Robot motion performance evaluation system and method based on optical motion capture

Abstract

The invention relates to the technical field of motion capture technologies, and discloses a robot motion performance evaluation system and method based on optical motion capture, and the system comprises an optical motion capture subsystem which is used for collecting the three-dimensional coordinates of a reflective mark point in a space according to a specified frequency; the rigid body adaptation assembly comprises a plurality of independent structural parts which are detachably fixed to the designated position of the detected robot, and each structural part is loaded with at least three reflective mark points forming the unique geometric topology; the data processing unit is used for mapping the rigid body posture into physical motion parameters of robot joints; and the synchronous calibration module is used for realizing time synchronization of data acquired by the optical motion capture subsystem and instruction data of the robot controller. The rigid body posture can be accurately mapped into joint physical motion parameters, the problem of errors caused by asynchronism of collected data and instruction data in traditional evaluation is solved, and comprehensive and reliable basic data support is provided for robot motion performance evaluation.

Description

Technical Field

[0001] This application relates to the technical field of motion capture technology, and in particular to a robot motion performance evaluation system and method based on optical motion capture. Background Technology

[0002] With the accelerated industrialization of humanoid and multi-jointed robots, the market demands increasingly stringent requirements for motion control accuracy, gait stability, and batch consistency. Current methods for evaluating robot motion performance primarily rely on sensors integrated within the robot, such as encoders and inertial measurement units (IMUs), but these solutions have significant technical limitations: Firstly, the encoder can only provide position information at the motor end, and cannot detect mechanical transmission errors such as gearbox backlash or belt loosening. It also has difficulty capturing structural deformation of the connecting rod after it is subjected to force, resulting in distorted end motion accuracy measurement. Secondly, IMUs are prone to integral drift during long-term operation, making it impossible to achieve accurate measurement of absolute pose. Third, the traditional external sensor installation process is cumbersome and is affected by differences in human operation, resulting in poor consistency in installation location and a lack of comparability of data from different test objects.

[0003] At the same time, existing optical motion capture technology is mainly designed for human skeletal motion. Its algorithms have a certain tolerance for soft tissue deformation, and the measurement accuracy is far from meeting the micron-level evaluation requirements of industrial robots. There is an urgent need for a high-precision, standardized evaluation scheme specifically designed for the rigid structure of robots. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides a robot motion performance evaluation system and method based on optical motion capture.

[0005] In a first aspect, this application provides a robot motion performance evaluation system based on optical motion capture, comprising: An optical motion capture subsystem is used to acquire the three-dimensional coordinates of reflective markers in space at a specified frequency. The rigid body adapter assembly includes multiple independent, detachable structural components that are fixed to a designated position on the robot under test, each of which carries at least three reflective marker points that constitute a unique geometric topology; The data processing unit is used to map the rigid body pose to the physical motion parameters of the robot joints; The synchronization calibration module is used to synchronize the time of data acquired by the optical motion capture subsystem with the command data of the robot controller.

[0006] By adopting the above technical solution and combining it with rigid body adapter components installed on key parts of the robot, non-contact capture of the robot's motion state is achieved, avoiding interference from contact measurement on the robot's motion. The synergistic effect of the data processing unit and the synchronous calibration module can not only accurately map the rigid body pose to the joint physical motion parameters, but also solve the error problem caused by the asynchrony between the collected data and the command data in traditional evaluation. It effectively makes up for the shortcomings of internal sensors in capturing mechanical transmission errors and structural deformation, and provides comprehensive and reliable basic data support for the evaluation of robot motion performance.

[0007] Optionally, it also includes auxiliary positioning fixtures for guiding the standardized installation of the rigid body adapter components on the robot link, having a negative-shaped internal cavity that perfectly matches the geometry of the outer surface of the robot under test.

[0008] By adopting the above technical solution, the installation position of the rigid body adapter component is defined from a physical structure perspective. Standardized installation of the rigid body component can be achieved without relying on the experience and judgment of operators. This completely solves the problem of data incomparability caused by inconsistent installation positions of traditional external sensors. At the same time, it simplifies the installation process, improves installation efficiency, and provides physical assurance for the consistency of batch robot evaluation.

[0009] Optionally, the auxiliary positioning fixture has a magnetic positioning structure and a color code management mechanism; The magnetic positioning structure generates a repulsive force when the structural component is installed in the wrong direction. The structural component in different parts is equipped with a unique color identifier corresponding to the magnetic positioning structure.

