzy tasks, such as "tidy up the living room" in home use / "organize the shelves" in commercial use / "put the parts back in their place" in industry); ≥85% (for cross-scenario fuzzy tasks, such as "tidy up here and prepare for the next task").

[0238] Decision rationality under multiple constraints: In scenarios with multiple overlapping safety constraints, environmental constraints, and operational constraints, whether the execution decision made conforms to the principles of "optimal path, minimum loss, and highest efficiency" is quantified by a decision rationality score (1-5 points). ≥4.5 points (single constraint); ≥4 points (double constraint); ≥3.5 points (triple constraint, such as "getting a water cup + avoiding hitting children + avoiding debris on the ground").

[0239] Uninstructed autonomous decision triggering rate: When there is a clear need in the scenario but no user has issued an instruction (such as a water cup falling on the table / goods falling off the shelf), the success rate of autonomously triggering decisions and executing them is ≥90%.

[0240] Decision response delay: The time from perceiving complete information to making an execution decision is ≤1s (simple tasks / ≤2s (complex tasks).

[0241] 4.2. Scene Autonomous Adaptation Performance Indicators

[0242] Intelligent scene autonomous adaptation is the key to the implementation of embodied intelligence in real-world scenarios. It is reflected in the ability to autonomously adapt to dynamic, unfamiliar, and changing scenarios without the need for manual debugging or reprogramming, focusing on environment adaptation, task adaptation, and working condition adaptation.

[0243] Success rate of adapting to dynamic changes in known scenarios: In a known scenario that has been learned, the success rate of autonomously adjusting execution strategies and completing tasks when dynamic changes occur in the environment (such as furniture movement, the addition of obstacles, or changes in the location of the task target). ≥98% (minor changes, such as furniture movement ≤30cm); ≥95% (moderate changes, such as the addition of ≤3 obstacles); ≥90% (severe changes, such as adjustments to the layout of the work area);

[0244] Zero-debugging adaptation success rate in unfamiliar scenarios: When entering an unfamiliar scenario without prior training, the success rate of completing basic and general tasks solely through on-site perception (basic tasks: obstacle avoidance, simple retrieval; general tasks: desktop organization, shelf restocking) is as follows: ≥90% (basic tasks in unfamiliar scenarios); ≥85% (general tasks in unfamiliar scenarios); ≥80% (complex general tasks in unfamiliar scenarios).

[0245] Autonomous Adaptability to Different Working Conditions: The ability to autonomously adjust motion parameters, operational force, and perception strategies while maintaining task completion quality in the face of different working conditions (such as changes in load, ambient lighting, and target material). This is quantified by the accuracy retention rate of working condition adaptation. ≥95% (minor working condition changes, such as load ±1kg / lighting ±50lux); ≥90% (moderate working condition changes, such as load ±2kg / lighting ±100lux).

[0246] Cross-scenario job reuse rate: The execution strategy reuse rate of the same type of job (such as "picking and putting items") in different scenarios is ≥90% (no need to relearn the core actions);

[0247] Scene adaptation learning time: For new and complex scenes, the time to complete basic adaptation learning is ≤5 minutes (autonomous perception learning, no manual annotation).

[0248] 4.3. Learning Evolutionary Performance Indicators

[0249] Learning evolution is the core manifestation of embodied intelligence. Unlike the traditional robots whose capabilities are "fixed at the factory," it is characterized by autonomous learning, experience accumulation, and capability evolution. It does not require manual labeling or model retraining and focuses on trial-and-error learning, experience reuse, and small-sample learning.

[0250] Success rate of autonomous learning on unknown tasks with small sample sizes: This refers to the success rate of learning and independently reproducing an unknown task through a small number of demonstrations (1-3 human demonstrations) or autonomous trial and error (≤5 trial and error attempts). ≥95% (learning a simple unknown task with 1 human demonstration); ≥90% (learning a complex unknown task with 3 human demonstrations); ≥85% (learning a basic unknown task through 5 autonomous trial and error attempts).

