51 results about "Embodied intelligence" patented technology

The application relates to the field of natural language processing and decision intelligence, and particularly relates to a large language model reasoning method based on time series difference learning and rule enhancement, which is widely applied to automatic planning, intelligent question answering, embodied intelligence and the like. The method comprises task trajectory sampling, domain knowledge induction, domain rule extraction and large language model reasoning based on rule enhancement. For test task data, the most relevant historical task is matched based on vector retrieval, and a domain rule set corresponding to the task is obtained. The rule is rewritten in natural language by the large language model itself, so that the rule is more interpretable and adaptable. Finally, the optimized rule set is integrated into the large language model reasoning prompt text, so as to optimize the reasoning quality and stability.

The application relates to the field of embodied intelligence and autonomous navigation, in particular to a wheelchair navigation method and device, equipment and medium, wherein the method comprises the following steps: acquiring a current environment image, a navigation instruction and a historical environment image of a wheelchair; extracting an environment structure feature of the current environment image, and determining whether a target environment is an indoor environment or an outdoor environment based on the environment structure feature; when the target environment is the indoor environment, generating a bird's-eye view feature map according to the current environment image, and combining the navigation instruction to navigate; when the target environment is the outdoor environment, predicting a three-dimensional point cloud, a relative pose and a driving trajectory based on the current environment image and the historical environment image, and combining the navigation instruction to navigate. Therefore, the problems of poor unified representation capability of perception, mapping and decision, insufficient spatial geometric expression and real-time performance, and lack of continuous environment updating and memory capability in the related art are solved.

The application provides an entropy-aware robot control method and system based on expert demonstration, and relates to the field of embodied intelligence technology, which comprises obtaining a task instruction input by a user, observation data and ontology perception data of a robot at a current time; cleaning and feature extracting the data to obtain multiple features, and retrieving a reference video in an expert reference video database; encoding and splicing the multiple features to generate a multi-modal feature representation; feature extracting the reference video to obtain reference video features; processing the multi-modal feature representation and the reference video features through a skill generation model to obtain a discrete skill codebook index probability; calculating an information entropy of the probability, and determining a sampling candidate number of a current candidate skill based on the information entropy; generating a skill token sequence based on the discrete skill codebook index probability and the sampling candidate number; and finally decoding the skill token sequence to obtain a robot action sequence. The control method provided by the application can guarantee accurate control of a robot in a complex environment.

The application belongs to the technical field of embodied intelligence and cloud edge collaboration, and specifically discloses a cloud edge collaborative edge control device and method supporting AI natural language communication. The device disclosed by the application comprises a SoC master control unit, an edge computing power unit, a multi-mode communication unit, a natural language dialogue interaction unit, an audio and video output unit and a data feedback unit; the edge computing power unit is bidirectionally connected with the SoC master control unit, local processing sensors and voice signals are processed, an actuator is controlled and data feedback is received, and an embodied intelligence closed loop is formed; the multi-mode communication unit supports at least two communication modes and dynamically switches; the natural language dialogue interaction unit runs a lightweight AI model, and realizes multi-round dialogue of non-fixed format natural language. The application solves the problems of single communication, insufficient computing power, lack of video interface and open loop control. The application can be widely applied in the fields of smart home, industrial Internet of Things and robots.

The application discloses a kind of robot task planning system, method and device, it is related to embodied intelligence field, to improve the success rate of robot long-range task planning efficiency and implementation.System includes expert skill base, encapsulates multiple independent skill modules;Vector knowledge base, maintain historical scene data;Decision brain, receive and understand natural language instruction, implement instruction based on historical scene data;Determine the skill module required to be called in implementation instruction;In the process of implementing instruction, according to current task progress, only activate the skill module required currently, load required resource for the activated skill module immediately, resource includes at least one of historical scene data, historical context data, current scene data;Current activated skill module and the operation result of current skill module are written into historical context data.The application greatly reduces the reasoning burden, improves the success rate of long-range task planning efficiency and execution.