[0010] By adopting the above technical solutions, the magnetic positioning structure achieves installation error prevention through polarity matching, which physically eliminates the directional installation errors of rigid body adapter components and reduces the operation error rate; the color code management mechanism achieves rapid sorting and matching of tooling and rigid body components through unique color identification, which is especially suitable for batch robot assembly scenarios, greatly shortens the assembly time, improves the assembly accuracy, and further enhances the consistency and efficiency of batch evaluation.

[0011] Optionally, the data processing unit includes: The rigid body solution module is used to solve the 6-DOF pose of the rigid body based on the three-dimensional coordinates, combined with the rigid body kinematics model and redundant feature constraints. The joint axis calibration module, based on geometric characteristics, adopts a single-axis isolated rotation fitting strategy and solves the true physical rotation center of the joint by least-squares fitting of multiple sets of trajectory point clouds. The performance analysis module is used to quantify and obtain the motion performance indicators of the robot under test.

[0012] By adopting the above technical solutions, the rigid body solution module combines rigid body kinematics models with redundant feature constraints to ensure the accuracy of 6-DOF pose solving and eliminate the influence of abnormal data on measurement results; the single-axis isolated rotation fitting strategy of the joint axis calibration module can accurately locate the real physical rotation center of the joint, solving the pose mapping deviation problem caused by installation errors; the modular design of the performance analysis module realizes the quantitative extraction of motion performance indicators, breaking through the limitations of traditional evaluation indicators that are vague and difficult to quantify, and providing multi-dimensional and accurate judgment basis for robot performance evaluation.

[0013] Optionally, the single-axis isolated rotation fitting strategy includes: S1: Lock all degrees of freedom of the robot except for the joint to be calibrated, and control the joint to be calibrated to perform reciprocating rotational motion within the mechanical limit range; S2: Use the optical motion capture subsystem to record the motion trajectory point cloud of rigid body markers on the link in space; S3: Fit the point cloud using the least squares method; S4: Calculate the fitted center or sphere coordinates as the true physical rotation center of the joint, and calculate the transformation matrix of the origin of the rigid body coordinate system relative to this center, which is used to correct the robot's URDF kinematic model.

[0014] By adopting the above technical solutions, the single-axis isolated rotation fitting strategy ensures the purity of the motion trajectory of the joint to be calibrated by locking other degrees of freedom, and avoids mutual interference of multi-joint motions. The least-squares fitting of multiple sets of trajectory point clouds combined with targeted spatial circle or spatial sphere fitting methods improves the accuracy of joint axis positioning. By generating transformation matrices to correct the URDF kinematic model, the influence of installation errors and model errors on the evaluation results is further eliminated, making the calculation of joint motion parameters more consistent with the actual motion state of the robot, and providing algorithmic support for high-precision evaluation.

[0015] Optionally, the performance analysis module includes: Joint accuracy evaluation unit: Calculates the relative rotation angle between the rigid bodies of the parent link and the child link, compares it with the time-domain feedback value of the robot motor encoder, and outputs the hysteresis error and linearity error; Head stability assessment unit: Analyzes the six-degree-of-freedom pose of the robot's head rigid body during walking, and extracts the high-frequency jitter component and the low-frequency sway envelope area; Gait evaluation unit: Based on the contact state between the rigid body of the foot and the ground and the mass distribution of each link, calculates the relationship between the robot's synthetic center of mass trajectory and the zero moment point in real time; Structural rigidity assessment unit: detects the actual positional deviation of the link end of the robot under load and evaluates the structure's resistance to deformation.

[0016] By adopting the above technical solutions, the joint accuracy evaluation unit accurately quantifies hysteresis error and linearity error through time-domain alignment comparison, filling the gap that traditional sensors cannot detect errors such as mechanical backlash; the head stability evaluation unit achieves quantitative evaluation of head motion stability by separating high-frequency jitter and low-frequency sway components, providing specific directions for robot motion control optimization; the gait evaluation unit comprehensively reflects the stability and anti-tipping ability of the robot's gait based on the relationship analysis between the center of mass trajectory and the zero-moment point; and the structural rigidity evaluation unit intuitively evaluates the deformation resistance of the linkage structure through position deviation detection under load. The four evaluation units work together to cover the core dimensions of robot motion performance, achieving comprehensive and in-depth evaluation.

[0017] Optionally, the operation of the synchronization calibration module includes: Receive the acquisition timestamp from the optical motion capture subsystem and the command issuance timestamp from the robot controller; Time base alignment is achieved through hardware trigger signals or GPS timing synchronization; The data transmission delay and acquisition delay are detected and calculated, and the two data streams are time-domain calibrated using a delay compensation algorithm.

