38 results about "Motion generation" patented technology

This invention belongs to the field of vibration control technology and discloses a viscous damper for controlling vertical vibration. The viscous damper adopts a double-cylinder, unsealed structure. The two cylinders are connected by a connecting rod hole in the middle of the cylinder partition. One end of the connecting rod is rigidly connected to a piston inside the second cylinder, and the other end passes through the cylinder partition and the top plate of the first cylinder before connecting to a structure. The bottom of the second cylinder is connected to another structure. When the two structures generate a vertical relative velocity, the connecting rod drives the piston to move in the damping fluid of the second cylinder, generating viscous resistance and suppressing the relative motion of the two structures. This invention does not require strict sealing, has a simple structure, low manufacturing cost, and better durability. The damping coefficient can be dynamically adjusted as needed by the comb teeth or the gradually changing inner cavity of the cylinder, resulting in higher efficiency compared to traditional dampers. This invention can effectively control the vertical or slightly inclined relative motion between various large structures such as bridges, cables, building internal structures, and marine engineering projects caused by various dynamic loads.

This invention provides a laser cutting machine and its vibration reduction method, relating to the field of laser cutting. The solution includes a machine bed with a gantry movable in the X-direction, and a cutting component movable in the Y-direction on the gantry. It also includes: a counterweight, movable in the X-direction within the machine bed, capable of moving synchronously with the gantry; a drive assembly, mounted on the machine bed, driving the counterweight; and a controller electrically connected to both the gantry's drive motor and the drive assembly. The controller can acquire the gantry's motion parameters in real time and calculate the counterweight's X-direction motion parameters, ensuring that the counter-inertial force generated by the counterweight's X-direction movement is equal in magnitude and opposite in direction to the impact force generated by the gantry's movement. This invention reduces the impact of processing vibration on processing quality and equipment lifespan.

The present invention relates to a device and method for generating context-aware motion using text and emotion embeddings. An electronic device according to one embodiment of the present invention includes a memory and a processor connected to the memory, and the processor receives input information including a text prompt instructing the content of a character's action and emotion information to be assigned to the action for generating three-dimensional motion from a user terminal, and vectorizes the received text prompt and the emotion information, respectively, to generate a text embedding for the content of the action and an emotion embedding that determines the nuance of the action, and generates a motion sequence based on the text embedding using a Generative AI model, but reflects the emotion embedding as a condition of the generation process to assign an emotional context to the character's action, and can generate the motion sequence with the assigned emotional context as final motion data.

The application discloses a kind of capacitive displacement sensor calibration device and method, belong to metrological testing technical field, comprising: host computer is started linear displacement table by controller module and carries out the linear motion required in calibration work, multiple degrees of freedom error and displacement value generated by linear motion are measured simultaneously using optical measurement module, and are transmitted to host computer by data acquisition module, host computer is adjusted by controller module and is located in the multiple degrees of freedom microstage pose of linear displacement table bottom to compensate multiple degrees of freedom error in linear motion, realizes to capacitive displacement sensor calibration.The application effectively breaks through the bottleneck that traditional capacitive displacement sensor calibration technique precision is limited, measurement and compensation function are integrated, eliminate multiple degrees of freedom error coupling influence, can satisfy the calibration demand of high-precision capacitive displacement sensor.

This invention provides a method for online force assessment of characters in a virtual reality environment, aiming to achieve real-time quantitative assessment, physical accuracy, and high robustness of force on humanoid character movements in virtual reality environments. This method combines a motion generation model and a physics optimizer, using an end-to-end integrated approach to dynamically estimate key mechanical parameters such as joint torque and ground reaction force. Specific steps include: acquiring user motion trajectory data from sparse sensors (such as VR headsets and controllers); generating preliminary full-body motion postures using a motion generation model based on discrete latent space learning; performing physical correction and force estimation on the motion using a physics optimization module (employing a dual proportional-differential controller and a motion tracking optimizer); integrating motion data and force parameters; and outputting character animation and force visualization results in real time through a rendering engine. Through the collaborative design of motion generation and physical force assessment, this invention significantly improves the realism, real-time performance, and adaptability of virtual character movements, making it suitable for VR applications such as security monitoring, medical training, and sports simulation, enhancing immersion and interaction efficiency.

