Intelligent analysis method and system for operation data of construction equipment

Abstract

The application provides a kind of construction equipment operation data intelligent analysis method and system, it is related to data processing technical field, the method includes: extracting historical positioning data in space calibration interval, according to the position offset vector set between the trajectory point on predicted trajectory and actual trajectory point, with the weighted square sum of position offset vector as optimization goal, to obtain three-dimensional space transformation parameter optimization model;According to optimization model, solve optimization model using quasi-newton optimization algorithm, generate final calibration parameter set;Based on final calibration parameter set, correct predicted motion trajectory, analyze spatial relationship between corrected trajectory and space risk field, calculate collision risk probability and accident severity level, generate comprehensive risk score, and trigger response action according to preset threshold value.The application improves construction safety level and management efficiency.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for intelligent analysis of building equipment operation data. Background Technology

[0002] With the intelligent transformation of the construction industry, the efficient operation and safe management of construction equipment have become crucial for improving construction efficiency and ensuring personnel safety. Traditional equipment monitoring mainly relies on sensors to collect data in real time, but it has significant shortcomings in risk prediction and precise control in dynamic environments. For example, existing technologies often determine equipment status based on single sensor data or simple thresholds, which is insufficient to meet the needs of multi-source data fusion and dynamic risk assessment in complex construction scenarios.

[0003] Furthermore, the prediction of the motion trajectory of building equipment and the modeling of environmental risks often rely on independent physical models or data-driven methods, resulting in insufficient prediction accuracy or poor adaptability. Some traditional methods have not achieved closed-loop control of risk assessment and trajectory correction, which may make it difficult to dynamically adjust the equipment operation strategy to reduce risks. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for intelligent analysis of building equipment operation data, which improves construction safety and management efficiency.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] In a first aspect, a method for intelligent analysis of building equipment operation data, the method comprising:

[0007] Step 1: Acquire device status data and environmental perception data;

[0008] Step 2: Based on the device status data and environmental perception data, obtain the predicted motion trajectory of the device;

[0009] Step 3: Integrate the environmental perception data with the preset building information model and real-time obstacle scanning data to generate a spatial risk field;

[0010] Step 4: Based on the spatial risk field and the predicted motion trajectory, select two spatially fixed calibration reference points within the target analysis domain, and generate a reference direction axis according to the spatial relationship between the two calibration reference points; divide continuous spatial calibration intervals along the reference direction axis; extract historical positioning data within the spatial calibration intervals; calculate the set of position offset vectors between the predicted trajectory points and the actual trajectory points based on the trajectory points on the predicted motion trajectory; use minimizing the weighted sum of squares of the position offset vectors as the optimization objective to obtain a three-dimensional spatial transformation parameter optimization model; based on the optimization model, use a quasi-Newton optimization algorithm to solve the optimization model and generate the final calibration parameter set.

[0011] Step 5: Correct the predicted motion trajectory based on the final calibration parameter set, analyze the spatial relationship between the corrected trajectory and the spatial risk field, calculate the collision risk probability and accident severity level, generate a comprehensive risk score, and trigger response actions according to preset thresholds.

[0012] Further, in step 1, acquire device status data and environmental perception data, including:

[0013] Step 11: Acquire the original voltage signal through the strain gauge array, and perform analog-to-digital conversion on the original voltage signal to generate discrete stress data; input the discrete stress data and the joint rotation pulse number acquired by the encoder into the forward kinematics process to generate real-time spatial pose coordinates; process the real-time spatial pose coordinates, hydraulic pressure sensor readings, motor current sampling values ​​and control operation commands to generate a structured equipment state data matrix.

[0014] Step 12: Acquire raw point cloud data in polar coordinate system using LiDAR, and perform coordinate transformation on the raw point cloud data to generate a three-dimensional point set in global coordinate system; based on the three-dimensional point set, filter out noise points based on the spatial distribution characteristics of adjacent frames to generate denoised point cloud data; acquire Doppler frequency shift signals using millimeter-wave radar, and perform moving target detection calculation on the frequency shift signals to generate obstacle radial velocity vectors; fuse the denoised point cloud data, obstacle radial velocity vectors, and RGB-D depth maps acquired by binocular vision sensors to generate an environmental perception data package;

[0015] Step 13 involves aligning the structured device status data matrix with the environmental awareness data packet using time synchronization. This is done by interpolating based on the hardware clock synchronization signal to generate time-synchronized device status data frames and environmental awareness data frames.

[0016] Further, step 2, based on device status data and environmental perception data, obtains the predicted motion trajectory of the device, including:

[0017] Step 21: Receive the time synchronization device status data frame and extract the real-time spatial pose coordinates and control operation command codes from it; input the real-time spatial pose coordinates into the device dynamic equation library and generate an initial motion trajectory segment based on the rigid body kinematics principle; input the initial motion trajectory segment and control operation command codes into the motion constraint filter and combine them with the device mechanical structure parameters to generate a physically compliant trajectory segment.

[0018] Step 22: Input the physically compliant trajectory segment into the pre-trained trajectory feature extractor to decompose it into a three-dimensional feature vector representing the trajectory characteristics; match the three-dimensional feature vector with the actual trajectory pattern library under similar working conditions to generate a trajectory offset set; input the trajectory offset set into the adaptive weighting unit, and dynamically allocate weights to generate a comprehensive correction factor based on the environmental features contained in the time-synchronized environmental perception data frame.

[0019] Step 23: Input the physically compliant trajectory segment and the comprehensive correction factor into the trajectory synthesizer, perform frame-by-frame pose correction through time axis alignment to obtain the corrected trajectory segment; perform smoothness optimization processing on the corrected trajectory segment to eliminate trajectory jitter, and generate an optimized continuous motion trajectory as the predicted motion trajectory of the device.

[0020] Further, in step 3, the environmental perception data is fused with the preset building information model and real-time obstacle scanning data to generate a spatial risk field, including:

[0021] Step 31: Receive time-synchronized environmental awareness data frames and extract denoised point cloud data from them; input the denoised point cloud data into the spatial registration process, align it with the coordinate system of the preset building information model, and generate a registration point cloud; input the registration point cloud into the voxel processor, divide it into three-dimensional mesh units of equal size, and mark the static obstacle occupancy area.

[0022] Step 32: Extract the radial velocity vector and RGB-D depth map of the obstacle from the time-synchronized environmental perception data frame; generate the future time series location set of the dynamic obstacle from the radial velocity vector, RGB-D depth map and denoised point cloud data; calculate the spatiotemporal distribution probability of the dynamic obstacle in the three-dimensional grid cell based on the future time series location set, and generate the spatiotemporal probability distribution of the dynamic obstacle.

[0023] Step 33: Generate an initial risk field based on the static obstacle occupancy area markings and the spatiotemporal probability distribution of dynamic obstacles; normalize the initial risk field to generate a spatial risk field.

[0024] Further, in step 4, based on the spatial risk field and the predicted motion trajectory, two spatially fixed calibration reference points are selected within the target analysis domain, and a reference direction axis is generated according to the spatial relationship between the two calibration reference points; continuous spatial calibration intervals are divided along the reference direction axis; historical positioning data within the spatial calibration intervals are extracted, and the set of position offset vectors between the predicted trajectory points and the actual trajectory points is calculated based on the trajectory points on the predicted motion trajectory; the optimization objective is to minimize the weighted sum of squares of the position offset vectors to obtain a three-dimensional spatial transformation parameter optimization model; based on the optimization model, a quasi-Newton optimization algorithm is used to solve the optimization model to generate the final calibration parameter set, including:

[0025] Step 41: Based on the three-dimensional spatial distribution characteristics of the spatial risk field, select two calibration reference points with fixed spatial positions within the target analysis domain; generate a reference direction axis based on the spatial coordinates of the two calibration reference points and calculate the direction cosine; divide the reference direction axis into equal intervals and generate a continuous spatial calibration interval with the division points as boundaries.

[0026] Step 42: Extract historical positioning data of equipment operation within each spatial calibration interval to obtain the actual trajectory point set; extract the predicted trajectory point set that matches the timestamp of the historical positioning data based on the predicted motion trajectory; calculate the spatial offset between the predicted trajectory point and the actual trajectory point point point by point to generate a position offset vector set.

[0027] Step 43: Normalize and weight the position offset vector set to generate a weighted offset vector set; take minimizing the sum of squares of the weighted offset vectors as the objective function, and combine the constraints of the three-dimensional space transformation parameters to obtain the three-dimensional space transformation parameter optimization model.

[0028] Step 44: Input the objective function into the optimization solver, initialize the parameter estimates of the rotation matrix, translation vector, and scaling factor; iteratively execute the quasi-Newton optimization algorithm, calculate the gradient vector of the objective function under the parameters, update the approximate value of the Hessian inverse matrix, and perform adaptive parameter updates along the gradient in the opposite direction; terminate the iteration when the change in the objective function is lower than the convergence threshold, and output the final calibration parameter set composed of the rotation matrix, translation vector, and scaling factor; terminate the iteration when the change in the objective function is lower than the convergence threshold to obtain the final calibration parameter set.

[0029] Further, in step 5, the predicted trajectory is corrected based on the final calibration parameter set. The corrected trajectory is then spatially analyzed in relation to the spatial risk field to calculate the collision risk probability and accident severity level, generating a comprehensive risk score. A response action is then triggered according to a preset threshold, including:

[0030] Step 51: Extract the rotation matrix, translation vector, and scaling factor parameters from the final calibration parameter set; input the trajectory point set of the predicted motion trajectory into the space transformer, apply the parameters to perform coordinate transformation calculation, and generate the preliminary corrected trajectory; perform time-series filtering on the preliminary corrected trajectory to eliminate high-frequency jitter noise and generate the final corrected motion trajectory.