[0018] By adopting the above technical solution, the synchronous calibration module effectively eliminates the delay error in the data transmission and acquisition process through time reference alignment and delay compensation algorithms, and realizes microsecond-level synchronization between optical capture data and command data. This ensures the temporal consistency of data during subsequent error comparison and performance analysis, avoids the distortion of evaluation results caused by data asynchrony, and provides a reliable time reference guarantee for high-precision performance quantification.

[0019] Secondly, this application provides a method for evaluating robot motion performance based on optical motion capture, including: A detachable rigid body adapter assembly is installed at a designated location on the robot under test. Each structural component in the rigid body adapter assembly carries at least three reflective marker points that constitute a unique geometric topology. The three-dimensional coordinates of reflective markers in the space are collected at a specified frequency to achieve time synchronization between the collected data and the robot controller command data. The rigid body pose is mapped to the physical motion parameters of the robot joints, and the motion performance is evaluated based on the physical motion parameters.

[0020] By adopting the above technical solution, the evaluation method strictly matches the hardware composition and core functions of the system. Through standardized rigid body component installation, precise data acquisition and synchronization, and scientific parameter mapping and performance evaluation process, the operability and repeatability of the technical solution are ensured. It is suitable for the fine-grained verification of robot motion performance in the R&D stage, and can also meet the efficiency requirements of batch evaluation in the mass production stage, realizing the standardization and normalization of robot motion performance evaluation in different application scenarios.

[0021] Optionally, a gold sample transfer strategy may also be included: Establish benchmark: Select a standard preset robot as the gold sample for fine calibration of all degrees of freedom, and establish a standard preset rigid body topology template that includes the relative positional relationship of all joint axes; Tooling-assisted installation: Using auxiliary positioning tooling, the rigid body adapter component is installed onto the mass-production robot to be tested; Static deviation calculation: Control the mass production robot under test to execute a single standard static posture, collect the current rigid body coordinates and perform spatial registration with the standard preset rigid body topology template; Automatic compensation and judgment: Calculate the spatial deviation between the current rigid body coordinates and the standard preset rigid body topology template; if the deviation is less than the preset threshold, automatically generate a compensation matrix to correct the measurement model without dynamic motion calibration; if the deviation is greater than the threshold, trigger an assembly error alarm.

[0022] By adopting the above technical solutions, a unified benchmark is provided for batch robot evaluation, avoiding the tedious process of performing complex dynamic calibration on each robot and greatly improving the efficiency of batch evaluation. The static deviation calculation and automatic compensation mechanism not only solves the problem of evaluation deviation caused by minor differences in robot structure during batch production, but also triggers assembly error alarms through threshold judgment to promptly detect installation or structural abnormalities, taking into account the efficiency, accuracy and fault tolerance of batch evaluation, and providing an efficient and reliable solution for robot mass production quality inspection.

[0023] In summary, this application includes at least one of the following beneficial technical effects: By deeply integrating optical motion capture technology with rigid body kinematics and synchronous calibration technology, a fully closed-loop, high-precision robot motion performance evaluation solution has been constructed. This solution effectively solves the problems of traditional evaluation methods, such as missing mechanical error detection, poor installation consistency, data asynchrony, and low efficiency of batch testing, which rely on internal sensors. The combination of the optical motion capture subsystem and the rigid body adapter component can directly acquire the real motion state of the robot links, accurately capture mechanical transmission errors and structural deformation, and further improve the measurement reliability in complex environments by using asymmetric high-reflectivity reflective markers. Attached Figure Description

[0024] Figure 1 This is a schematic diagram illustrating the modular composition principle of the robot motion performance evaluation system in this application; Figure 2 This is a schematic diagram of a practical scenario for the robot motion performance evaluation system based on optical motion capture in this application; Figure 3 This is an exploded view of the rigid body adapter component in this application. Detailed Implementation