[0251] Accuracy of Accumulated and Reusable Work Experience: This refers to the accuracy rate of reusing previously completed task experience (such as work paths, operational intensity, and strategies for handling scenario constraints) in subsequent similar tasks. ≥98% (Reuse of experience for similar tasks in the same scenario); ≥93% (Reuse of experience for similar tasks in similar scenarios); ≥90% (Reuse of experience for similar tasks in different scenarios).

[0252] Trial-and-error learning resilience: When minor execution failures occur during the task (such as falling objects, path planning errors), the success rate of independently analyzing the reasons for the failure, adjusting strategies, and re-executing is ≤2 times. ≥95% (successfully corrected after 1 failure); ≥90% (successfully corrected after 2 failures), with no 3 or more repeated failures.

[0253] Efficiency of capability evolution: For the same type of task, the efficiency of completing the task for the 10th time is ≥30% higher than that for the first time (reduced time / optimized steps);

[0254] Unsupervised self-learning capability: Success rate of self-learning and optimization of recurring scenarios / tasks ≥80% without human intervention or demonstration;

[0255] Learning and memory retention time: For a learned but unknown task, without repeated execution, the memory retention time is ≥7 days and the success rate of re-execution is ≥90%.

[0256] 4.4. Intelligent Collaborative Performance Indicators

[0257] The intelligence of embodied intelligence is not only reflected in the autonomous operation of a single robot, but also in the intelligent collaboration with humans and other robots. The core is intention collaboration, action collaboration, and resource collaboration, avoiding "mechanical cooperation" and realizing "intelligent linkage", which is suitable for commercial cluster operations, industrial human-machine collaboration, and home multi-device linkage scenarios.

[0258] Human-robot collaborative operation intelligence coordination level: In human-robot collaborative operations, the robot's ability to proactively perceive human operational intentions, coordinate with human movement rhythms, and compensate for human shortcomings is quantified by a coordination level score (1-5 points). ≥4.5 points (light industrial assembly / commercial shelving organization); ≥4 points (complex household chores collaboration)

[0259] Success rate of collaborative decision-making in multi-robot swarm operations: When multiple embodied intelligent robots form a swarm to complete distributed tasks (such as multi-area replenishment in supermarkets, multi-workstation handling in warehouses, and multi-process assembly in industries), the success rate of autonomous task allocation, path planning, and resource scheduling collaborative decision-making is as follows: ≥95% (2-3 robot swarms, simple distributed tasks); ≥90% (4-5 robot swarms, complex distributed tasks).

[0260] Cross-device intelligent linkage accuracy: The accuracy rate of autonomous linkage with intelligent devices in the scenario (such as smart home devices, commercial smart shelves, and industrial smart production line equipment) without manual configuration, triggering equipment actions based on operational needs. ≥98% (linkage with intelligent devices of the same brand); ≥93% (linkage with universal intelligent devices across brands).

[0261] Human-machine collaborative work efficiency improvement rate: Compared with human solo work, the efficiency improvement of human-machine intelligent collaborative work is ≥20% (reduced time / increased workload).

[0262] Multi-machine cluster operation load balancing rate: The load deviation of each robot in the cluster is ≤20% (to avoid individual robots being overloaded and others being idle);

[0263] Cross-device linkage response latency: The time from when the robot triggers the linkage command to when the smart device executes the action is ≤500ms.

[0264] 4.5. Global Planning and Problem Solving Performance Indicators

[0265] Embodied intelligence's global planning and problem-solving capabilities are manifested in task planning from a global perspective, autonomous resolution of unexpected problems, and dynamic optimization of the operation process, focusing on global planning, self-resolvement of faults, and dynamic optimization.

[0266] Global planning optimality for multi-objective tasks: When faced with multi-objective, multi-regional tasks, the system autonomously optimizes global path planning and task execution sequence planning, quantified by a planning optimality score (1-5 points) (optimality: shortest path, least time, no duplicate paths). ≥4.5 points (2-3 objectives); ≥4 points (4-5 objectives); ≥3.5 points (more than 5 objectives).