[0031] Step 52: Map the spatial position sequence of the final corrected motion trajectory to the grid coordinate system of the spatial risk field to generate the mapping relationship between trajectory points and grid cells; extract the real-time risk value of the grid cell where each trajectory point is located based on the mapping relationship.

[0032] Step 53: Calculate the dynamic collision probability value based on the real-time risk value and the geometric distance from the trajectory point to the nearest obstacle; perform an integral operation on the dynamic collision probability value along the trajectory time window to generate the overall collision risk probability value;

[0033] Step 54: Extract the predefined structural importance coefficients from the grid cells, and query the preset accident level comparison table in combination with the overall collision risk probability value to generate the accident severity level; input the overall collision risk probability value and the accident severity level into the weighted fusion unit to generate a comprehensive risk score; compare the comprehensive risk score with the preset multi-level risk threshold sequence, and output a response command to trigger the response action.

[0034] Further, in step 4, historical positioning data within the spatial calibration interval is extracted. Based on the trajectory points on the predicted motion trajectory, the set of position offset vectors between the predicted trajectory points and the actual trajectory points is calculated. The optimization objective is to minimize the weighted sum of squares of the position offset vectors to obtain an optimized model for the three-dimensional spatial transformation parameters. Based on the optimized model, a quasi-Newton optimization algorithm is used to solve the optimization model, generating the final calibration parameter set, including:

[0035] Step 61: Based on the timestamp alignment results of the actual trajectory point set and the predicted trajectory point set, perform point-by-point spatial position matching; calculate the straight-line distance difference of each matching point pair in the three-dimensional coordinate system and decompose it into X, Y, and Z axial components; combine each axial component to generate a single-point position offset vector, and summarize the position offset vectors of all timestamps to form a position offset vector set.

[0036] Step 62: Assign incremental weight coefficients to the set of position offset vectors according to the time decay characteristics; perform squaring operation on the weighted position offset vectors and sum them up, and construct a three-dimensional spatial transformation parameter optimization model with the goal of minimizing the weighted sum of squares.

[0037] Step 63: Initialize the rotation angle, translation distance, and scaling parameters; iterate based on the quasi-Newton optimization algorithm: calculate the gradient vector of the objective function under the current parameters, update the approximate value of the Hessian inverse matrix, and adjust the parameters adaptively along the gradient in the opposite direction; terminate the iteration when the change in the objective function is lower than the convergence threshold, and output a set of calibration parameters consisting of the final rotation angle, translation distance, and scaling ratio.

[0038] Secondly, a building equipment operation data intelligent analysis system includes:

[0039] The data acquisition module is used to acquire device status data and environmental perception data;

[0040] The input module is used to obtain the predicted motion trajectory of the device based on device status data and environmental perception data;

[0041] The fusion module is used to fuse environmental perception data with preset building information models and real-time obstacle scanning data to generate a spatial risk field;

[0042] The calculation module is used to select two spatially fixed calibration reference points within the target analysis domain based on the spatial risk field and predicted motion trajectory, generate a reference direction axis according to the spatial relationship between the two calibration reference points, divide continuous spatial calibration intervals along the reference direction axis, extract historical positioning data within the spatial calibration intervals, calculate the set of position offset vectors between the predicted trajectory points and the actual trajectory points based on the trajectory points on the predicted motion trajectory, use minimizing the weighted sum of squares of the position offset vectors as the optimization objective to obtain a three-dimensional spatial transformation parameter optimization model, and solve the optimization model using a quasi-Newton optimization algorithm to generate the final calibration parameter set.

[0043] The output module is used to correct the predicted motion trajectory based on the final calibration parameter set, analyze the spatial relationship between the corrected trajectory and the spatial risk field, calculate the collision risk probability and accident severity level, generate a comprehensive risk score, and trigger response actions according to preset thresholds.

[0044] Thirdly, a computing device includes:

[0045] One or more processors;

[0046] A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0047] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0048] The above-described solution of the present invention has at least the following beneficial effects:

[0049] This method utilizes a sensor array to collect real-time equipment status data (such as vibration, temperature, and rotational speed) and environmental perception data (such as temperature, humidity, dust concentration, and light intensity). It then combines this data with the 3D spatial coordinates of a pre-defined Building Information Model (BIM) and real-time obstacle scanning data (such as dynamic mapping results from LiDAR and visual sensors), achieving a high-dimensional fusion of "equipment-environment-space" ternary data. On one hand, the BIM model provides spatial parameters throughout the equipment's lifecycle (such as installation location and operational trajectory planning path), providing a precise spatial benchmark for environmental data.

[0050] On the other hand, real-time obstacle scanning data dynamically updates information on temporary obstacles (such as construction material stockpiles and temporary equipment) in the built environment, avoiding the environmental lag problem of traditional static BIM models. It also solves the problem of cumulative errors caused by multipath effects and signal blockage in traditional positioning technology in steel structure construction and underground space operation scenarios, providing a reliable position reference for automated construction equipment. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of a method for intelligent analysis of building equipment operation data provided by an embodiment of the present invention.

[0052] Figure 2 This is a schematic diagram of an intelligent analysis system for building equipment operation data provided in an embodiment of the present invention. Detailed Implementation

[0053] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0054] like Figure 1 As shown, an embodiment of the present invention proposes an intelligent analysis method for building equipment operation data, the method comprising the following steps:

[0055] Step 1: Acquire device status data and environmental perception data;

[0056] Step 2: Based on the device status data and environmental perception data, obtain the predicted motion trajectory of the device;

[0057] Step 3: Integrate the environmental perception data with the preset building information model and real-time obstacle scanning data to generate a spatial risk field;

[0058] Step 4: Based on the spatial risk field and the predicted motion trajectory, select two spatially fixed calibration reference points within the target analysis domain, and generate a reference direction axis according to the spatial relationship between the two calibration reference points; divide continuous spatial calibration intervals along the reference direction axis; extract historical positioning data within the spatial calibration intervals; calculate the set of position offset vectors between the predicted trajectory points and the actual trajectory points based on the trajectory points on the predicted motion trajectory; use minimizing the weighted sum of squares of the position offset vectors as the optimization objective to obtain a three-dimensional spatial transformation parameter optimization model; based on the optimization model, use a quasi-Newton optimization algorithm to solve the optimization model and generate the final calibration parameter set.

[0059] Step 5: Correct the predicted motion trajectory based on the final calibration parameter set, analyze the spatial relationship between the corrected trajectory and the spatial risk field, calculate the collision risk probability and accident severity level, generate a comprehensive risk score, and trigger response actions according to preset thresholds.

[0060] In this embodiment of the invention, the method uses a sensor array to collect real-time equipment status data (such as vibration, temperature, and rotational speed) and environmental perception data (such as temperature, humidity, dust concentration, and light intensity). It then combines this data with the three-dimensional spatial coordinates of a pre-defined Building Information Model (BIM) and real-time obstacle scanning data (such as dynamic mapping results from LiDAR and visual sensors), achieving a high-dimensional fusion of "equipment-environment-space" three-dimensional data. On one hand, the BIM model provides spatial parameters throughout the equipment's lifecycle (such as installation location and operational trajectory planning path), providing a precise spatial benchmark for environmental data.

[0061] On the other hand, real-time obstacle scanning data dynamically updates information on temporary obstacles (such as construction material stockpiles and temporary equipment) in the built environment, avoiding the environmental lag problem of traditional static BIM models. It also solves the problem of cumulative errors caused by multipath effects and signal blockage in traditional positioning technology in steel structure construction and underground space operation scenarios, providing a reliable position reference for automated construction equipment.

[0062] In a preferred embodiment of the present invention, step 1, acquiring device status data and environmental perception data, includes:

[0063] Step 11: Acquire the original voltage signal through the strain gauge array, and perform analog-to-digital conversion on the original voltage signal to generate discrete stress data; input the discrete stress data and the joint rotation pulse number acquired by the encoder into the forward kinematics process to generate real-time spatial pose coordinates; process the real-time spatial pose coordinates, hydraulic pressure sensor readings, motor current sampling values ​​and control operation commands to generate a structured equipment state data matrix.

[0064] Step 12: Acquire raw point cloud data in polar coordinate system using LiDAR, and perform coordinate transformation on the raw point cloud data to generate a three-dimensional point set in global coordinate system; based on the three-dimensional point set, filter out noise points based on the spatial distribution characteristics of adjacent frames to generate denoised point cloud data; acquire Doppler frequency shift signals using millimeter-wave radar, and perform moving target detection calculation on the frequency shift signals to generate obstacle radial velocity vectors; fuse the denoised point cloud data, obstacle radial velocity vectors, and RGB-D depth maps acquired by binocular vision sensors to generate an environmental perception data package;

[0065] Step 13 involves aligning the structured device status data matrix with the environmental awareness data packet using time synchronization. This is done by interpolating based on the hardware clock synchronization signal to generate time-synchronized device status data frames and environmental awareness data frames.

[0066] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows:

[0067] In step 11 above, the original voltage signal of the building equipment structure is acquired in real time by a strain gauge array. The continuous voltage signal is converted into a discrete digital signal by an analog-to-digital converter (ADC) to form discrete stress data related to the stress distribution of the equipment. The data reflects the stress state of each part of the equipment.