[0025] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0026] In the description of this specification, the references to "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples" refer to specific features, structures, materials, or characteristics described in connection with the described embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0027] This application discloses a robot motion performance evaluation system based on optical motion capture, referring to... Figure 1 and Figure 2 As shown, it includes: The optical motion capture subsystem uses an array of eight high-speed infrared cameras with a resolution of 2048×2048 pixels and an infrared emission wavelength of 850nm. The acquisition frequency is locked at 120Hz through system integration software (which can be adjusted to 240Hz according to the evaluation accuracy requirements). During deployment, the cameras are arranged within a 3-5 meter range around the robot under test to form a full-view, unobstructed acquisition area, ensuring that the reflective markers are captured by at least three cameras simultaneously during the robot's movement. The acquired three-dimensional coordinate data is transmitted in real time via gigabit Ethernet to the data processing unit for acquiring the three-dimensional coordinates of the reflective markers in space according to the specified frequency. The rigid body adapter component consists of multiple custom-designed independent structural parts tailored to the structural dimensions of the robot under test. Each component is injection-molded from high-strength ABS engineering plastic, with a weight controlled between 100-300g to avoid placing additional load on the robot's movement. These components are secured to key locations on the robot's head, torso, shoulder joint links, elbow joint links, hip joint links, knee joint links, and ankle joint links using detachable bolts or clips. Each component has three reflective markers fixed to it via threaded connections. These markers are arranged in an asymmetrical triangular pattern (ensuring geometric topological uniqueness; an anti-reflective coating can also be applied to the surface of the markers). The spacing between the markers is 50-80mm to meet the accuracy requirements of optical capture. The data processing unit uses an industrial computer equipped with an Intel Core i9 processor, 32GB of memory and an NVIDIA RTX 4090 graphics card. It comes pre-installed with dedicated data processing software developed based on C++. The software integrates core algorithms for rigid body calculation, joint axis calibration and performance analysis. It establishes data communication with the optical motion capture subsystem and robot controller through the PCIe interface to realize the real-time reception, processing and calculation output of three-dimensional coordinate data and motion parameters. The synchronization calibration module uses an independent synchronization control board, which is connected to the trigger terminal of the optical motion capture subsystem through the GPIO interface and communicates with the robot controller through the RS485 interface. It receives the timestamp information of the two data sources in real time. The board has a built-in GPS timing module (supporting Beidou dual-mode timing) to ensure the uniformity of the time reference. The synchronization control board and the data processing unit transmit the calibrated synchronization data through the USB3.0 interface.

[0028] During the installation of the rigid body adapter component, an auxiliary positioning fixture can be used to guide the standardized installation of the rigid body adapter component on the robot link. In this embodiment, the auxiliary positioning fixture, in addition to having a negative-shaped inner cavity that perfectly matches the geometry of the outer surface of the robot under test, also has a magnetic positioning structure and a color code management mechanism.

[0029] Specifically, the auxiliary positioning fixture is designed for the linkage structure of a specific robot model. It is manufactured using photosensitive resin 3D printing technology, with manufacturing data derived from the 3D CAD model of the robot's linkage, ensuring a perfect fit between the fixture's negative-shaped inner cavity and the robot's outer surface. The fixture has a semi-enclosed structure, with an inner cavity depth of 1 / 3 of the linkage diameter. A 0.5mm thick sponge buffer layer is attached to the inner wall to prevent scratching the robot's surface during installation. The outer side of the fixture features a strip-shaped guide groove that matches the shape of the rigid body adapter component. The guide groove's length is 1.2 times the length of the rigid body adapter component, and its width is 0.2mm wider than the rigid body component, ensuring smooth and secure sliding. During installation, the operator first fastens the fixture onto the robot linkage. Due to the unique match between the negative-shaped inner cavity and the linkage, the fixture is fixed in a single position without additional positioning. Then, the rigid body adapter component is slid into the guide groove to the preset limit. After securing, the fixture is removed, achieving standardized installation of the rigid body adapter component.

[0030] The magnetic positioning structure of the auxiliary positioning fixture uses neodymium iron boron magnets. One magnet with a diameter of 8mm and a thickness of 3mm is embedded at each end of the fixture's guide groove. Magnets with opposite polarities are embedded at the ends of the corresponding rigid body adapter components. The magnets are embedded to a depth of 2mm, with their surfaces flush with the structural components to avoid affecting the installation fit. When the rigid body component is installed in the correct orientation, the opposite poles of the magnets attract each other, generating an attraction force of 0.5-1N, assisting in the positioning and locking of the rigid body component. When the installation orientation is incorrect, the like poles of the magnets repel each other, generating a repulsive force greater than 2N, preventing the operator from pushing the rigid body component into the guide groove, thus physically eliminating installation errors. The color-coded management mechanism is implemented by spraying wear-resistant paint onto the outer surface of the fixture and the sides of the rigid body components. The fixture and rigid body component corresponding to the left limb are painted red, the right limb green, and the fixture and rigid body component corresponding to the head yellow. The painted area is a 20×20mm square mark, facilitating quick sorting and matching by the operator, especially suitable for batch robot assembly scenarios, improving assembly efficiency.