The application provides an entropy-aware robot control method and system based on expert demonstration, and relates to the field of embodied intelligence technology, which comprises obtaining a task instruction input by a user, observation data and ontology perception data of a robot at a current time; cleaning and feature extracting the data to obtain multiple features, and retrieving a reference video in an expert reference video database; encoding and splicing the multiple features to generate a multi-modal feature representation; feature extracting the reference video to obtain reference video features; processing the multi-modal feature representation and the reference video features through a skill generation model to obtain a discrete skill codebook index probability; calculating an information entropy of the probability, and determining a sampling candidate number of a current candidate skill based on the information entropy; generating a skill token sequence based on the discrete skill codebook index probability and the sampling candidate number; and finally decoding the skill token sequence to obtain a robot action sequence. The control method provided by the application can guarantee accurate control of a robot in a complex environment.

This invention relates to the field of embodied intelligence technology and discloses an embodied intelligence edge-cloud collaborative control method and system. The method includes: receiving environmental perception data and task instructions collected by an embodied intelligence device; determining the task category based on the task instructions; establishing an edge-side execution record queue and a cloud-side execution record queue for each task category; the execution record queue is a fixed-length circular queue structure of length N, used to record the results of the most recent N task executions; calculating the edge-side execution success rate and the cloud-side execution success rate based on the queue data; when the difference between the two is greater than a preset threshold T, selecting the side with the higher success rate as the current task execution position; and updating the corresponding circular queue after the task execution is completed. This invention achieves the updability and stability of edge-cloud execution decisions through a fixed-length statistical structure.

The application discloses a double-constraint reinforcement fine-tuning method and system for an ultrasonic robot VLA model, and relates to the field of embodied intelligence. The double-constraint reinforcement fine-tuning method first uses an expert demonstration data set to perform offline initialization on a VLA strategy network and a value network, and obtains an offline basic strategy through offline consistency policy Q learning optimization. When the robot is online exploring, a time-series value model is independently trained, and sparse final task rewards are converted into dense time-series rewards for each step. Then, a proximal policy optimization framework is adopted, a consistency constraint is introduced, and the online strategy is limited to deviate from the basic strategy through KL divergence, so that the safe exploration range of the probe contact process is ensured. Meanwhile, the contact force is monitored in real time, and a penalty and manual reset are triggered when the contact force exceeds a limit. The application solves the problems that the existing VLA model is difficult to utilize sparse rewards and lacks safe constraints in ultrasonic application, and improves the learning efficiency and stability under the premise of ensuring operation safety.

This invention relates to the field of network and information security technology, specifically to a method for adversarial prompt purification and semantic perturbation defense against embodied intelligence. It involves constructing a jailbreak token lexicon and an embodied intelligence synonym lexicon to define attack characteristics. Then, it receives user commands, purifies them by removing obvious malicious elements, and standardizes the commands. Furthermore, it applies semantic perturbation defense to the purified commands by replacing words with synonyms. This disrupts the adversarial token sequence while preserving the original semantic intent, thereby reducing the likelihood of jailbreak attacks. This invention proactively weakens the effectiveness of jailbreak attacks at the input source, transforming potentially malicious prompts into neutral ones to eliminate hidden or confusing commands while retaining the meaning of the dialogue. Unlike direct rejection, this invention cleans up commands before the embodied intelligence model uses them.

The application discloses a native creativity realization method, system and circuit of an electronic system, and belongs to the technical field of hardware management and control chips. The method comprises the following steps: when a behavior intention is generated but there is no directly-matched execution scheme, a hardware logic circuit which is independent of instruction execution automatically retrieves a relevant experience entry from an experience set in the same physical chip package and recombines the experience entry to generate a brand-new candidate behavior scheme; the candidate scheme must be verified by an independent decision circuit which is physically separated from a behavior generation path before execution; the scheme which fails to pass the verification is permanently prohibited from output under standard working conditions; and the new scheme which passes the verification and is successfully executed is automatically added to the experience set. The application realizes the cross-scene recombination of experiences and the safe generation of brand-new behavior schemes by means of hardware, and is suitable for electronic systems such as intelligent robots, embodied intelligence and deep space exploration which need to cope with unknown scenes offline and autonomously.