[0068] The discrete stress data is combined with the joint rotation pulse count collected by the encoder and processed by the forward kinematics algorithm. The process is based on the mechanical structure model of the equipment. Each joint angle (converted from the pulse count) is substituted into the kinematic equation to calculate the real-time spatial pose coordinates of the end effector or key component of the equipment.

[0069] Angle conversion: Based on the encoder's resolution (e.g., 1000 pulses per revolution), convert the number of pulses into the actual angle. For example, if the encoder outputs 1000 pulses per revolution, then each pulse corresponds to 0.36° (360° / 1000). Assuming a joint collects 500 pulses, the rotation angle is 500 × 0.36° = 180°. The encoder's initial position (zero position) needs to be considered to ensure the consistency of the angle calculation reference.

[0070] Starting from the equipment base (the origin of the coordinate system), the angle of the first joint (derived from the encoder) is substituted into the calculation to determine the coordinate transformation (such as rotation or translation) of the joint relative to the base, thus obtaining the position and orientation of the end of the first link.

[0071] Using the end of the first link as the new reference point, and substituting the angle of the second joint, along with the link length parameter, calculate the coordinate transformation of the end of the second link (for example, how the position of the end of link 2 in the coordinate system of joint 1 changes after joint 2 rotates by an angle of θ2). Process each joint in turn, and gradually transmit the angle changes of each joint to the end effector through a "coordinate transformation recursion". For example, for a three-joint robotic arm, calculate the transformation of joint 1 → joint 2 → joint 3 in turn, and finally obtain the position (X, Y, Z) and attitude (pitch angle, yaw angle, roll angle) of the end effector in the base coordinate system.

[0072] The real-time spatial pose coordinates, readings from hydraulic pressure sensors (reflecting hydraulic load), motor current sampling values ​​(reflecting motor operating status), and control operation commands (such as movement, start / stop signals) are encoded and integrated into a structured equipment status data matrix according to time series and data type. Each element in the matrix corresponds to multi-dimensional equipment status parameters at a certain moment.

[0073] In step 12 above, the lidar acquires raw point cloud data in polar coordinates and converts it into a three-dimensional point set in global coordinates through a coordinate transformation matrix (including rotation and translation parameters), so that the point cloud data at different times have a unified spatial reference; dynamic noise reduction is performed based on the spatial distribution characteristics of point clouds in adjacent frames (such as point density and distance continuity), and outlier noise points are filtered out through clustering or statistical filtering methods to generate denoised point cloud data, thereby improving the accuracy of environmental modeling.

[0074] Millimeter-wave radar collects the Doppler frequency shift signal reflected by the target, extracts the frequency shift features through frequency domain analysis (such as FFT transformation), and calculates the radial velocity vector of the obstacle in combination with the radar detection principle to determine the speed and direction of moving objects in the environment.

[0075] The noise-reduced point cloud data, obstacle radial velocity vectors, and RGB-D depth maps acquired by a binocular vision sensor are fused together. Through coordinate registration (aligning visual data and point cloud data to the same coordinate system) and data association (matching multi-source data with the same physical target), an environmental perception data package containing the 3D structure of the environment, target position, and motion state is generated.

[0076] In step 13 above, the device status data matrix and the environmental perception data packet are input to the time alignment module. Based on the hardware clock synchronization signal (such as GPS clock or synchronization pulse), the sensor data with different sampling frequencies are interpolated and aligned (such as linear interpolation or spline interpolation) to eliminate the time offset of each sensor data and generate device status data frames and environmental perception data frames under the same timestamp, so as to ensure data timing consistency.

[0077] The time-synchronized device status data frames are output to the physical-machine learning fusion model for analyzing device operating status or predicting faults; the environmental perception data frames are output to the spatial risk field construction module for building a real-time environmental model and identifying potential risk areas.

[0078] In this embodiment of the invention, a sensor array covers both equipment status and environmental perception, enabling comprehensive monitoring of the operational status of building equipment and improving the real-time performance and completeness of data acquisition. Forward kinematics and point cloud denoising ensure the accuracy of equipment pose calculation and 3D environmental modeling, providing a reliable basis for equipment control and risk assessment. The fusion of LiDAR, millimeter-wave radar, and visual data overcomes the limitations of single sensors (such as the advantages of LiDAR in low light and the color information of vision), enhancing the perception of dynamic obstacles and complex environments. Hardware clock-based interpolation alignment eliminates data timing deviations, avoiding decision-making errors caused by time asynchrony, providing a unified time reference for subsequent model analysis and risk map construction, and improving the reliability and safety of decision-making.

[0079] In a preferred embodiment of the present invention, step 2, obtaining the predicted motion trajectory of the device based on device status data and environmental perception data, includes:

[0080] Step 21: Receive the time synchronization device status data frame and extract the real-time spatial pose coordinates and control operation command codes from it; input the real-time spatial pose coordinates into the device dynamic equation library and generate an initial motion trajectory segment based on the rigid body kinematics principle; input the initial motion trajectory segment and control operation command codes into the motion constraint filter and combine them with the device mechanical structure parameters to generate a physically compliant trajectory segment.

[0081] Step 22: Input the physically compliant trajectory segment into the pre-trained trajectory feature extractor to decompose it into a three-dimensional feature vector representing the trajectory characteristics; match the three-dimensional feature vector with the actual trajectory pattern library under similar working conditions to generate a trajectory offset set; input the trajectory offset set into the adaptive weighting unit, and dynamically allocate weights to generate a comprehensive correction factor based on the environmental features contained in the time-synchronized environmental perception data frame.

[0082] Step 23: Input the physically compliant trajectory segment and the comprehensive correction factor into the trajectory synthesizer, perform frame-by-frame pose correction through time axis alignment to obtain the corrected trajectory segment; perform smoothness optimization processing on the corrected trajectory segment to eliminate trajectory jitter, and generate an optimized continuous motion trajectory as the predicted motion trajectory of the device.

[0083] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows:

[0084] Step 21 above extracts two types of core information from the time-synchronized device status data frame: real-time spatial pose coordinates: including the device's current three-dimensional position (x, y, z) - representing its absolute position in the global coordinate system, and attitude parameters (such as pitch angle, yaw angle, roll angle) - describing the device's own orientation and rotation state; control operation command encoding: such as motor speed commands, joint angle adjustment commands, etc., reflecting the actions that the control operation expects the device to perform (e.g., "bend the left arm 30 degrees" "move forward at a speed of 0.5 m / s").

[0085] Input the real-time pose coordinates into the device's dynamic equation library (built-in rigid body kinematics model). Based on the principles of rigid body kinematics, calculate the predicted motion state for the next time period (e.g., 0.1 seconds or 0.5 seconds in the future) in three steps: Based on the current pose change rate (e.g., the position difference from the previous moment to the current moment, the amount of attitude change), and combined with the velocity / acceleration target in the operation command (e.g., the command requires "accelerate to 1 m / s"), estimate the current motion velocity (unit: m / s or degrees / second) and acceleration (unit: m / s²). 2 or degrees / second 2The speed range is determined by the maximum operating speed of the equipment (e.g., the maximum speed of a wheeled robot is 2m / s, and the maximum rotation speed of the robotic arm joint is 180 degrees / second), while the acceleration is limited by the power of the equipment (e.g., the maximum acceleration corresponding to the maximum torque of the motor).

[0086] Based on the current position, velocity, and acceleration, assuming the motion is a uniformly accelerated or variablely accelerated process (smoothly transitioning according to the operation command), calculate the three-dimensional position at each time point in the next time interval. For example, if the current velocity is 0.5 m / s and the acceleration is 0.2 m / s²,... 2 The position within the next 0.1 seconds is calculated as follows: Position = Current Position + Velocity × Time + 0.5 × Acceleration × Time 2 "Logical recursion ensures that the position coordinates are within the device's workspace (e.g., the end effector of the robotic arm must not enter prohibited areas, and the mobile robot must not exceed the preset boundaries).

[0087] By combining the rate of change of attitude angles and the attitude adjustment target in the operation command (such as "turn 15 degrees to the right"), the attitude parameters (angles) in the next time period are calculated through Euler angle transformation or quaternion interpolation to ensure continuous attitude change (such as avoiding non-physical movements such as instantaneous flipping).

[0088] Finally, an initial motion trajectory segment is generated, which is a sequence of continuous predicted positions and attitude points (e.g., one point every 0.01 seconds), describing the motion trend of the device under the ideal physical model. The initial trajectory segment and the operation command code are input into the motion constraint filter, and double verification is performed in combination with the device's mechanical structure parameters (e.g., factory-set hardware limitations): checking whether each predicted pose point meets the device's hardware limitations: for the robotic arm, verifying whether the joint angle is within the allowable range (e.g., shoulder joint rotation angle -90° to +90°), and whether the linkage interferes (e.g., the end effector must not collide with the base); for the mobile device, checking whether the wheel speed exceeds the maximum speed, and whether the steering angle causes a risk of slippage (e.g., the minimum turning radius limit of Ackerman steering).

[0089] Remove or correct pose points that exceed constraints (e.g., adjust out-of-bounds joint angles to the threshold boundary); if the velocity suddenly increases beyond the maximum allowable range of acceleration at a certain time point (e.g., the equipment specifies that the acceleration must not exceed 1 m / s²), 2 If a transition point is inserted, the gradient of velocity change conforms to the constraints; for "jumps" in attitude changes (such as instantaneous 180-degree rotation), intermediate attitudes are generated through interpolation to ensure smooth and executable motion.