[0031] Optionally, such as Figure 3 As shown, the rigid body adapter component includes: The rigid body adapter component features a split-type bidirectional V-shaped slot with a 120-degree V-angle design. This angle, optimized through mechanical simulation, enables self-centering clamping of cylindrical, square (side length 20-150mm), or polygonal (inscribed circle diameter 20-150mm) connecting rods within a diameter range of 20mm-150mm, ensuring stable fixation of connecting rods with different cross-sectional types. The flexible buffer anti-slip layer inside the V-shaped slot is injection molded from TPU95A material with a thickness of 3mm. The surface has evenly distributed anti-slip textures (texture depth 0.5mm, spacing 2mm), which not only prevents crushing of the robot's surface paint during clamping but also increases friction with the connecting rods, preventing relative slippage. The quick-release ratchet locking mechanism consists of a ratchet, pawl, and elastic steel band. The steel band is 15mm wide and 1mm thick. The ratchet is made of plastic. Locking and unlocking are controlled by a press-type pawl. When locked, it can provide a constant clamping force of 5-10N. This force has been tested and can effectively eliminate the relative micro-movement between the rigid body and the connecting rod without damaging the connecting rod structure. When unlocking, simply press the pawl to quickly remove the component, making it easy to operate.

[0032] Optionally, the data processing unit includes: The rigid body solution module establishes a rigid body kinematic model based on a homogeneous transformation matrix. It takes the three-dimensional coordinates of the optically captured reflective markers as input and combines redundant feature constraints (each rigid body has at least 3 markers, and abnormal data is eliminated through redundancy verification of the coordinates of multiple markers). It uses the iterative least squares method to solve the 6-DOF pose of the rigid body (X, Y, Z axis translation and rotation angles around the three axes). The solution frequency is kept consistent with the optical capture frequency (120Hz) to ensure the real-time performance of the pose data. The joint axis calibration module integrates a single-axis isolated rotation fitting strategy, which automatically switches the fitting mode for different types of robot joints (single-axis joints, ball joints). The algorithm uses CPU core threads during runtime, and accelerates the fitting calculation through GPU to ensure that the calibration process is completed efficiently. The calibration results are stored in the system database in real time for subsequent URDF model correction. The performance analysis module is used to quantify and obtain the motion performance indicators of the robot under test. It adopts a modular design, with each evaluation unit operating independently and data communicating with each other. By calling the hardware computing resources of the data processing unit, it synchronously analyzes the solved rigid body pose data, robot controller command data and encoder feedback data, and outputs multi-dimensional motion performance indicators. The indicator results can be displayed in real time through the software interface or exported as an Excel file.

[0033] Specifically, the single-axis isolated rotation fitting strategy includes: 1. By issuing commands through the robot controller, lock all joint degrees of freedom except the joint to be calibrated (for example, when calibrating the left elbow joint, lock the shoulder joint, wrist joint and other limb joints), and control the joint to be calibrated to perform reciprocating rotational motion within its mechanical limit range (such as -90° to +90°). The movement speed is gradually increased from 5° / s to 30° / s, and 3-5 reciprocating movements are completed to ensure that the full range of motion of the joint is covered.

[0034] S2: During joint movement, the optical motion capture subsystem continuously collects the three-dimensional coordinates of rigid body markers on the link. Each set of reciprocating motions collects no less than 1,000 trajectory points to form high-density point cloud data. The point cloud data is accompanied by a collection timestamp to ensure accurate correspondence with the joint movement state.

[0035] S3: Call the least squares fitting algorithm in the data processing unit. For single-axis joints (such as elbow and knee joints), project the point cloud data onto a plane perpendicular to the joint rotation direction and fit a spatial circle. For ball joints (such as hip and shoulder joints), directly fit the three-dimensional point cloud to a spatial sphere. During the fitting process, outliers are removed through residual analysis (points with residuals greater than 3 times the standard deviation are considered outliers) to improve fitting accuracy.

[0036] S4: Calculate the center of the spatial circle or the center of the spatial sphere based on the fitting results. This coordinate is the true physical rotation center of the joint. Establish a rigid body coordinate system with the geometric center of the rigid body component as the origin. Calculate the translation vector T_offset and the rotation transformation matrix R_align of the origin relative to the joint rotation center. Combine the two to generate the calibration matrix M_calib=[R_align|T_offset]. Write the calibration matrix into the robot's URDF kinematic model through the software interface to complete the model correction.