[0267] Success rate of autonomous problem resolution: When unexpected problems arise during operation (such as missing task targets, minor equipment malfunctions, or sudden environmental changes), the robot can autonomously analyze the problem, develop solutions, and complete the task without human intervention. ≥95% (minor unexpected problems, such as target position deviation or minor sensor interference); ≥90% (moderate unexpected problems, such as missing task targets or partial path blockages); ≥85% (severe unexpected problems, such as minor robot malfunctions or temporary closure of the work area).

[0268] Dynamic optimization capability during job execution: During job execution, the system autonomously and dynamically adjusts its planning strategy based on real-time scenario changes (such as sudden path congestion, insertion of new tasks, or low job efficiency) to improve job efficiency and reduce resource consumption. This is quantified by the dynamic optimization efficiency improvement rate: ≥15% (dynamic optimization efficiency improvement under minor scenario changes); ≥10% (dynamic optimization efficiency improvement under moderate scenario changes).

[0269] Untargeted global autonomous inspection success rate: When there is no clear operational objective, the success rate of completing the global autonomous inspection according to the priority of "scenario safety, equipment status, and environmental cleanliness" is ≥90%;

[0270] Fault self-diagnosis and early warning accuracy: The self-diagnosis accuracy for minor faults (such as low battery or slight sensor deviation) is ≥99%, and the timeliness of early warning is ≥95% (early warning time ≥30s).

[0271] Multi-task priority autonomous determination accuracy: When multiple tasks are triggered simultaneously, the accuracy of autonomously determining the execution priority based on the principles of "safety first, user needs first, and efficiency first" is ≥95%.

[0272] It should be noted that the above-described method embodiments, or the various possible implementations of the method embodiments, can be executed individually, or, provided there is no conflict, they can be combined with each other. The specific implementation can be determined according to actual usage requirements, and this application embodiment does not impose any restrictions on this.

[0273] As can be seen, the above mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, the embodiments of this application provide corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the modules and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0274] This application embodiment can divide the embodied intelligent device control device into functional modules according to the above method example. For example, each function can be divided into a separate functional module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. Optionally, the module division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0275] In some embodiments, this application also provides a avatar smart device control apparatus. This avatar smart device control apparatus may include one or more functional modules for implementing the avatar smart device control method of the above method embodiments.

[0276] For example, Figure 6 This is a schematic diagram of the structure of a embodied intelligent device control device provided in an embodiment of this application. Figure 6 As shown, the embodied intelligent device control device 900 includes: a first transmitting module 901, a receiving module 902, a comparison module 903, and a second transmitting module 904.

[0277] The first sending module 901 is used to send a test request to the embodied smart device, and the test request includes embodied smart test cases.

[0278] The receiving module 902 is configured to receive the embodied intelligence test response sent by the embodied intelligence device; wherein the embodied intelligence test response includes test values ​​of at least one target task indicator, the test values ​​being the test values ​​output by the embodied intelligence device after executing the embodied intelligence test example; the at least one target task is determined from a preset task indicator set of the embodied intelligence device based on the embodied intelligence test example; the preset task indicator set includes task indicators corresponding to the perception capability, interaction capability, motion control capability, and intelligent decision-making capability of the embodied intelligence device under the dimensions of physical ontology, environmental perception and interaction, and intelligence level, respectively.

[0279] The comparison module 903 is used to compare the test value corresponding to each target task indicator with the preset benchmark value to obtain comparison result information. The comparison result information is used to characterize the capabilities of the aforementioned embodied intelligent device.

[0280] The second sending module 904 is used to send control commands to the aforementioned embodied intelligent device based on the comparison result information.