[0090] Step 22 above inputs the physically compliant trajectory segment into a pre-trained trajectory feature extractor (based on deep learning or a statistical model) to extract a three-dimensional feature vector from the trajectory's time-series data. These features characterize the core attributes of the trajectory: motion trend features reflect whether the trajectory is moving in a straight line, turning at a curve, or reciprocating, quantified by information such as trajectory inflection point density and direction change frequency; velocity fluctuation features describe the stability of velocity at each time point, embodied by indicators such as velocity standard deviation and acceleration extrema; spatial distribution features characterize the concentration or dispersion of the trajectory in the pose space, measured by parameters such as the distribution range of position coordinates and the fluctuation amplitude of attitude angles. The final generated three-dimensional feature vector, with each dimension's value range determined according to the specific feature definition (e.g., values ​​after 0-1 normalization), comprehensively describes the current trajectory's behavior pattern.

[0091] The 3D feature vector is input into the historical behavior analysis module and matched with the historically stored actual trajectory pattern library for similarity. Each record in the pattern library corresponds to a typical working condition (such as unloaded flat ground operation, loaded uphill climbing, high-speed obstacle avoidance, etc.), and each record contains the feature vector and actual motion trajectory under the working condition. The similarity calculation logic is as follows: the similarity is determined by calculating the "closeness" between the current feature vector and the historical feature vector. For example, the distance between the vectors is calculated (the smaller the distance, the higher the similarity), or the consistency of the vector directions is calculated (the closer the directions, the higher the similarity). After normalization, the similarity result ranges from 0 to 1 (0 represents complete dissimilarity, and 1 represents complete consistency).

[0092] Select 3-5 most similar historical operating conditions from high to low similarity, retrieve the corresponding actual trajectories, and compare the physical compliant trajectories with the historical trajectories point by point in time to calculate positional deviations (differences in front / back, left / right, and up / down distances) and attitude deviations (differences in pitch angle, yaw angle, and roll angle), forming a set of trajectory offsets (each similar operating condition corresponds to a set of offsets). Input the set of trajectory offsets for each similar operating condition into an adaptive weighted unit. The unit assigns a weight between 0 and 1 to the offset of each similar operating condition based on real-time environmental characteristics (such as obstacle distance, equipment load, ground roughness, etc.) (the sum of all weights is 1).

[0093] For work conditions with high similarity and matching environment: If the feature vector similarity of a historical work condition is ≥0.8, and the current environmental parameters (such as load weight) are relatively close to the historical environmental records of the work condition, a higher weight is assigned (e.g., 0.3-0.4, with a single work condition weight not exceeding 0.5); For work conditions with medium similarity or differences in environment: If the similarity is between 0.5 and 0.8, or if the environmental parameters have changed to some extent, a medium weight is assigned (e.g., 0.1-0.3); For work conditions with low similarity: When the similarity is <0.5, the weight approaches 0, and the offset is ignored.

[0094] The environmental complexity will adjust the weighting bias as a whole: when the environmental sensor detects complex situations such as approaching obstacles or ground bumps, the total weight of all similar working condition offsets will be increased (e.g., the total weight will be increased from 0.5 to 0.7) to enhance the correction effect of historical experience on the trajectory; when the environment is stable, the total weight will be reduced (e.g., the total weight will be 0.3-0.4), and the prediction of the physical model will be the main factor; each offset will be multiplied by its corresponding weight and then summed to generate a comprehensive correction factor that includes the direction and magnitude of position and attitude correction.

[0095] In step 23 above, the physically compliant trajectory segment and the comprehensive correction factor are input into the trajectory synthesizer, the trajectory points are aligned according to the time axis, and the correction factor is applied to each pose point. Position correction: the (x, y, z) coordinates are adjusted based on the offset weighting; attitude correction: the rotation parameters are adjusted based on the attitude deviation, and the corrected trajectory segment is generated.

[0096] The corrected trajectory segment is smoothed to eliminate trajectory jitter or abrupt changes caused by the correction, and finally outputs a continuous and smooth optimized motion trajectory, which is transmitted to the calibration reference point processing flow (such as for path planning or control execution).

[0097] In this embodiment of the invention, an initial trajectory is generated through a dynamic model and combined with mechanical constraint filtering to ensure the trajectory is physically feasible and avoid mechanical damage caused by exceeding the limits of equipment movement. Historical actual trajectories are used to correct deviations and compensate for nonlinear factors (such as friction loss and sensor errors) not considered by the pure physical model, making the predicted trajectory closer to the actual operating state. Based on environmental characteristics, the correction weights are dynamically adjusted to enable trajectory prediction to adapt to different working conditions (such as complex terrain and load changes), reducing trajectory deviations caused by environmental uncertainties and improving the stability and reliability of equipment movement. Eliminating trajectory jitter can reduce the impact load during equipment movement and extend mechanical life; at the same time, a continuous and smooth trajectory helps improve the positioning accuracy of the end effector (such as robot grasping and welding tasks), reducing operational errors and ensuring the safety of task execution. The physical model provides basic dynamic constraints, reducing the search space for machine learning; machine learning compensates for the modeling errors of the physical model, and the combination of the two improves the long-term accuracy of trajectory prediction while ensuring computational efficiency.

[0098] In a preferred embodiment of the present invention, environmental perception data is fused with a preset building information model and real-time obstacle scanning data to generate a spatial risk field, including:

[0099] Step 31: Receive time-synchronized environmental awareness data frames and extract denoised point cloud data from them; input the denoised point cloud data into the spatial registration process, align it with the coordinate system of the preset building information model, and generate a registration point cloud; input the registration point cloud into the voxel processor, divide it into three-dimensional mesh units of equal size, and mark the static obstacle occupancy area.

[0100] Step 32: Extract the radial velocity vector and RGB-D depth map of the obstacle from the time-synchronized environmental perception data frame; generate the future time series location set of the dynamic obstacle from the radial velocity vector, RGB-D depth map and denoised point cloud data; calculate the spatiotemporal distribution probability of the dynamic obstacle in the three-dimensional grid cell based on the future time series location set, and generate the spatiotemporal probability distribution of the dynamic obstacle.

[0101] Step 33: Generate an initial risk field based on the static obstacle occupancy area markings and the spatiotemporal probability distribution of dynamic obstacles; normalize the initial risk field to generate a spatial risk field.

[0102] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows:

[0103] In step 31 above, after obtaining the raw point cloud from the time-synchronized environmental perception data frame (such as LiDAR or millimeter-wave radar data), noise is removed through two filtering processes, and outlier filtering is performed:

[0104] Set a threshold for the number of neighboring points (usually 5-50, adjusted according to the density of the environmental point cloud; use a smaller value for dense scenes such as indoors, and a larger value for open scenes such as outdoors). Calculate the average distance between each point and its neighboring points. If the distance of a point exceeds "average + 1-3 times the standard deviation" (1 times to filter out mild noise, 3 times to retain more points), it is identified as an outlier and removed, retaining only valid point clouds that conform to statistical laws. Voxel mesh downsampling: Divide the 3D space into equal-sized cubic meshes (voxels). The mesh side length is set according to sensor accuracy and computational efficiency (usually 0.05-0.5 meters; use smaller sizes for high-precision scenes such as robot navigation, and larger sizes for macroscopic modeling). Only one representative point (such as the center point) is retained for each voxel, reducing the amount of data while preserving the environmental contour.

[0105] Spatial location matching is performed between the denoised point cloud and the preset building information model (BIM): significant structural features (such as wall corners, column apex, stair corners) in the point cloud and corresponding features in the BIM model are extracted, and feature point pairs are established through manual labeling or automatic identification by algorithms (such as point A in the point cloud corresponding to point A′ in the BIM).

[0106] Based on the positional differences of feature point pairs, rotation and translation parameters are calculated. Rotation parameters: The point cloud orientation is adjusted by Euler angles (pitch angle, yaw angle, roll angle) to make the axis of the point cloud coordinate system consistent with the global coordinate system of BIM (e.g., the X-axis points to the east side of the building, and the Z-axis is vertically upward). The angle range is limited to -180° to 180° to avoid directional confusion. Translation parameters: The positional offset of the point cloud as a whole on the X, Y, and Z axes is calculated (e.g., usually -100 meters to 100 meters in a building scene), and the numerical stability is ensured by normalization (scaling according to the maximum size of the scene). Finally, the point cloud coordinate system is aligned to the global coordinate system of BIM through rotation and translation transformation matrices to generate a registered point cloud (the two are completely overlapped in spatial position).

[0107] The registered point cloud is input into the voxelization processor, which divides the 3D space into equal-sized grid cells (e.g., 0.1m × 0.1m × 0.1m, adjusted according to scene accuracy requirements). Each grid cell is checked one by one. If it contains point cloud data (i.e., occupied by static objects such as walls or fixed equipment), it is marked as a "static obstacle occupied area"; if it is empty (no point cloud), it is marked as a "passable area".

[0108] Through processing, continuous point cloud data is converted into a discrete three-dimensional mesh matrix, with each mesh carrying a "placeholder" or "passage" status label, forming a structured representation of the static environment.

[0109] Step 32 above involves separating dynamic obstacles from time-synchronized environmental perception data frames (such as two consecutive frames of LiDAR point cloud or camera RGB-D image) through the following steps: comparing the point cloud data of the current frame with that of the previous frame, marking points whose positions have changed (i.e., dynamic points), and ensuring that the point cloud positions of stationary objects are basically overlapping (allowing for minor noise errors); using a density clustering algorithm (such as clustering nearby dynamic points into a group based on the point cloud spacing) to divide the dynamic points into independent obstacle targets (such as pedestrians, vehicles, and mobile robotic arms), with each cluster corresponding to one moving obstacle.