This invention discloses a stylized human motion generation method based on spatiotemporal attention, comprising the following steps: using a publicly available human motion style dataset, preprocessing it, and then dividing it into training and testing sets; constructing a stylized motion generation network model based on spatiotemporal attention; training the stylized motion generation network model based on spatiotemporal attention using the training set and updating its parameters; using the trained human motion generation model, taking the original motion content sequence and target style information as input, and outputting a stylized human motion sequence that retains the original action semantics and possesses the target style characteristics. This invention overcomes the problems of style mean-averaging and loss of fine-grained dynamics caused by temporal averaging aggregation in existing technologies, as well as the lack of adaptive attention to key spatiotemporal positions. It achieves fine modeling of multi-scale, non-uniformly distributed style cues, thereby generating human motion sequences with distinct styles, rich details, and well-preserved semantic structure.

The embodiments present a system and method for generating motion of a 3D character using rotation range constraints based on joint-specific biomechanical information based on artificial intelligence. A method according to one embodiment receives a text prompt for generating motion of a 3D character from a user terminal, inputs the text prompt into a text encoder to generate a text condition vector, obtains skeleton configuration parameters including a character type, an active joint set, and a character proportional parameter based on the text condition vector, generates skeleton structure information including joint list information, joint connection information, joint characteristic information, and joint rotation axis information based on the skeleton configuration parameters, inputs the character proportional parameter and the skeleton structure information into a biomechanics encoder to generate a biomechanics context vector including an allowable rotation range per joint, a joint type, and a bone length, inputs the text condition vector and the biomechanics context vector into a diffusion-based motion generation model to generate a rotation sequence per joint, transmits 3D motion data generated by applying the rotation sequence per joint to the user terminal, and performs scale adjustment and shift adjustment on the output feature vector generated in the graph transformer layer of the diffusion-based motion generation model using the biomechanics context vector.

The utility model discloses a kind of walking welding robots, including platform, platform one end is connected with stand, the other end is connected with diagonal brace, first guide mechanism is installed in platform, first sliding table is installed in first guide mechanism, first drive mechanism is connected with first sliding table, second guide mechanism is installed in first sliding table, second drive mechanism is installed in first sliding table, second sliding table is installed in second guide mechanism, and is connected with second drive mechanism, manipulator is installed in second sliding table, and welding torch is connected in manipulator.Platform is extended from the upper side of stand to the side of stand and forms overhanging structure, and positioning of platform is strengthened by diagonal brace, and larger movement space is left for the box to be welded, interference with the movement of box is avoided, and manipulator and welding torch with two degrees of freedom are hoisted by platform, can be welded to box at different angles and directions, compared with artificial welding, greatly improve welding precision and welding efficiency.

The application provides a multi-stage semantic emotion-driven digital human action generation method, and relates to the technical field of digital human action generation. The application extracts audio features and text semantic features and completes time alignment and fusion through multi-modal preprocessing of input audio and text to obtain a multi-modal feature sequence. Then, the sequence is used to identify semantic intent and analyze emotion to obtain a semantic intent sequence and an emotion curve containing categories, intensity and change trend. Subsequently, a global action script containing action type, intensity, duration and priority is generated according to a preset action mapping rule. The script is encoded into a feature vector and input into a lightweight neural network to obtain an action parameter sequence containing facial expressions, mouth shapes, skeletal actions and gaze directions. After time sequence smoothing and physical constraint checking, the final digital human driving parameters are output, which can realize low-power real-time operation, smooth action, accurate mouth shape synchronization and emotion and semantic matching.