[0281] The embodied intelligent device control device provided in this application can proactively send test requests for test cases to the embodied intelligent device, and receive test responses containing target task indicator test values ​​after the device executes the test cases. The target task indicators are selected as needed from a preset set of task indicators. This indicator set comprehensively covers three dimensions: physical entity, environmental perception and interaction, and level of intelligence, fully encompassing evaluation indicators corresponding to various capabilities of the device, including perception, interaction, motion control, and intelligent decision-making. Accurate test values ​​are obtained based on multi-dimensional measured data. Subsequently, the test values ​​of each indicator are compared with preset benchmark values ​​to form a comparison result that truly and objectively reflects the actual comprehensive capabilities of the device. Adaptive control commands are then generated and sent to the device based on this accurate comparison result. The entire process abandons the traditional single-indicator evaluation mode, relying on multi-dimensional test data to ensure the authenticity and reliability of the evaluation results. Based on this, the device performance parameters are optimized and adjusted, gradually improving the accuracy of performance parameter adjustments for the embodied intelligent device.

[0282] In some embodiments, the aforementioned capability assessment information includes first capability information of the perception capability of the embodied intelligent device, second capability information of the interaction capability of the embodied intelligent device, third capability information of the motion control capability of the embodied intelligent device, and fourth capability information of the intelligent decision-making capability of the embodied intelligent device.

[0283] The second sending module is specifically used to generate control commands for the embodied intelligent device based on the first capability information, the second capability information, the third capability information, and the fourth capability information.

[0284] The aforementioned control instructions include at least one of the following: a control strategy for the perception capabilities of the aforementioned embodied intelligent device, a control strategy for the interaction capabilities of the aforementioned embodied intelligent device, a control strategy for the motion control capabilities of the aforementioned embodied intelligent device, and a control strategy for the intelligent decision-making capabilities of the aforementioned embodied intelligent device.

[0285] In other embodiments, the above-described embodied smart device control device further includes:

[0286] The third sending module is used to send a task execution instruction to the aforementioned embodied intelligent device after receiving the control feedback sent by the aforementioned embodied intelligent device. The task execution instruction indicates that the target task to be executed corresponds to the same application scenario as the aforementioned embodied intelligent test case.

[0287] In some other embodiments, the aforementioned preset task indicator set includes perception performance indicators, interaction performance indicators, body motion and control performance indicators, and intelligent decision-making performance indicators. Specifically, the perception performance indicators characterize the perception capability of the embodied intelligent device in at least one of the following dimensions: physical body dimension, environmental perception and interaction dimension, and intelligence level dimension; the interaction performance indicators characterize the interaction capability of the embodied intelligent device in at least one of the following dimensions: physical body dimension, environmental perception and interaction dimension, and intelligence level dimension; the body motion and control performance indicators characterize the motion control capability of the embodied intelligent device in at least one of the following dimensions: physical body dimension, environmental perception and interaction dimension, and intelligence level dimension; and the intelligent decision-making performance indicators characterize the intelligent decision-making capability of the embodied intelligent device in at least one of the following dimensions: physical body dimension, environmental perception and interaction dimension, and intelligence level dimension.

[0288] In some other embodiments, the above-mentioned perception performance indicators include at least one of the following: environmental perception indicators, target perception indicators, force / touch perception indicators, and perception fusion and dynamic perception indicators.

[0289] In some other embodiments, the above-mentioned interaction performance indicators include at least one of the following: voice interaction indicators, visual interaction indicators, physical contact interaction indicators, multimodal fusion interaction indicators, and contextualized semantic interaction indicators.

[0290] In some other embodiments, the above-mentioned body motion and control performance indicators include at least one of the following: body pose accuracy indicator, motion stability and smoothness indicator, dynamic characteristic indicator, load and endurance indicator, and adaptability and reliability indicator.

[0291] In some other embodiments, the above-mentioned intelligent decision-making performance indicators include at least one of the following: cognitive decision-making performance indicators, scene autonomous adaptation performance indicators, learning evolution performance indicators, intelligent collaboration performance indicators, and global planning and problem-solving performance indicators.

[0292] It should be noted that the embodied intelligent device control device can implement all the processes implemented in the above method embodiments and achieve the same beneficial effects. To avoid repetition, it will not be described again here.

[0293] In the case where the functions of the integrated modules described above are implemented in hardware, this application provides a possible structural schematic diagram of the electronic device involved in the above embodiments. For example... Figure 7 As shown, the electronic device 90 includes: a processor 92, a communication interface 93, and a bus 94. Optionally, the electronic device 90 may also include a memory 91.