[0110] Extract the cluster center coordinates of the obstacle in the current frame (e.g., 3D position (x1, y1, z1)) and the corresponding center coordinates (x0, y0, z0) of the previous frame, and calculate the displacement vector between the two frames (Δx=x1-x0, Δy=y1-y0, Δz=z1-z0). The direction of the vector is the direction of the obstacle's movement (e.g., from (0, 0, 0) to (1, 0, 0) represents movement along the positive X-axis).

[0111] Given that the time interval between two frames of data is Δt (e.g., 0.1 seconds), the magnitude of the velocity is the magnitude of the displacement vector divided by Δt, that is, the velocity is equal to the square root of the sum of the squares of Δx, Δy, and Δz divided by Δt, and the unit is meters per second (if it is a two-dimensional planar motion, then the Z-axis is ignored).

[0112] If the calculated speed exceeds the device's preset reasonable range (e.g., pedestrian speed exceeding 5m / s is considered noise), the current value is replaced with the average speed of the previous few frames to ensure the reasonableness of the speed data.

[0113] Combining the RGB-D depth map (which provides obstacle shape and distance information) with the aforementioned velocity vector: the obstacle boundary is identified by the grayscale values ​​of the depth map pixels, and the length, width, and height dimensions are determined (e.g., a pedestrian is approximately a cylinder 1.8m high and 0.5m wide); assuming that the obstacle maintains its current direction and speed (uniform linear motion) in the short term, multiple predicted position points are generated according to the time series (e.g., in the future 0.5 seconds, 1 second, 1.5 seconds) (the position at each time point = current position + speed × time), forming a set of future time series positions.

[0114] Random perturbations (simulating possible speed fluctuations or direction changes) are added to each predicted location, and the probability of each 3D grid cell being occupied by an obstacle at different times is calculated. The closer the grid is to the current location and the shorter the time, the higher the probability of it being occupied (e.g., the probability of the grid at the current location being occupied in the next 0.5 seconds is set to 0.8, and after 1 second, it drops to 0.6 due to increased uncertainty). The probability values ​​are represented by color intensity (0 is white, 1 is red), forming a 3D heatmap that includes the time dimension (XY axis is spatial coordinates, Z axis is time axis), which intuitively shows the high-risk areas where dynamic obstacles may appear in the future.

[0115] In step 33 above, the static obstacle-occupied area (marked as high risk) and the spatiotemporal distribution heat map are input into the risk overlay device, and the comprehensive risk value is calculated by grid cell: the risk value of the grid occupied by static obstacles is directly set to the maximum value; the risk value of the grid that may be occupied by dynamic obstacles is linearly mapped according to the probability of the heat map (e.g., probability 0.8 corresponds to risk value 0.8); and the original risk distribution map is generated (each grid cell has a risk value between 0 and 1).

[0116] By combining load-bearing structural data (such as the location of load-bearing walls and columns) in the building information model, the original risk distribution map is reinforced in key areas: grids close to load-bearing walls and columns, regardless of whether there are obstacles, have their risk value increased by 20%-50% (because collisions may cause structural damage); in key passage areas such as staircases and elevator shafts, the risk value of dynamic obstacles is given an additional weight (such as multiplied by 1.5 times) to ensure safety priority.

[0117] The enhanced risk distribution map is globally normalized, mapping the risk values ​​of all grids to the 0-1 range (e.g., using linear transformation to make the maximum value 1 and the minimum value 0), generating a standardized spatial risk field. The map includes the location of static obstacles, the future risk distribution of dynamic obstacles, and the risk level of key building areas, and is output to the calibration benchmark processing flow (e.g., for avoiding high-risk areas when using path planning).

[0118] In this embodiment of the invention, the spatiotemporal distribution of both static building structures and dynamic obstacles is considered simultaneously, avoiding the omission of dynamic risks (such as suddenly intruding moving objects) caused by relying solely on static maps, thus improving the real-time performance and completeness of environmental modeling. By combining the load-bearing structural information of the BIM model, risk weighting is applied to key areas, prioritizing the protection of building safety (such as avoiding collisions with load-bearing walls), making it suitable for scenarios with high structural protection requirements, such as factories and buildings. The uncertainty of dynamic obstacles is quantified using a probability diffusion model, and the risk distribution is visually represented in the form of a heatmap, facilitating dynamic path adjustment by the planning algorithm based on risk levels (such as detouring through high-risk areas and proceeding through low-risk areas). The normalized risk map uses a unified numerical standard (0-1) to represent risk levels, allowing direct interface with path planning, motion control, and other modules, reducing data conversion costs between different areas and improving overall decision-making efficiency.

[0119] In a preferred embodiment of the present invention, step 4 involves selecting two spatially fixed calibration reference points within the target analysis domain based on the spatial risk field and the predicted motion trajectory; generating a reference direction axis based on the spatial relationship between the two calibration reference points; dividing continuous spatial calibration intervals along the reference direction axis; extracting historical positioning data within the spatial calibration intervals; calculating the set of position offset vectors between the predicted trajectory points and the actual trajectory points based on the trajectory points on the predicted motion trajectory; minimizing the weighted sum of squares of the position offset vectors as the optimization objective to obtain a three-dimensional spatial transformation parameter optimization model; and solving the optimization model using a quasi-Newton optimization algorithm to generate the final calibration parameter set, including:

[0120] Step 41: Based on the three-dimensional spatial distribution characteristics of the spatial risk field, select two calibration reference points with fixed spatial positions within the target analysis domain; generate a reference direction axis based on the spatial coordinates of the two calibration reference points and calculate the direction cosine; divide the reference direction axis into equal intervals and generate a continuous spatial calibration interval with the division points as boundaries.

[0121] Step 42: Extract historical positioning data of equipment operation within each spatial calibration interval to obtain the actual trajectory point set; extract the predicted trajectory point set that matches the timestamp of the historical positioning data based on the predicted motion trajectory; calculate the spatial offset between the predicted trajectory point and the actual trajectory point point point by point to generate a position offset vector set.

[0122] Step 43: Normalize and weight the position offset vector set to generate a weighted offset vector set; take minimizing the sum of squares of the weighted offset vectors as the objective function, and combine the constraints of the three-dimensional space transformation parameters to obtain the three-dimensional space transformation parameter optimization model.

[0123] Step 44: Input the objective function into the optimization solver and initialize the parameter estimates of the rotation matrix, translation vector, and scaling factor; iteratively execute the quasi-Newton optimization algorithm, calculate the gradient vector of the objective function under the parameters, update the approximate value of the Hessian inverse matrix, and perform adaptive parameter updates along the gradient in the opposite direction; terminate the iteration when the change in the objective function is lower than the convergence threshold, and output the final calibration parameter set composed of the rotation matrix, translation vector, and scaling factor.

[0124] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows:

[0125] Step 41 above involves selecting static structural feature points (such as the tops of building columns, corners of fixed equipment, and intersections of walls) from the spatial risk field. These selections must meet three conditions: 1) Fixed location: They must be immovable, permanent structures, excluding dynamic objects (such as pedestrians or mobile devices); 2) Identifiability: They must be reliably detected by sensors (such as lidar or vision cameras) and have distinct features (such as right-angle inflection points or height abrupt changes); 3) Reasonable spacing: The distance between the two points must be moderate (usually 5-20 meters) to ensure the accuracy of the directional axis while avoiding overly dense calibration intervals due to excessive proximity. For example, in a factory scenario, the top endpoints of two load-bearing columns in a workshop can be selected as reference points A and B.

[0126] Vector definition: Taking reference point A as the starting point and reference point B as the ending point, a three-dimensional spatial vector is formed (i.e., a directed line segment from the coordinates of A to the coordinates of B). Calculate the components of the vector on the three coordinate axes X, Y, and Z (i.e., subtracting the coordinates of point A from the coordinates of point B to obtain Δx, Δy, and Δz); calculate the total length of the vector (i.e., the three-dimensional distance of the line segment); divide the components on the X, Y, and Z axes by the total length respectively to obtain the three direction cosine values. These three values ​​represent the cosines of the angles between the vector and the X, Y, and Z axes, respectively, reflecting the "projection ratio" of the vector in the three coordinate axis directions.

[0127] The cosine value for each direction is between -1 and 1. When a vector is exactly in the same direction as a coordinate axis, the corresponding direction cosine is 1 (same direction) or -1 (opposite direction); when a vector is perpendicular to a coordinate axis, the corresponding direction cosine is 0.

[0128] Divide the data into equal intervals along the reference direction axis (i.e., the line connecting A and B): Set the interval distance according to the scene accuracy requirements (usually 0.5-2 meters). For example, choose a 0.5-meter interval in a high-precision positioning scene and a 1-meter interval in a large-scale scene.

[0129] Centered on each dividing point, the system extends outwards into a plane perpendicular to the reference direction axis, forming a continuous cylindrical calibration interval (such as a cylinder with a radius of 1 meter centered on the direction axis). Each interval covers a certain range of three-dimensional space for subsequent local calibration.

[0130] Number the intervals according to the division order (e.g., interval 1, interval 2, etc.) to ensure that each interval is unique and continuous on the reference direction axis.

[0131] In step 42 above, the actual positioning trajectory point set within the current calibration interval is extracted (position points collected in real time by sensors such as GPS and LiDAR, with timestamps), and the predicted motion trajectory point set with the corresponding timestamp is retrieved to ensure that the two sets of points correspond one-to-one in time order and the time difference does not exceed a preset threshold (such as 50 milliseconds).