[0037] The performance analysis module includes: Joint accuracy evaluation unit: Extracts the relative rotation angle θ_mocap between the rigid bodies of the parent and child links calculated by the data processing unit, synchronously acquires the robot motor encoder feedback angle θ_enc and the controller command angle θ_cmd, performs time-domain alignment of the three data based on the timestamp after synchronous calibration (using linear interpolation to fill the data sampling interval difference), calculates the difference between θ_mocap and θ_cmd to obtain the hysteresis error, analyzes the linear correlation between θ_mocap and θ_enc through linear regression, and outputs the linearity error; Head stability assessment unit: Real-time acquisition of six-degree-of-freedom pose data of the rigid body of the head during robot walking. The Kalman filter algorithm is used to separate the high-frequency jitter component (frequency greater than 10Hz) and low-frequency sway component (frequency less than 1Hz) in the pose data. The amplitude range of the high-frequency jitter is calculated, and the envelope area of ​​the low-frequency sway in the XY plane is calculated by numerical integration to quantitatively evaluate the head motion stability. Gait evaluation unit: Based on the change of the Z-axis coordinate of the foot rigid body, the contact state between the foot and the ground is determined (Z-axis coordinate less than 5mm is considered ground contact, greater than 5mm is considered off-ground). The mass parameters of each link are input (pre-measured by weighing and stored in the system). The three-dimensional trajectory of the robot's synthetic center of mass (CoM) is calculated in real time using the composite rigid body method. At the same time, the zero moment point (ZMP) is calculated by combining the ground support polygon. The gait stability is evaluated by analyzing the relative positional relationship between the CoM trajectory and the ZMP (such as the projection deviation of the CoM above the ZMP). Structural rigidity assessment unit: This unit detects the actual positional deviation of the robot's link ends under load and evaluates the structure's resistance to deformation. A standard load (the load weight is determined according to the robot's rated load, such as 5kg or 10kg) is suspended at the robot's link ends (such as the end of an arm or leg). The three-dimensional coordinates of the link ends under load are collected by the optical motion capture subsystem and compared with the coordinates under no-load conditions. The positional deviations in the X, Y, and Z directions are calculated. The magnitude of the deviation directly reflects the deformation resistance of the link structure, thus completing the structural rigidity assessment.

[0038] Optionally, the operation of the synchronization calibration module includes: Receive the acquisition timestamp from the optical motion capture subsystem and the command issuance timestamp from the robot controller; Time base alignment is achieved through hardware trigger signals or GPS timing synchronization; The data transmission delay and acquisition delay are detected and calculated, and the two data streams are time-domain calibrated using a delay compensation algorithm.

[0039] Secondly, this application provides a method for evaluating robot motion performance based on optical motion capture, including: The first step is to install the rigid body adapter components: First, clean the mounting surfaces of the robot's head, torso, and joint links (remove dust and oil). Select the corresponding rigid body structural components according to the robot's structure. Use bolts or clips to detachably fix the structural components to the preset positions, ensuring that each structural component is installed firmly without shaking. After installation, use auxiliary positioning fixtures to verify the accuracy of the installation position.

[0040] Step 2, Data Acquisition and Synchronization: Start the optical motion capture subsystem, set the acquisition frequency to 120 Hz, and start the synchronization calibration module after the system stabilizes to complete the time synchronization of the optical capture data and the robot controller command data; control the tested robot to execute preset motion tasks (such as static standing, walking, joint flexion and extension, etc.), and the optical capture subsystem continuously collects the three-dimensional coordinates of the reflective marker points and synchronously transmits them to the data processing unit.

[0041] Step 3, Motion Performance Evaluation: The data processing unit solves the 6-degree-of-freedom pose of each rigid body through the rigid body calculation module, and maps it to the physical motion parameters of the robot joints (such as joint angles, motion speeds, etc.) after correction by the joint axis calibration module. Based on these parameters, the performance analysis module conducts a quantitative evaluation from four dimensions: joint accuracy, head stability, gait performance, and structural rigidity, and finally outputs an evaluation report, including the specific values of various performance indicators and the qualified judgment results.

[0042] In terms of the production line, it also includes the golden sample transfer strategy: Establish a benchmark: Select a standard preset robot that has passed the factory inspection and has qualified motion performance as the golden sample, complete the fine calibration of all degrees of freedom according to the single-axis isolation rotation fitting strategy, collect data such as the relative position relationship of the joint axes of the golden sample robot and the installation coordinates of the rigid body structural parts, and establish a standard rigid body topology template. The template is stored in JSON format and contains key parameters such as marker point coordinates, joint axis coordinates, and rigid body relative positions.