[0294] Processor 92 may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 92 may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 92 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0295] Communication interface 93 is used to connect with other devices via a communication network. This communication network can be Ethernet, wireless access network, wireless local area network (WLAN), etc.

[0296] The memory 91 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0297] As one possible implementation, the memory 91 can exist independently of the processor 92. The memory 91 can be connected to the processor 92 via a bus 94 and is used to store instructions or program code. When the processor 92 calls and executes the instructions or program code stored in the memory 91, it can implement the personalized intelligent device control method provided in the embodiments of this application.

[0298] In another possible implementation, memory 91 can also be integrated with processor 92.

[0299] Bus 94 can be an Extended Industry Standard Architecture (EISA) bus, etc. Bus 94 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0300] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the service calling device can be divided into different functional modules to complete all or part of the functions described above.

[0301] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described embodiments of the intelligent device control method and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0302] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0303] This application also provides a readable storage medium storing a program or instructions that, when executed by a computer, implement the embodied intelligent device control method provided in the above embodiments. It is understood that all or part of the processes in the above method embodiments can be executed by computer instructions instructing related hardware; the readable storage medium can be any of the foregoing embodiments or memory; the readable storage medium can also be an external storage device of the service invocation device, such as a plug-in hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, flash card, etc., equipped on the service invocation device. Further, the readable storage medium can include both internal storage units of the service invocation device and external storage devices. The readable storage medium is used to store the computer program and other programs and data required by the service invocation device. The readable storage medium can also be used to temporarily store data that has been output or will be output.

[0304] This application also provides a computer program product, which is stored in a storage medium and, when executed by a computer, implements the embodied intelligent device control method provided in the above embodiments.

[0305] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0306] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0307] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A somatic intelligent device control method, characterized by, include: Send a test request to the embodied intelligent device, the test request including embodied intelligent test cases; The system receives an embodied intelligence test response sent by the embodied intelligence device; wherein the embodied intelligence test response includes a test value of at least one target task indicator, the test value being the test value output by the embodied intelligence device after executing the embodied intelligence test example; the at least one target task is determined from a preset task indicator set of the embodied intelligence device based on the embodied intelligence test example; the preset task indicator set includes task indicators corresponding to the perception capability, interaction capability, motion control capability, and intelligent decision-making capability of the embodied intelligence device under the dimensions of physical ontology, environmental perception and interaction, and intelligence level, respectively. The test value corresponding to each target task indicator is compared with a preset benchmark value to obtain comparison result information, which is used to characterize the capabilities of the embodied intelligent device. Based on the comparison results, control commands are sent to the embodied smart device.

2. The embodied intelligence device control method of claim 1, wherein, The capability assessment information includes first capability information of the perception capability of the embodied intelligent device, second capability information of the interaction capability of the embodied intelligent device, third capability information of the motion control capability of the embodied intelligent device, and fourth capability information of the intelligent decision-making capability of the embodied intelligent device. Sending control commands to the embody smart device based on the comparison result information includes: Based on the first capability information, the second capability information, the third capability information, and the fourth capability information, control commands are sent to the embodied intelligent device; The control instructions include at least one of the following: a control strategy for the perception capabilities of the embodied intelligent device, a control strategy for the interaction capabilities of the embodied intelligent device, a control strategy for the motion control capabilities of the embodied intelligent device, and a control strategy for the intelligent decision-making capabilities of the embodied intelligent device.

3. The method for controlling a personal intelligent device according to claim 1 or 2, characterized in that, After sending the control command to the embodied smart device based on the comparison result information, the method further includes: After receiving control feedback from the embodied intelligent device, a task execution instruction is sent to the embodied intelligent device. The task execution instruction indicates that the target task to be executed corresponds to the same application scenario as the embodied intelligent test case.