This invention discloses an intelligent dance choreography method and system based on generative AI and a cloud-based motion library, relating to the field of artificial intelligence technology. It constructs a cloud-based motion library and combines it with dance music time frames for candidate motion matching. It utilizes dancer spatial relationships to generate dynamic graph adjacency weights and performs path search to complete motion sequence planning. Based on this, it performs spatial conflict correction and keyframe interpolation to generate continuous motions. Furthermore, it introduces spectrogram theory to enhance and optimize the adjacency structure, achieving secondary motion planning and spatial constraint correction. This generates a target dance motion sequence with good formation structure and temporal continuity, solving the problems of disconnect between motion generation and spatial formation, and insufficient expressive ability of dynamic formation structure in traditional dance choreography.

The application provides a multi-modal semantic driven string performance motion generation method and system, the method comprising: obtaining text data and audio data, and respectively preprocessing to obtain text sequence features and audio timing features; encoding the text sequence features and the audio timing features respectively to obtain text embedding representation and audio embedding representation; time aligning and fusing the text embedding representation and the audio embedding representation to obtain unified multi-modal latent representation; extracting performance style features and injecting them into the multi-modal latent representation to obtain style modulated latent representation; decoding and post-processing the style modulated latent representation to generate string performance motion sequence, which not only overcomes the short board of pure audio driving in semantic understanding, but also flexibly adjusts the motion performance according to the user's text instruction and reference style, realizes high expressiveness, high controllability and string performance motion generation conforming to physical laws.

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.

The application provides a controllable dance action generation method based on a standardized generation flow, comprising: constructing a dance action generation model based on a standardized generation flow, and training the dance action generation model using an open-source audio-dance dataset; setting a key frame of a target dance action sequence, generating a control signal from the key frame and target audio, and using the dance action generation model to implicitly reflect an original dance action sequence into original implicit variables in a Gaussian space; and using the control signal to conditionally generate the target dance action sequence from target implicit variables sampled from the Gaussian space by the dance action generation model. The application also provides a controllable dance action generation system based on a standardized generation flow, and a data processing device for realizing controllable dance action generation based on a standardized generation flow.

This invention belongs to the field of artificial intelligence technology and relates to an audio-driven real-time motion generation method and a robot full-body motion control method. The robot full-body motion control includes: extracting music conditional features from the current audio stream data; mapping the target's historical motion sequence to a low-dimensional target motion sequence; using a target diffusion model to predict human motion based on multi-conditional noise, music conditional features, and the target low-dimensional motion sequence, obtaining the human motion velocity term corresponding to the multi-conditional noise; decoding the human motion velocity term under the motion window into real-time human motion; generating the robot's target joint motion based on the real-time human motion; and driving the joint motion with the motor torque generated by the target joint motion based on the target joint motion. This invention achieves real-time closed-loop control of the robot's full-body motion while stably perceiving real-time audio and generating human motion online for extended periods, while simultaneously ensuring motion continuity and rapid response to audio changes.

This invention discloses a robotic arm grasping method and apparatus based on a multi-scale sequence model using imitation learning. The method includes: collecting disordered grasping robotic arm motion trajectory data through a robotic arm hardware platform to establish a dataset; establishing a sequence multi-scale module based on an attention mechanism to extract multi-scale features from the dataset; integrating the extracted multi-scale features based on a teach-and-learn strategy to obtain a final corrected feature vector; establishing a motion generation strategy model based on a conditional autoencoder to obtain prediction results at different scales; and integrating the prediction results at different scales to obtain the final corrected predicted motion. This invention improves the efficiency of robotic arm motion trajectory planning. Furthermore, the addition of the multi-scale module allows the model to better learn information from different dimensions, increasing the model's generalization ability. Therefore, compared with traditional imitation learning algorithms, this method is more flexible and can better adapt to different control tasks.

This invention discloses a continuous motion control method and system for a multi-degree-of-freedom robot based on a streaming policy. The continuous motion control method includes a motion generation step to generate a final motion command, and drives the multi-degree-of-freedom robot to perform the corresponding action based on the final motion command. The motion generation step includes: A1: acquiring the current state observation information, current intermediate motion variables, and current flow time step information of the multi-degree-of-freedom robot; A2: using a velocity network with a sequence model structure to calculate the output velocity field based on the information acquired in A1; A3: acquiring initial noise samples, and iteratively executing steps A1 to A2 based on the initial noise samples to perform a motion update step, generating the final motion command. In each iteration, the velocity field is used to update the current intermediate motion variables. This invention avoids the problems of training instability and policy convergence difficulties caused by ill-conditioned gradients in multi-step flow expansion when training streaming policies using off-policy reinforcement learning.