This invention discloses a method, system, terminal, and medium for generating embodied intelligence training corpus based on adversarial data governance, relating to the field of robot learning technology. The key technical points are: acquiring multimodal data streams generated by robots during human-machine collaborative interaction; performing time alignment processing on the multimodal data streams to obtain aligned standardized data streams; identifying high-value data segments from the standardized data streams using a preset value evaluation function; and performing coordinate regularization processing on the identified high-value data segments to generate training corpus. This invention introduces an adversarial operator role into the human-machine collaborative loop, systematically injecting multidimensional perturbations and actively generating high-difficulty edge scenario data containing instability-recovery logic. This results in training corpus that not only has extremely high information density, covering failure-recovery manifolds that are difficult to collect using traditional data, but also possesses strong generalization characteristics.

This invention relates to the field of robotics, and in particular to an AI inspection robot integrating embodied intelligence and deep learning. The robot includes a base plate with rollers mounted on its bottom surface, a mobile chassis system, a posture adjustment and perception system, and an embodied intelligence decision-making system. The rollers are connected to drive motors. A lifting frame is mounted on the upper surface of the base plate, and the posture adjustment and perception system is installed on the lifting frame. A camera is mounted on the lifting frame, and a second motor is mounted on the top of the lifting frame for adjusting the camera's height. This invention enables natural language interaction, one-time batch removal of obstructed I-beams, and unobstructed comprehensive inspection, significantly improving the intelligence level and operational efficiency of I-beam sampling inspection.

This application relates to the field of embodied intelligence technology and discloses a multi-condition adaptive task scheduling method and system for cutting and grinding processes. The method includes: inputting collected multi-agent state information, workshop dust environment information, and standardized cutting and grinding task data into an initialized customized scheduling model to generate multiple feasible scheduling schemes, selecting the optimal scheme, and retaining multiple alternative schemes; each agent executes the cutting and grinding task according to the execution instructions, and the feedback information is processed hierarchically; if a trigger is detected, the adjusted scheduling scheme is transmitted to each agent for execution; a standardized training dataset is constructed, and a reinforcement learning training framework with the core requirements of the cutting and grinding process scheduling as the reward objective is built; the standardized training dataset is input into the framework for iterative training. This application improves the collaborative operation efficiency and stability of multiple embodied agents in the cutting and grinding process.

The application provides a kind of intelligent fire-fighting robot of fusion multi-modal embodied intelligence and its application control method, it aims at deeply fusing multi-modal perception fusion, hierarchical reinforcement learning, whole body motion planning, multi-agent collaboration and other frontier technologies, designs and realizes a kind of mobile fire duty robot (fusion multi-modal embodied intelligent intelligent fire-fighting robot and its application control method) with embodied intelligence, promotes the paradigm change of fire operation from "manual experience driving" to "machine autonomous intelligence". The mobile fire duty robot with embodied intelligence is a new generation of fire-fighting equipment form created by deeply fusing artificial intelligence, robotics and control theory. It realizes the leap from "perception-control" separation to "perception-cognition-decision-action" integration, is a key step to promote fire rescue to high intelligence and autonomy, and has great social value and application prospect.

This invention relates to the field of medical robotics and embodied intelligence, specifically to a dual-arm robot control method and electronic device based on VLA model and cognitive alignment of thought chain. The method includes the following steps: S1, multimodal task scene perception; S2, full-modal space alignment based on multidimensional state lexicalization and low-rank adaptive alignment; S3, dynamic intervention mechanism based on cognitive alignment assessment; S4, end-to-end joint space motion generation based on reparameterization; S5, compliant execution of cognitive-motor dual-loop closed-loop. This invention constructs a multimodal state vector by fusing robotic arm motion data, a six-dimensional force sensor at the end effector, and a third-view camera. Based on a deep learning fusion model, it performs feature extraction and correlation analysis to objectively and quantitatively assess the user's dual-arm motor ability. This allows for the capture of the subject's motion details and neuromuscular coordination characteristics, quantitative characterization of the subject's dual-arm motor ability, dynamic monitoring of the user's dual-arm motor ability progress, and scientific quantification.