[0043] Tooling-assisted installation: For the mass-produced robots to be tested, use the auxiliary positioning tooling to quickly install the rigid body adapter components to the corresponding link positions. During the installation process, ensure the correct installation direction and accurate position through the magnetic adsorption positioning structure and color code management mechanism, without manual repeated adjustment.

[0044] Static deviation calculation: Control the mass-produced robot to be measured to execute a single standard static posture (such as the T-pose, that is, both arms are naturally extended horizontally, both legs are straight, and the body is straight), and the optical motion capture subsystem collects the three-dimensional coordinates of the marker points on the current rigid body structural parts. The data processing unit performs spatial registration of this coordinate data with the standard rigid body topology template (realize registration using the ICP algorithm) and calculates the spatial deviation amount ΔT between the two.

[0045] Automatic compensation and judgment: First, determine the preset threshold. Select 10 golden sample robots for multiple calibrations (the interval between each calibration is 24 hours, and a total of 10 calibrations are performed), calculate the deviation mean μ and standard deviation σ of the 10 calibration results, and substitute them into the formula Threshold = μ + 2σ to obtain the preset threshold; compare the spatial deviation amount ΔT of the robot to be measured with the preset threshold: If ||ΔT|| < Threshold, the system automatically generates a compensation matrix through the formula The measurement model is corrected, eliminating the need for complex dynamic motion calibration of the mass-produced robot; in, The standard rigid body pose data (or its corresponding spatial transformation matrix) obtained by optical motion capture and joint calibration of the gold sample robot under standard static posture is the "benchmark reference data" for batch evaluation. The rigid body pose transformation matrix / coordinate data corrected for the robot under test: It is the rigid body pose data (or transformation matrix) corresponding to the usable measurement model obtained after the robot under test has undergone "spatial deviation calculation + compensation matrix correction" - simply put, it is the effective data after aligning the rigid body pose of the robot under test to the gold sample reference through the compensation algorithm, which will be used for motion performance evaluation in the future.

[0046] If ||ΔT||>Threshold, the system will trigger an audible and visual alarm and simultaneously display an assembly error message on the software interface, informing the operator to check whether the auxiliary positioning fixture is damaged or whether the robot linkage structure is deformed.

[0047] This invention deeply integrates optical motion capture technology with rigid body kinematics and synchronous calibration technology to construct a fully closed-loop, high-precision robot motion performance evaluation scheme. This effectively solves the problems of traditional evaluation methods, such as missing mechanical error detection, poor installation consistency, data asynchrony, and low batch testing efficiency due to reliance on internal sensors. The combination of the optical motion capture subsystem and the rigid body adapter component can directly acquire the real motion state of the robot links, accurately capturing mechanical transmission errors and structural deformations. Combined with asymmetric high-reflectivity reflective markers, it further improves measurement reliability in complex environments. The negative-shape design, magnetic anti-foolproof structure, and color-code management mechanism of the auxiliary positioning fixture ensure the standardization and consistency of rigid body component installation from a physical perspective, significantly reducing human error. The synchronous calibration module achieves microsecond-level synchronization of measurement data and command data through time base alignment and delay compensation, providing a precise data foundation for error comparison and analysis. The core algorithm and gold sample transfer strategy of the data processing unit not only meet the high-precision verification requirements of robot motion performance during the R&D stage but also simplify the evaluation process in the mass production stage through static registration and automatic compensation, significantly improving batch quality inspection efficiency. The entire system takes into account measurement accuracy, ease of operation, and applicability, and can be adapted to different types and specifications of robots. It provides reliable support for the quantitative evaluation of robot motion performance and helps robot R&D iteration and industrialization.

[0048] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A robot motion performance evaluation system based on optical motion capture, characterized in that, include: An optical motion capture subsystem is used to acquire the three-dimensional coordinates of reflective markers in space at a specified frequency. The rigid body adapter assembly includes multiple independent, detachable structural components that are fixed to a designated position on the robot under test, each of which carries at least three reflective marker points that constitute a unique geometric topology; The data processing unit is used to map the rigid body pose to the physical motion parameters of the robot joints; The synchronization calibration module is used to synchronize the time of data acquired by the optical motion capture subsystem with the command data of the robot controller.