[0132] For each pair of corresponding points, calculate the three-dimensional spatial offset between the predicted trajectory point and the actual trajectory point (Δx = actual x - predicted x, Δy = actual y - predicted y, Δz = actual z - predicted z), forming a set of position offset vectors. For example, if the actual coordinates of a point are (10, 5, 2) and the predicted coordinates are (10.1, 4.9, 2.1), then the offset vector is (-0.1, 0.1, -0.1).

[0133] In step 43 above, the offset vector is weighted according to the risk value of each point in the spatial risk field (the higher the risk value, the greater the weight). For example, the risk value is mapped to a weight coefficient of 0.5-1.5 (risk value 1 corresponds to weight 1.5, risk value 0 corresponds to weight 0.5), so that the offset of high-risk areas (such as near load-bearing walls) is corrected first.

[0134] With the objective function of "minimizing the sum of squares of the weighted offset vectors", a three-dimensional spatial transformation parameter optimization model is constructed, including rotation matrix parameters: three Euler angles (pitch, yaw, roll) describing the rotation of the coordinate system; translation vector parameters: three-dimensional coordinate offsets (tx, ty, tz); scaling factor parameters: global scale adjustment coefficient (to handle sensor scale error, usually with an initial value of 1 and a range of 0.95-1.05).

[0135] The model must meet the following constraints: the rotation matrices must be orthogonal (to ensure the rationality of the transformation), and the scaling factor must be non-negative.

[0136] In step 44 above, the rotation matrix parameter is initially set to "no rotation", meaning that the X, Y, and Z axes in 3D space are completely consistent with the global coordinate system (equivalent to the object not rotating at all); the translation vector parameter is initially set to (0, 0, 0), indicating that there is no positional offset between the predicted trajectory coordinate system and the actual trajectory coordinate system in the initial stage; the scaling factor parameter is initially set to 1, assuming that the predicted trajectory and the actual trajectory have the same scale (no magnification or reduction error).

[0137] The initial parameters are assumed to have "zero error" as the starting point for iterative optimization, and then gradually corrected through data-driven approaches.

[0138] Calculate the gradient of the objective function (weighted sum of squared offsets) with respect to the rotation angle, translation amount, and scaling factor under the given parameters. The gradient is a vector, and its direction indicates the direction in which the objective function grows the fastest. During iteration, the final parameters are searched along the opposite direction of the gradient (the direction of fastest descent). For example, if the gradient of the rotation angle θ is positive, it means that decreasing θ will make the objective function decrease. Therefore, the next step is to adjust in the direction of decreasing θ.

[0139] The quasi-Newton method does not directly calculate the complex Hessian matrix (second derivative matrix). Instead, it iteratively approximates the inverse matrix using historical gradient information (used to adjust the "curvature" of the search direction); it records the difference (Δx) between the current parameter vector and the previous parameter vector; it records the difference (Δg) between the current gradient and the previous gradient; and it uses Δx and Δg to update the approximate value of the Hessian inverse matrix. The approximate value of the Hessian inverse matrix is ​​used to transform the gradient vector into a better search direction, thus accelerating the convergence speed.

[0140] Try different step sizes along the search direction (inverse gradient direction combined with Hessian inverse matrix) to find the step size that maximizes the decrease in objective function and ensures computational stability; preset a maximum step size, and then decrease it proportionally (e.g., multiply by 0.5 each time); for each step size, calculate the objective function value under the new parameters, requiring that it meets the "sufficient decrease" condition (e.g., the decrease in function value exceeds a threshold related to the step size); select the maximum step size that meets the condition as the update step size for this iteration; avoid step sizes that are too large, causing parameter oscillation, or step sizes that are too small, causing slow convergence.

[0141] Add "search direction × final step size" to the current parameters to obtain the new rotation matrix parameters, translation vector parameters, and scaling factor parameters. For example, if the current rotation angle is 0.1°, the search direction is decreasing by 0.05°, and the step size is 1, then the new rotation angle will be updated to 0.05°.

[0142] If the change in the objective function is less than the preset threshold (e.g., the change in function value between two iterations is less than 0.001), it indicates that parameter adjustment can no longer significantly improve calibration accuracy; if the number of iterations reaches the upper limit (e.g., 100 times), avoid excessive computation time.

[0143] When the termination condition is met, the final set of calibration parameters is output. For example, rotation matrix parameters: corresponding to three Euler angles (such as pitch angle 0.1°, yaw angle -0.2°, roll angle 0.3°) to correct the directional deviation of the predicted trajectory; translation vector parameters: such as (0.05m, -0.03m, 0.01m) to compensate for the positional offset of the predicted trajectory; scaling factor parameters: such as 1.02 to adjust the overall scale of the predicted trajectory to be consistent with the actual one.

[0144] In this embodiment of the invention, static structural point selection based on an environmental risk map ensures that the reference point positions are fixed and reproducible, avoiding interference from dynamic obstacles and improving the long-term stability of the calibration model. Calibration intervals are divided along the reference axis, and segmented optimization is performed based on the offset characteristics at different locations (e.g., the difference between near and far offsets), addressing the problem that global calibration cannot cover local errors. Higher weights are assigned to trajectory offsets in high-risk areas, prioritizing the correction of positioning errors that may lead to collisions (e.g., offsets when approaching obstacles), optimizing the motion trajectory while ensuring safety. The quasi-Newton method reduces computation by approximating the Hessian matrix. Simultaneous optimization of rotation, translation, and scaling parameters can handle sensor installation deviations (rotation / translation) and scale errors (scaling), making it suitable for joint calibration of different types of positioning devices (e.g., LiDAR, visual cameras).

[0145] In a preferred embodiment of the present invention, step 5 involves correcting the predicted motion trajectory based on the final calibration parameter set, performing spatial relationship analysis between the corrected trajectory and the spatial risk field, calculating the collision risk probability and accident severity level, generating a comprehensive risk score, and triggering a response action according to a preset threshold, including:

[0146] Step 51: Extract the rotation matrix, translation vector, and scaling factor parameters from the final calibration parameter set; input the trajectory point set of the predicted motion trajectory into the space transformer, apply the parameters to perform coordinate transformation calculation, and generate the preliminary corrected trajectory; perform time-series filtering on the preliminary corrected trajectory to eliminate high-frequency jitter noise and generate the final corrected motion trajectory.

[0147] Step 52: Map the spatial position sequence of the final corrected motion trajectory to the grid coordinate system of the spatial risk field to generate the mapping relationship between trajectory points and grid cells; extract the real-time risk value of the grid cell where each trajectory point is located based on the mapping relationship.

[0148] Step 53: Calculate the dynamic collision probability value based on the real-time risk value and the geometric distance from the trajectory point to the nearest obstacle; perform an integral operation on the dynamic collision probability value along the trajectory time window to generate the overall collision risk probability value;

[0149] Step 54: Extract the predefined structural importance coefficients from the grid cells, and query the preset accident level comparison table in combination with the overall collision risk probability value to generate the accident severity level; input the overall collision risk probability value and the accident severity level into the weighted fusion unit to generate a comprehensive risk score; compare the comprehensive risk score with the preset multi-level risk threshold sequence, and output a response command to trigger the response action.

[0150] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows:

[0151] In step 51 above, three types of parameter rotation matrices are extracted from the final calibration parameter set to correct the directional deviation of the predicted trajectory (e.g., rotating the predicted trajectory around the X-axis by 0.1° to align with the actual coordinate system); translation vector: to compensate for the positional offset of the predicted trajectory (e.g., translating all predicted points along the Y-axis by -0.03 meters to eliminate sensor installation position errors); scaling factor: to adjust the overall scale of the predicted trajectory (e.g., magnifying the predicted distance by 1.02 times to match the ranging accuracy of the actual sensor).

[0152] Three types of parameters are applied to each predicted trajectory point to generate a preliminary corrected trajectory point (the coordinate transformation process is similar to a combination of "rotation, scaling, and translation"). The preliminary corrected trajectory point sequence is then subjected to time series filtering (such as moving average filtering or median filtering): the filter window length is set (e.g., including the current point and two points before and after it, for a total of five points); the coordinates of the current point are replaced with the average (or median) coordinates of the points within the window to eliminate high-frequency jitter (such as millimeter-level position jumps) caused by fluctuations in calibration parameters, thus generating the final smooth corrected motion trajectory.

[0153] In step 52 above, each continuous location point (x, y, z) in the final corrected trajectory is converted into a grid coordinate system of the spatial risk field: given the grid size of the risk map (e.g., 0.1 m × 0.1 m × 0.1 m) and the global coordinate origin, the grid cell index corresponding to each location point is calculated (e.g., the x-coordinate is divided by the grid size and rounded to obtain the grid number in the X-axis direction).

[0154] If a location point is exactly located at the intersection of two grids, it is simultaneously mapped to the adjacent grid (risk values ​​are assigned according to distance weight). Based on the mapping results, the risk value of each grid cell is extracted (a value between 0 and 1, where 0 is no risk and 1 is the highest risk): grids occupied by static obstacles are directly marked with a risk value of 1; for grids in the dynamic obstacle heatmap, the risk value is updated in real time according to the probability distribution (e.g., if the probability of a grid being occupied in the next second is 0.7, then the risk value is marked as 0.7).

[0155] In step 53 above, for each trajectory point in the grid, the geometric distance to the nearest obstacle in the grid (such as the distance to a static wall or the distance to the predicted position of a dynamic pedestrian) is calculated: the closer the distance, the higher the probability of dynamic collision (e.g., the probability is 0.3 when the distance is 0.5 meters and rises to 0.8 when the distance is 0.2 meters).