This application discloses a robot, its trajectory correction method, and a training device for the correction capability of an embodied intelligent model. The trajectory correction method includes: acquiring perception data and task instructions after the robot makes an execution error; inputting the perception data and task instructions into a pre-trained VLA model with trajectory correction capability; wherein the VLA model is obtained by learning a correction trajectory; processing the perception data and task instructions using the VLA model to generate a correction action sequence; controlling the robot to sequentially execute multiple correction actions in the correction action sequence, so that the robot's actual trajectory gradually approaches and eventually returns to the standard trajectory. Applying this application can provide effective correction capabilities for embodied intelligent models and robots, improving the system robustness and practicality of embodied intelligent models and robots.

This invention provides a task execution method, apparatus, electronic device, and storage medium, relating to the field of embodied intelligence technology. The method includes: determining redundant feature components from the first task reference information based on the feature sources of feature components and the importance values ​​of each feature source, wherein the importance value of each feature source represents the magnitude of the influence of the feature component of that feature source on the successful output of the robot's expected action by the first VLA model; performing redundancy removal processing on the redundant feature components in the first task reference information to obtain processed first task reference information; inputting the processed first task reference information into the first VLA model to obtain the predicted action of the robot output by the first VLA model; and controlling the robot to execute the task based on the predicted action. Applying the solution provided by this invention can improve the robot's motion control frequency, thereby improving task execution performance.

This application discloses a cause-driven intention constraint calculation method and system, comprising: S210, detecting the reasoning intention of the embodied intelligence system; S220, identifying the decision of the embodied intelligence system; S230, verifying whether the decision of the embodied intelligence system conforms to constraint one; S240, if the decision of the embodied intelligence system does not conform to constraint one, verifying whether the reasoning intention of the embodied intelligence system conforms to constraint two; S250, if the reasoning intention of the embodied intelligence system conforms to constraint two, then replanning the decision of the embodied intelligence system. Based on the above steps, the reasoning evolution and control process of the embodied intelligence system can be realized around a preset core goal (such as environmental adaptation, task achievement, and continuous cognitive evolution), rather than aimless trial and error.

This invention relates to the field of embodied intelligence technology, and discloses a method for autonomous obstacle avoidance and task rescheduling for robotic dogs operating in high-risk areas of cement plants. This method employs multimodal sensor fusion to achieve real-time environmental perception, and constructs a dynamic risk field based on an offline-calibrated 2.5D grid and a lookup table method. Then, it uses hierarchical path planning—global A*-guided low-risk paths and local DWA real-time obstacle avoidance—for task path planning and execution. Combined with asynchronous task rescheduling, task sequence optimization is placed in an independent thread, supporting timeout returns of suboptimal solutions, and directly triggering preset templates for emergency events, improving the timeliness of response to sudden emergencies. This invention ensures the operational safety of robotic dogs in high-risk environments such as cement plants, improves the navigation robustness and task response of robotic dogs in dynamic high-risk environments, and enhances overall inspection efficiency.

This invention relates to the fields of embodied intelligence, multimodal computing, and unmanned navigation, specifically to a visual language navigation method and apparatus that integrates semantic enhancement and hierarchical decision-making. The method includes the following steps: receiving RGB images, depth images, and text commands as input; obtaining a semantic map with scene attributes, which is then further distilled; constructing a cross-modal semantic representation with the text commands; and then, using a global planning-local control collaborative strategy, fusing predicted actions, coordinates, and progress, and completing navigation under the guidance of a motion decision algorithm. The apparatus is used to execute the steps described in the above method. This invention effectively preserves the semantic space topology, solves the problem of multi-scale feature splicing imbalance, and simultaneously achieves a closed-loop fusion of command understanding, environmental perception, and motion control through the collaboration of global path planning and local behavior control, improving navigation coherence and increasing navigation success rate.