This binocular device for a robot comprises: a base; a left camera installed on the upper side of the base; a right camera installed on the upper side of the base so as to be spaced apart from the left camera; a motion generation device comprising a main shaft and formed to enable the main shaft to undergo linear movement, rotation, and spiral movement; a left-right tilting mechanism connected to the motion generation device and formed to simultaneously tilt the left camera and the right camera at a predetermined angle in the left direction or the right direction when the main shaft linearly moves in a direction parallel to the base; an up-down tilting mechanism connected to the motion generation device and formed to simultaneously tilt the left camera and the right camera at a predetermined angle in an up direction or a down direction with respect to the base when the main shaft spirally moves; a focus adjustment mechanism connected to the motion generation device and formed to enable adjusting a focus angle between the left camera and the right camera when the main shaft rotates; and a first motor and a second motor for operating the motion generation device.

This invention provides a method and system for automatically generating the motions of a welding robot, which can reduce pre-work of the welding robot and automatically generate the robot's motions on-site. The method for automatically generating the motions of the welding robot (1) in this embodiment includes: a setting step (Step 1), in which a 3D measurement sensor (4) is configured in a predetermined location; a measurement step (Step 2), in which the welding location is divided into multiple measurement areas and measurement is performed using the 3D measurement sensor (4); a point group processing step (Step 3), in which an extraction surface is generated based on the point group data that can be identified as a plane from the measured point group data, and an imaginary surface is generated based on the invalid areas that cannot be identified due to the lack of point group data; a measurement confirmation step (Step 4), in which it is confirmed whether all measurements in the welding location have been completed; and a three-dimensional model generation step (Step 5), in which a three-dimensional model of the welding location is generated based on the extraction surface and the imaginary surface.

The application discloses a motion generation method and device of an electric power robot, electronic equipment and a storage medium, and belongs to the technical field of artificial intelligence. The method comprises the following steps: acquiring instruction data and environment data in the working process of the electric power robot, performing feature fusion to form multi-modal semantic features; inputting the multi-modal semantic features into a motion generation model to generate an initial action sequence; calculating an inference loss function value of the initial action sequence based on an inference loss function containing a hardware constraint penalty term; when the inference loss function value does not exceed a preset constraint penalty threshold, directly determining the initial action sequence as a target action sequence; when the inference loss function value exceeds the threshold, iteratively correcting the initial action sequence until a target action sequence meeting a hardware constraint set is generated; and driving the electric power robot to perform an action according to the target action sequence. Through implementation of the application, the problem of low motion generation efficiency of the electric power robot in the prior art can be solved.

The application provides a large-scale ring structure hoisting deformation monitoring method and system based on panoramic vision and deep learning. The method uses a single panoramic camera to collect video streams synchronously with the structure, and reconstructs a full-circumferential distortion-free window in real time through double-cubic projection technology. To meet the real-time requirements of the hoisting site, a lightweight YOLOv8 model integrating partial convolution and efficient attention mechanism is used to improve the dynamic response speed of the edge computing terminal for target identification. During the hoisting operation, the system uses sub-pixel positioning of the inner and outer targets and uses a difference algorithm to immediately eliminate the rigid body displacement interference caused by the movement of the camera with the structure, thereby directly calculating the real elastic deformation of the structure under the hoisting state. The application realizes an automatic operation mode of "hoisting, monitoring and feedback at the same time", effectively solving the problem of real-time and accurate acquisition of deformation data of large-scale components in a moving base environment.