4. The method for controlling a personal intelligent device according to claim 1, characterized in that, The preset task indicator set includes perception performance indicators, interaction performance indicators, body motion and control performance indicators, and intelligent decision-making performance indicators. The perception performance index is used to characterize the perception capability of the embodied intelligent device in at least one of the dimensions of physical ontology, environmental perception and interaction, and intelligence level; the interaction performance index is used to characterize the interaction capability of the embodied intelligent device in at least one of the dimensions of physical ontology, environmental perception and interaction, and intelligence level; the ontology motion and control performance index is used to characterize the motion control capability of the embodied intelligent device in at least one of the dimensions of physical ontology, environmental perception and interaction, and intelligence level; and the intelligent decision-making performance index is used to characterize the intelligent decision-making capability of the embodied intelligent device in at least one of the dimensions of physical ontology, environmental perception and interaction, and intelligence level.

5. The method for controlling a personal intelligent device according to claim 4, characterized in that, The perception performance indicators include at least one of the following: environmental perception indicators, target perception indicators, force / touch perception indicators, and perception fusion and dynamic perception indicators. The interaction performance metrics include at least one of the following: voice interaction metrics, visual interaction metrics, physical contact interaction metrics, multimodal fusion interaction metrics, and contextualized semantic interaction metrics. The body motion and control performance indicators include at least one of the following: body pose accuracy, motion stability and smoothness, dynamic characteristics, load and endurance, and adaptability and reliability. The intelligent decision-making performance indicators include at least one of the following: cognitive decision-making performance indicators, scenario autonomous adaptation performance indicators, learning and evolution performance indicators, intelligent collaboration performance indicators, and global planning and problem-solving performance indicators.

6. A control device for a personal intelligent device, characterized in that, include: The first sending module is used to send a test request to the embodied intelligent device, the test request including embodied intelligent test cases; A receiving module is configured to receive an embodied intelligence test response sent by the embodied intelligence device; wherein, the embodied intelligence test response includes test values ​​of at least one target task indicator, the test values ​​being the test values ​​output by the embodied intelligence device after executing the embodied intelligence test example; the at least one target task is determined from a preset task indicator set of the embodied intelligence device based on the embodied intelligence test example; the preset task indicator set includes task indicators corresponding to the perception capability, interaction capability, motion control capability, and intelligent decision-making capability of the embodied intelligence device under the dimensions of physical ontology, environmental perception and interaction, and intelligence level, respectively; The comparison module is used to compare the test value corresponding to each target task indicator with a preset benchmark value to obtain comparison result information, which is used to characterize the capabilities of the embodied intelligent device. The second sending module is used to send control commands to the embodied smart device based on the comparison result information.

7. The embodied intelligent device control device according to claim 6, characterized in that, The capability assessment information includes first capability information of the perception capability of the embodied intelligent device, second capability information of the interaction capability of the embodied intelligent device, third capability information of the motion control capability of the embodied intelligent device, and fourth capability information of the intelligent decision-making capability of the embodied intelligent device. The second sending module is specifically used to send control commands to the embodied smart device based on the first capability information, the second capability information, the third capability information, and the fourth capability information; The control instructions include at least one of the following: a control strategy for the perception capabilities of the embodied intelligent device, a control strategy for the interaction capabilities of the embodied intelligent device, a control strategy for the motion control capabilities of the embodied intelligent device, and a control strategy for the intelligent decision-making capabilities of the embodied intelligent device.

8. The embodied intelligent device control device according to claim 6 or 7, characterized in that, The embodied intelligent device control device also includes: The third sending module is used to send a task execution instruction to the embodied intelligent device after receiving the control feedback sent by the embodied intelligent device. The task execution instruction indicates that the target task to be executed corresponds to the same application scenario as the embodied intelligent test case.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing programs or instructions that can run on the processor, the programs or instructions being executed by the processor to implement the embodied intelligent device control method as described in any one of claims 1-5.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a computer, implement the embodied intelligent device control method as described in any one of claims 1-5.

11. A computer program product, characterized in that, The computer program product is stored in a storage medium, and when executed by a computer, the computer program product implements the embodied intelligent device control method as described in any one of claims 1-5.

Images