[0156] By combining the grid risk value (basic risk) and the geometric distance (dynamic correction), a dynamic collision probability value (range 0-1) for a single trajectory point is generated.

[0157] Set a time window length (e.g., trajectory points within the next 2 seconds), and integrate (sum or weighted average) the dynamic collision probability values ​​of all trajectory points within the window: assign higher weights (e.g., weight 0.8) to recent trajectory points (e.g., within the next 0.5 seconds), and lower weights (e.g., weight 0.2) to distant points (e.g., after the next 1.5 seconds); to obtain the overall collision risk probability (e.g., 0.6 means there is a 60% probability of a collision within the next 2 seconds).

[0158] In step 54 above, the "structural importance coefficient" (predefined parameter, such as 1.5 for load-bearing wall areas and 1.0 for ordinary areas) of the grid where the trajectory point is located is extracted from the risk map. Combined with the preset accident level comparison table (such as coefficient ≥1.5 corresponding to "major accident risk", 1.0-1.5 corresponding to "general accident risk"), the accident severity level is generated (divided into three levels: high, medium and low).

[0159] The collision risk probability (0-1) and accident severity level (quantified by a coefficient of 0.5-1.5) are weighted and fused (e.g., risk probability accounts for 60% weight, severity accounts for 40%) to obtain a comprehensive score (0-1.5). This score is then compared with a preset three-level threshold sequence (e.g., high risk threshold 1.2, medium risk 0.8, low risk 0.5): Score ≥ 1.2: Emergency stop command is triggered, and the equipment brakes immediately; 0.8 ≤ score < 1.2: Deceleration command is triggered, and the equipment reduces its operating speed by 50%; 0.5 ≤ score < 0.8: Early warning prompt is triggered, and a risk warning is sent to the control operation; Score < 0.5: No action is taken, and normal operation continues.

[0160] In this embodiment of the invention, calibration parameters compensate for sensor installation deviations and scale errors, improving the consistency between the corrected trajectory and the actual motion, and reducing the risk of collisions caused by inaccurate positioning. By combining the real-time position of trajectory points, obstacle distances, and the importance of environmental structures, collision risk is transformed into quantifiable probability values ​​and severity levels, avoiding the limitations of traditional binary "unsafe / safe" judgments and achieving refined risk management. Different levels of response actions (from warning to emergency stop) are triggered based on the comprehensive risk score, maximizing equipment operating efficiency while ensuring safety (e.g., normal operation in low-risk scenarios, deceleration for medium-risk scenarios, and immediate stop for high-risk scenarios). Not only are the immediate risks of dynamic obstacles considered, but the long-term safety priorities of static structures (e.g., protecting load-bearing walls) are also incorporated, forming a dual protection mechanism of "real-time dynamic risk calculation + static risk structural enhancement," suitable for industrial and building scenarios with extremely high safety requirements. Temporal filtering eliminates trajectory jitter, avoiding misjudgments caused by sensor noise; a weighted fusion algorithm balances different risk factors, enhancing adaptability to complex environments (e.g., multiple obstacles, complex structural areas).

[0161] In a preferred embodiment of the present invention, step 4 involves extracting historical positioning data within the spatial calibration interval; calculating the set of position offset vectors between the predicted trajectory points and the actual trajectory points based on the trajectory points on the predicted motion trajectory; minimizing the weighted sum of squares of the position offset vectors as the optimization objective to obtain a three-dimensional spatial transformation parameter optimization model; and solving the optimization model using a quasi-Newton optimization algorithm to generate the final calibration parameter set, including:

[0162] Step 61: Based on the timestamp alignment results of the actual trajectory point set and the predicted trajectory point set, perform point-by-point spatial position matching; calculate the straight-line distance difference of each matching point pair in the three-dimensional coordinate system and decompose it into X, Y, and Z axial components; combine each axial component to generate a single-point position offset vector, and summarize the position offset vectors of all timestamps to form a position offset vector set.

[0163] Step 62: Assign incremental weight coefficients to the set of position offset vectors according to the time decay characteristics; perform squaring operation on the weighted position offset vectors and sum them up, and construct a three-dimensional spatial transformation parameter optimization model with the goal of minimizing the weighted sum of squares.

[0164] Step 63: Initialize the rotation angle, translation distance, and scaling parameters; iterate based on the quasi-Newton optimization algorithm, calculate the gradient vector of the objective function under the current parameters, update the approximate value of the Hessian inverse matrix, and adjust the parameters adaptively along the gradient direction; terminate the iteration when the change in the objective function is lower than the convergence threshold, and output a set of calibration parameters consisting of the final rotation angle, translation distance, and scaling ratio.

[0165] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows:

[0166] Step 61 above, actual trajectory point set: location data collected in real time by device sensors (such as LiDAR, vision camera), each point containing three-dimensional coordinates (x, y, z). a y a , z a ) and timestamp t a Predicted trajectory point set: Based on the theoretical motion trajectory, each point contains a corresponding timestamp t. p 3D coordinates (x) p y p , z p ).

[0167] Find point pairs with the smallest time difference (typically no more than 50 milliseconds) to ensure that each pair of trajectory points represents the actual and predicted positions of the device at the same moment (e.g., actual point t). a =10:00:00.123 and the predicted point t p =10:00:00.125 is considered a matching pair.

[0168] For each pair of matching points, calculate the positional deviation (left-right offset) between the predicted trajectory point and the actual trajectory point along the three coordinate axes: the difference in positional deviation (left-right offset) along the X-axis: the X-coordinate of the actual point minus the X-coordinate of the predicted point, i.e., Δx = x a -x p If Δx is positive, it means the predicted point is to the left of the actual point; if it is negative, it means the predicted point is to the right.

[0169] Y-axis direction difference (forward and backward offset): actual Y-coordinate minus predicted Y-coordinate, i.e., Δy = y a -y p If Δy is positive, it means the predicted point is behind the actual point; if it is negative, it means the predicted point is in front (assuming the Y-axis is defined as the direction of equipment movement).

[0170] Z-axis direction difference (vertical offset): actual Z-coordinate minus predicted Z-coordinate, i.e., Δz = z a -z p If Δz is positive, it means the predicted point is below the actual point; if it is negative, it means the predicted point is above (applicable to scenarios with height changes, such as the vertical movement of a robotic arm).

[0171] The differences between the three coordinate axes are combined into a three-dimensional vector (Δx, Δy, Δz). The direction of the vector represents the offset direction of the predicted trajectory relative to the actual trajectory, and the length represents the straight-line distance offset. For example, if the actual point coordinates are (10, 5, 2) and the predicted point coordinates are (10.1, 4.9, 2.1), then Δx = -0.1 (the predicted point is 0.1 meters to the left), Δy = +0.1 (the predicted point is 0.1 meters behind), and Δz = -0.1 (the predicted point is 0.1 meters below), forming the offset vector (-0.1, 0.1, -0.1).

[0172] The offset vectors of all matching point pairs are aggregated to form a set of location offset vectors containing time series information (e.g., 100 offset vectors containing 100 time points, each vector recording the three-dimensional offset at the corresponding time).

[0173] In step 62 above, an increasing weight coefficient is assigned to each offset vector based on its timestamp (e.g., the weight of the most recent 10% of time points is 1, the weight of the next most recent 20% is 0.8, and so on, decreasing sequentially). The logic is that the closer the data is to the current moment, the greater its impact on real-time calibration, and the higher its weight, thus avoiding outdated data from interfering with the current error judgment.

[0174] For each weighted offset vector, calculate the squared values ​​of the X, Y, and Z components (e.g., (Δx × weight)). 2 Then, the squared values ​​of the three components are added together to obtain the weighted squared value of a single vector; the weighted squared values ​​of all time points are accumulated to obtain a total value, which is the optimization target value. The smaller the value, the smaller the overall deviation between the predicted trajectory and the actual trajectory.

[0175] With "minimizing the optimization target value" as the core, a three-dimensional spatial transformation model containing three types of parameters is constructed: rotation angle: corrects the directional deviation of the predicted trajectory (such as the rotation angle around the X, Y, and Z axes); translation distance: compensates for the positional deviation of the predicted trajectory (translation amount in the X, Y, and Z axis directions); scaling ratio: adjusts the overall scale of the predicted trajectory (such as global scaling caused by sensor ranging error).

[0176] In step 63 above, the rotation angle is initially set to 0°, assuming that the predicted trajectory and the actual trajectory are completely consistent in direction in the initial state (without rotation deviation).

[0177] Translation distance: initially set to (0, 0, 0) meters, meaning the origin of the predicted trajectory coordinate system coincides with the origin of the actual trajectory coordinate system, with no positional offset; scaling ratio: initially set to 1, assuming the scale of the predicted trajectory is the same as the actual trajectory (e.g., a predicted distance of 1 meter corresponds to an actual distance of 1 meter); key logic: the initial parameters adopt the assumption of "zero error" as the starting point for iterative optimization, and the deviation is gradually corrected through historical data.

[0178] Calculate the rate of change (gradient) of the target value (weighted sum of squared offsets) with respect to rotation angle, translation distance, and scaling ratio under the current parameters. For example, if the gradient of the rotation angle θ is positive, it means that increasing θ will increase the target value, so the next iteration should adjust in the direction of decreasing θ; if the gradient of the translation distance on the X-axis is negative, it means that increasing the X-axis translation will decrease the target value, so it should be adjusted in the positive direction; the direction of the gradient vector represents the direction in which the target value increases the fastest, and the final parameters are searched in the opposite direction (the direction of the fastest decrease) during iteration.