This invention relates to the field of embodied intelligent robots, and provides a management method and system for spatiotemporal multimodal data querying required for embodied intelligent training. The method includes: setting hardware metadata, which stores device definitions for multiple hardware devices, including performance definitions, category definitions, and flow mode definitions for multimodal devices; creating an embodied configuration, including a spatial graph, a hardware inventory registry, and hardware mounting configurations; automatically generating one or more data streams based on the device definitions and hardware mounting configurations, wherein the data streams continuously receive time-series observations generated by the hardware devices to form a time data stream, which includes a topology stream used to record topology events of hardware device installation and uninstallation; organizing the time data stream into segments, each segment being accompanied by a descriptive label representing a complete recording session; and, in response to receiving a data query request including one or more segments, traversing and executing the data query request to obtain the query results.

1. Name of the product in this design: Humanoid Robot. 2. Intended use of this design: for artificial intelligence education, humanoid robot education research, and embodied intelligence education research in primary and secondary schools or universities. 3. The key design feature of this product is its shape. 4. The image or photograph that best illustrates the design's key points: 3D rendering 1.

This invention discloses a grouped point cloud-driven arm-hand operating system. The arm-hand operating system includes a VR-based teleoperation data acquisition module, an environmental perception and encoding unit, and a two-layer feature fusion diffusion denoising model. Specifically: the VR-based teleoperation data acquisition module consists of a posture recognizer, a posture mapper, and a data storage module; the environmental perception and encoding unit consists of a point cloud grouper, a point cloud encoding module, and a state integrator; and the two-layer feature fusion diffusion denoising model consists of a two-layer feature fusion module and a diffusion denoising network. This arm-hand operating system can respond to environmental changes in real time, achieving comprehensive scene perception while quickly formulating task execution plans. This data acquisition-algorithm deployment collaborative framework effectively solves the challenge of balancing cost, speed, and accuracy in intelligent arm-hand operation, providing a cost-effective solution for embodied intelligence, robotic operations, and other fields.

This application discloses a dynamic adjustment method, electronic device, and storage medium for an embodied intelligence system. The method includes: acquiring real-time status data of target computing hardware deployed on an embodied intelligence carrier, and a target task execution plan; generating a scheduling strategy to guide resource allocation for the target computing hardware based on the real-time status data and the target task execution plan; dynamically adjusting the structure of a target neural network model running on the target computing hardware based on the scheduling strategy; and generating control instructions to drive the embodied intelligence carrier to execute the target task execution plan using the dynamically adjusted target neural network model. This application can dynamically adjust the structure of the target neural network model running on the computing hardware according to the real-time operating status of the computing hardware on the embodied intelligence carrier, enabling the embodied intelligence system to operate adaptively in resource-constrained environments.

The present application relates to a kind of embodied intelligent architecture and environmental adaptability learning method and device, a kind of embodied intelligent architecture, the embodied intelligent architecture includes sequentially connected from bottom to top core task anchoring layer, binary decision and check layer, dynamic strategy scheduling layer, basic cognitive function unit layer, reasoning state encoding layer and dynamic revision and reconstruction layer;A kind of embodied intelligent environmental adaptability learning method is applied to a kind of embodied intelligent architecture, comprising the following steps: S1, target anchoring;S2, pre-checking;S3, strategy routing;S4, module execution and state coding;S5, real-time checking;S6, dynamic revision.The present application is through six-layer closed-loop hierarchical cognitive architecture, signal interaction between levels, data transmission and collaborative work;Realize that embodied intelligence is actively monitored to the perception-decision-execution process, real-time checking, dynamic strategy scheduling and autonomous revision, improve the environmental adaptability of embodied intelligence, generalization ability, system reliability and explainability.