2. The robot motion performance evaluation system based on optical motion capture according to claim 1, characterized in that, It also includes auxiliary positioning fixtures for guiding the standardized installation of rigid body adapter components on the robot link, which have negative-shaped cavities that perfectly match the geometry of the outer surface of the robot under test.

3. The robot motion performance evaluation system based on optical motion capture according to claim 2, characterized in that, The auxiliary positioning fixture has a magnetic positioning structure and a color code management mechanism; The magnetic positioning structure generates a repulsive force when the structural component is installed in the wrong direction. The structural component in different parts is equipped with a unique color identifier corresponding to the magnetic positioning structure.

4. The robot motion performance evaluation system based on optical motion capture according to claim 1, characterized in that, The data processing unit includes: The rigid body solution module is used to solve the 6-DOF pose of the rigid body based on the three-dimensional coordinates, combined with the rigid body kinematics model and redundant feature constraints. The joint axis calibration module, based on geometric characteristics, adopts a single-axis isolated rotation fitting strategy and solves the true physical rotation center of the joint by least-squares fitting of multiple sets of trajectory point clouds. The performance analysis module is used to quantify and obtain the motion performance indicators of the robot under test.

5. The robot motion performance evaluation system based on optical motion capture according to claim 4, characterized in that, The single-axis isolated rotation fitting strategy includes: S1: Lock all degrees of freedom of the robot except for the joint to be calibrated, and control the joint to be calibrated to perform reciprocating rotational motion within the mechanical limit range; S2: Use the optical motion capture subsystem to record the motion trajectory point cloud of rigid body markers on the link in space; S3: Fit the point cloud using the least squares method; S4: Calculate the fitted center or sphere coordinates as the true physical rotation center of the joint, and calculate the transformation matrix of the origin of the rigid body coordinate system relative to this center, which is used to correct the robot's URDF kinematic model.

6. The robot motion performance evaluation system based on optical motion capture according to claim 4, characterized in that, The performance analysis module includes: Joint accuracy evaluation unit: Calculates the relative rotation angle between the rigid bodies of the parent link and the child link, compares it with the time-domain feedback value of the robot motor encoder, and outputs the hysteresis error and linearity error; Head stability assessment unit: Analyzes the six-degree-of-freedom pose of the robot's head rigid body during walking, and extracts the high-frequency jitter component and the low-frequency sway envelope area; Gait evaluation unit: Based on the contact state between the rigid body of the foot and the ground and the mass distribution of each link, calculates the relationship between the robot's synthetic center of mass trajectory and the zero moment point in real time; Structural rigidity assessment unit: detects the actual positional deviation of the link end of the robot under load and evaluates the structure's resistance to deformation.

7. The robot motion performance evaluation system based on optical motion capture according to claim 1, characterized in that, The operation process of the synchronous calibration module includes: Receive the acquisition timestamp from the optical motion capture subsystem and the command issuance timestamp from the robot controller; Time base alignment is achieved through hardware trigger signals or GPS timing synchronization; The data transmission delay and acquisition delay are detected and calculated, and the two data streams are time-domain calibrated using a delay compensation algorithm.

8. A method for evaluating robot motion performance based on optical motion capture, applied to the robot motion performance evaluation system based on optical motion capture as described in any one of claims 1-7, characterized in that, include: A detachable rigid body adapter assembly is installed at a designated location on the robot under test. Each structural component in the rigid body adapter assembly carries at least three reflective marker points that constitute a unique geometric topology. The three-dimensional coordinates of reflective markers in the space are collected at a specified frequency to achieve time synchronization between the collected data and the robot controller command data. The rigid body pose is mapped to the physical motion parameters of the robot joints, and the motion performance is evaluated based on the physical motion parameters.

9. The method for evaluating robot motion performance based on optical motion capture according to claim 8, characterized in that, This also includes the gold sample transfer strategy: Establish benchmark: Select a standard preset robot as the gold sample for fine calibration of all degrees of freedom, and establish a standard preset rigid body topology template that includes the relative positional relationship of all joint axes; Tooling-assisted installation: Using auxiliary positioning tooling, the rigid body adapter component is installed onto the mass-production robot to be tested; Static deviation calculation: Control the mass production robot under test to execute a single standard static posture, collect the current rigid body coordinates and perform spatial registration with the standard preset rigid body topology template; Automatic compensation and decision: Calculates the spatial deviation between the current rigid body coordinates and the standard preset rigid body topology template; If the deviation is less than the preset threshold, a compensation matrix will be automatically generated to correct the measurement model, without the need for dynamic motion calibration. If the deviation exceeds the threshold, an assembly error alarm will be triggered.

Images