This utility model relates to a novel robotic arm structure for clamping wafers, comprising: a clamping mechanism portion and a base portion; the clamping mechanism portion is located above the base portion, and the base portion includes a finger base and fingers; the clamping mechanism portion is provided with a first guide rod, the front end of which is provided with a contact roller that can push against the wafer, the middle part of the first guide rod is wrapped in a first bearing, the end of the first guide rod is connected to a connecting seat, the connecting seat is provided with a trigger baffle, and a slotted switch is provided below the trigger baffle; the other end of the connecting seat is connected to a second guide rod, the upper part of which is placed in a second bearing, the second guide rod is provided with a compression spring, the end of the second guide rod is provided with a push seat, the push seat is fitted with a third bearing, and one side of the push seat is connected to an electric push rod. This design reduces wiring difficulty and avoids cable wear; the electric push rod provides power output, avoiding the generation of particles from the piston-like movement of the cylinder; and the clamping force can be adjusted to accommodate wafers of different thicknesses.

The application discloses a robot long-sequence task skill splitting and action generation method based on visual prior and flow matching, comprising the following steps: constructing and training an action self-encoder to extract implicit action features; performing clustering operation on the implicit action features to construct an action category codebook; constructing and training an action category prediction network; training a VLA action generation model based on a contrast flow matching training mechanism; inputting collected multi-view images into the trained action category prediction network to obtain a predicted action category, sampling initial noise from the action category codebook of the action category and inputting the initial noise into the VLA action generation model to generate a continuous action sequence, and driving a robot to execute based on the continuous action sequence. The application can effectively improve the response speed and precision in a long-sequence task, realize accurate mapping of a complex action distribution, effectively inhibit the distribution confusion phenomenon, and reduce the misoperation rate in complex operations.

This invention discloses a semantically aware weakly supervised method for generating 3D motion based on 2D motion data. It utilizes the semantic features of 2D actions to achieve end-to-end optimization of the 3D action generation process under weak supervision. The method comprises three core steps: First, it constructs a shared semantic embedding feature space aligned between text, 2D actions, and 3D actions to achieve consistency in multimodal semantic representation; then, based on the weakly supervised semantic information contained in the 2D action embedding features, it optimizes and enhances the 3D action pseudo-labels to improve the effectiveness of the supervision signal; finally, it provides feedback signals at the embedding feature level to supervise and optimize the parameters of the text-3D action generation model. This invention reduces the model's dependence on text-3D action annotation data, relaxes the restrictions on 2D action data acquisition conditions, and has good scalability and compatibility, adapting to various existing and future text-to-3D action generation technologies.

Little adjustments and changes are made to a vehicle's designs to accommodate for wind turbines installed on the vehicles. The vehicle, the wind turbines and the shields are seamlessly combined, to increase their effective power and bring about the much-needed improvements needed to generate energy from the vehicle's movements. The uniqueness of this invention is in how it makes little adjustments and improvements to the different parts involved and put them together seamlessly to effectively work together and increase their combined power, efficiency, effectiveness etc., to get vehicles to generate energy from their own movements, all while allowing the vehicles to remain aesthetically pleasing. Wherein the vehicle designs, shapes, frame and body itself plays a major role in the process, by dedicating more of the vehicle space and areas for having the wind turbines.

This invention discloses a method and system for keyframe animation completion of complex motion based on neural networks. By providing precise footwork control through different foot contact information, this invention offers a stable and high-quality method for keyframe animation completion of complex movements. It eliminates the slippage problem during motion generation and exhibits strong robustness. This solves the problem in existing technologies that utilize deep learning to perform complex motion generation tasks, which suffer from severe slippage and ghosting issues, leading to significant manpower and resources required for subsequent motion correction.

This application discloses a vehicle control method and a vehicle, relating to the field of vehicle control technology. The method includes: acquiring perception information of the target vehicle while it is in motion, the perception information including first information representing the vehicle's inertial motion state, second information representing the wheel motion state, and third information representing the suspension height state; based on the first, second, and third information, removing interference components generated by the vehicle's own motion from the first information to obtain a gravity direction representation quantity; generating a slope estimate of the road where the target vehicle is located based on the first information and the gravity direction representation quantity; and controlling the target vehicle based on the slope estimate to adapt to the current road slope conditions. This application can improve the accuracy of obtaining road slope.