The application discloses a somatic intelligent data generation method and system based on physical feedback and large model enhancement, and belongs to the technical field of somatic intelligent data generation. The method comprises the following steps: establishing a rigid connection between a target object and an end effector of a mechanical arm in a simple environment and extracting initial visual features of the target object; generating a position true value label of the target object based on kinematic constraints of the mechanical arm and camera calibration parameters; actively collecting difficult example samples based on recognition confidence feedback; constructing a confusion scene according to the feature similarity of the target object and a background area; and generating natural language task instructions by using a multi-modal large model or an attribute recognition model, so as to form an image-text-action training data set. The application can effectively improve the label accuracy, difficult example coverage, background anti-interference capability and semantic richness of training data in an unstructured scene, and is suitable for somatic intelligent model training, robot grasping control and multi-modal interactive tasks.

The invention belongs to the technical field of water meter assembly. The invention provides a single-workbench water meter assembly method and system based on body intelligence. A four-layer framework of perception, prediction, planning and execution is adopted. The sensing module outputs a water meter part state and a structured scene map based on the 3D scene reconstruction network; the prediction module generates collision risk and assembly execution time prediction based on the Transform multi-task sequential network; the planning module generates an optimal assembly action sequence through hierarchical reinforcement learning; and the execution layer executes operation in a position-force hybrid control mode, and feeds back data in real time to form a closed loop. According to the scheme, the problems that traditional manual assembly is low in efficiency and existing automatic equipment lacks self-adaptive capacity are solved, the defects that sudden working conditions are difficult to deal with and multi-model adaptation is poor are overcome, the assembly efficiency and consistency are improved, manual intervention is reduced, assembly interruption and part damage are avoided, and flexible automation of the whole water meter assembly process is achieved.

The present application belongs to the field of power grid operation and maintenance robots, and discloses a multi-level man-machine cooperation embodied intelligent power grid maintenance robot system, which comprises a teleoperation hardware integrated module, an augmented reality guidance module, a data recording and intelligent learning module, and a multi-level man-machine cooperation decision module. By constructing four functional modules of teleoperation hardware integration, AR guidance, intelligent learning and multi-level man-machine cooperation decision, the embodied intelligence and multi-modal sensing technology are fused, a five-level man-machine cooperation pedigree is designed to realize dynamic allocation and smooth transition of control right, and the optimization control of the power maintenance robot is carried out in combination with the fault automatic degradation recovery and the operation data autonomous learning optimization mechanism, so as to dynamically adapt to the power line wiring, peeling and other maintenance operations, and improve the efficiency, safety and environmental adaptability of the power maintenance operation. The present application also discloses a power grid maintenance operation method based on the above-mentioned embodied intelligent power grid maintenance robot system.

The invention relates to a multi-modal large language model-based three-dimensional affordability positioning method and system for a body robot, and belongs to the technical field of robot perception and body intelligence. According to the scheme, a multi-modal large model availability positioning framework is provided, the framework effectively combines semantic understanding, multi-modal fusion and 3D space positioning through an end-to-end learning mode, and the features of 3D scene data and natural language instructions are deeply mined and fully utilized. A novel scene-level affordability data representation is constructed, so that robot interaction tasks, such as door opening, drawer pulling and object grabbing, in various complex unstructured environments become more accurate and efficient; according to the method, adaptive sampling and a coarse-to-fine reasoning thought are fused, the calculation efficiency and robustness of understanding of a robot on a large-scale three-dimensional scene are improved by optimizing a data retention strategy and enhancing space consistency, and the limitation of a traditional method under the complex background and the ambiguous instruction in the real world is effectively solved.

This invention provides a method and system for embodied intelligence target following trajectory annotation based on a large language model, belonging to the field of embodied intelligence technology. The method includes: acquiring image frames of an embodied agent following a target, and a question statement used to guide a large language model to identify the target in the image frames, and inputting these into the large language model for recognition to obtain the bounding box of the target; determining a set of point clouds within the image frame region, and filtering out the center point cloud corresponding to the center point of the bounding box from the point cloud set; determining the target's movement trajectory based on the center point cloud; generating the embodied agent's travel trajectory based on the embodied agent's positioning information; and concatenating the movement trajectory and the travel trajectory to obtain the ground truth trajectory of the target following. This invention addresses the technical problem that manually annotating the trajectory to obtain the data required for model training in embodied intelligence target following tasks results in low model iteration speed and affects task completion.