[0179] The quasi-Newton method approximates the curvature of the objective function using historical iteration data (alternatively replacing the calculation of the second derivative):

[0180] Record the difference between the current parameter vector and the previous parameter vector (e.g., when the rotation angle changes from 0.1° to 0.08°, the difference is -0.02°); record the difference between the current gradient and the previous gradient (e.g., when the X-axis gradient changes from 0.5 to 0.3, the difference is -0.2); use the two sets of differences to update the curvature approximation matrix (similar to "memorizing" the curvature of the objective function), so that the next search direction is closer to the final path. The curvature approximation matrix helps the algorithm determine the "curvature" direction of parameter adjustment and avoid getting trapped in a local finality.

[0181] By combining the gradient inverse direction with the curvature approximation matrix, the final parameter adjustment direction is obtained (e.g., adjusting the rotation angle and translation distance simultaneously). The step size is dynamically adjusted: a preset initial step size is used (e.g., adjusting the rotation angle by 0.1° each time, and the translation distance by 0.01 meters each time); different step sizes are tried to calculate the target value under the new parameters, requiring the target value to decrease by a preset condition (e.g., decreasing by at least 0.01); if the target value decreases significantly, the step size is increased to accelerate convergence; if the decrease is slow or increases, the step size is decreased to avoid oscillation (e.g., multiplying the step size by 0.5). Example: the first iteration step size is 0.1, the target value decreases by 0.1, the next step size is increased to 0.15; if the target value only decreases by 0.02, the step size is reduced to 0.05.

[0182] Add "adjustment direction × final step size" to the current parameters to get new parameter values: Rotation angle = current angle + adjustment direction × step size (e.g., from 0.1° to 0.05°); Translation distance = current distance + adjustment amount of each axis × step size (e.g., from 0.02 meters to 0.025 meters for the X-axis); Scaling ratio = current ratio + adjustment amount × step size (e.g., from 1.01 to 1.012).

[0183] If the change in the target value is less than the set threshold (e.g., 0.001) for three consecutive iterations, it indicates that the parameter adjustment has little effect on improving the optimization target; the number of iterations reaches the upper limit (e.g., 100 times) to avoid excessive computation time; when the termination condition is met, extract the parameters of the final iteration: rotation angle: such as 0.3° around the X-axis and -0.2° around the Y-axis, to correct the directional deviation of the predicted trajectory; translation distance: such as 0.03 meters around the X-axis and -0.01 meters around the Y-axis, to compensate for positional offset; scaling ratio: such as 1.02, to adjust the scale of the predicted trajectory to be consistent with the actual one; combine the parameters to generate the final calibration parameter set.

[0184] In this embodiment of the invention, by assigning higher weight to recent data, trajectory deviations in the current environment are corrected first, avoiding interference from early noise data on real-time calibration and improving the dynamic adaptability of the calibration model. The three-dimensional offset is decomposed into X, Y, and Z axes for separate optimization. The quasi-Newton algorithm reduces computation through curvature approximation, and the number of iterations typically converges within 10-20, making it suitable for online real-time calibration (e.g., updating calibration parameters every second), avoiding the time-consuming issues of traditional algorithms. Multiple error types are covered, optimizing rotation, translation, and scaling parameters. This can handle various errors such as sensor installation angle deviation (rotation), fixed position offset (translation), and ranging scale error (scaling), making it suitable for joint calibration of positioning devices of different brands and types. Automatic optimization based on historical trajectory data eliminates the need for manually preset error models and adapts to error characteristics in different scenarios (e.g., ranging deviations caused by multiple indoor reflections, and GPS signal drift outdoors), improving reliability in complex environments.

[0185] like Figure 2 As shown, embodiments of the present invention also provide an intelligent analysis system for building equipment operation data, comprising:

[0186] The data acquisition module is used to acquire device status data and environmental perception data;

[0187] The input module is used to obtain the predicted motion trajectory of the device based on device status data and environmental perception data;

[0188] The fusion module is used to fuse environmental perception data with preset building information models and real-time obstacle scanning data to generate a spatial risk field;

[0189] The calculation module is used to select two spatially fixed calibration reference points within the target analysis domain based on the spatial risk field and predicted motion trajectory, generate a reference direction axis according to the spatial relationship between the two calibration reference points, divide continuous spatial calibration intervals along the reference direction axis, extract historical positioning data within the spatial calibration intervals, calculate the set of position offset vectors between the predicted trajectory points and the actual trajectory points based on the trajectory points on the predicted motion trajectory, use minimizing the weighted sum of squares of the position offset vectors as the optimization objective to obtain a three-dimensional spatial transformation parameter optimization model, and solve the optimization model using a quasi-Newton optimization algorithm to generate the final calibration parameter set.

[0190] The output module is used to correct the predicted motion trajectory based on the final calibration parameter set, analyze the spatial relationship between the corrected trajectory and the spatial risk field, calculate the collision risk probability and accident severity level, generate a comprehensive risk score, and trigger response actions according to preset thresholds.

[0191] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0192] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0193] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0194] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent analysis of construction equipment operation data, characterized by, The method comprises Step 1, obtaining equipment state data and environment perception data; Step 2, obtaining equipment predicted motion trajectory according to equipment state data and environment perception data, comprising: Step 21, receiving time-synchronized equipment state data frames and extracting real-time spatial pose coordinates and control operation instruction codes therefrom; inputting the real-time spatial pose coordinates into a device dynamics equation library to generate an initial motion trajectory segment based on rigid body kinematics principles; inputting the initial motion trajectory segment and the control operation instruction codes into a motion constraint filter to generate a physically compliant trajectory segment in combination with device mechanical structure parameters; Step 22, inputting the physically compliant trajectory segment into a pre-trained trajectory feature extractor to decompose and obtain a three-dimensional feature vector representing trajectory characteristics; matching the three-dimensional feature vector with an actual trajectory pattern library under similar working conditions to generate a trajectory offset set; inputting the trajectory offset set into an adaptive weighting unit and dynamically assigning weights based on environment features contained in the time-synchronized environment perception data frames to generate a comprehensive correction factor; Step 23, inputting the physically compliant trajectory segment and the comprehensive correction factor into a trajectory synthesizer for frame-by-frame pose correction through time axis alignment to obtain a corrected trajectory segment; performing smoothness optimization processing on the corrected trajectory segment to eliminate trajectory jitter and generating an optimized continuous motion trajectory as the predicted motion trajectory of the equipment; Step 3, fusing environment perception data with a pre-set building information model and real-time obstacle scanning data to generate a spatial risk field; Step 4, based on the spatial risk field and the predicted motion trajectory, selecting two spatially fixed calibration reference points in the target analysis domain, generating a reference direction axis according to the spatial relationship of the two calibration reference points; dividing the spatial calibration interval along the reference direction axis; extracting historical positioning data within the spatial calibration interval and calculating a set of position offset vectors between predicted trajectory points and actual trajectory points according to trajectory points on the predicted motion trajectory; taking the weighted sum of squares of the position offset vectors as the optimization objective to obtain a three-dimensional space transformation parameter optimization model; solving the optimization model using a quasi-Newton optimization algorithm based on the optimization model to generate a final calibration parameter set; Step 5, correcting the predicted motion trajectory based on the final calibration parameter set, performing spatial relationship analysis on the corrected trajectory and the spatial risk field, calculating collision risk probability and accident severity level, generating a comprehensive risk score, and triggering a response action according to a pre-set threshold.

2. The method of claim 1, wherein, Step 1, obtaining equipment state data and environment perception data, comprising: Step 11, collecting raw voltage signals through a strain gauge array and performing analog-to-digital conversion on the raw voltage signals to generate discrete stress data; inputting the discrete stress data and joint rotation angle pulse numbers collected by an encoder into a kinematics forward solution process to generate real-time spatial pose coordinates; processing based on the real-time spatial pose coordinates, hydraulic pressure sensor readings, motor current sampling values, and control operation instructions to generate a structured equipment state data matrix; Step 12, collect polar coordinate system original point cloud data by laser radar, and perform coordinate transformation on the original point cloud data to generate a global coordinate system three-dimensional point set; according to the three-dimensional point set, filter out noise points based on the spatial distribution characteristics of adjacent frames to generate denoised point cloud data; collect Doppler frequency shift signals by millimeter wave radar, and perform motion target detection calculation on the frequency shift signals to generate obstacle radial velocity vectors; fuse the denoised point cloud data, obstacle radial velocity vectors, and RGB-D depth map collected by the binocular vision sensor to generate an environment perception data packet; Step 13, perform time alignment process on the structured device state data matrix and the environment perception data packet, and perform interpolation alignment based on the hardware clock synchronization signal to generate time-synchronized device state data frames and environment perception data frames.

3. The method of claim 2, wherein, Step 3, fuse the environment perception data with the preset building information model and real-time obstacle scanning data to generate a spatial risk field, including: Step 31, receive the time-synchronized environment perception data frames and extract the denoised point cloud data therefrom; input the denoised point cloud data into a spatial registration process to align with the coordinate system of the preset building information model to generate a registered point cloud; input the registered point cloud into a voxelization processor to segment into equal-size three-dimensional grid cells and mark static obstacle occupancy areas; Step 32, extract the obstacle radial velocity vectors and RGB-D depth map from the time-synchronized environment perception data frames; generate a future time sequence position set of dynamic obstacles based on the obstacle radial velocity vectors, RGB-D depth map, and denoised point cloud data; calculate the space-time distribution probability of dynamic obstacles in the three-dimensional grid cells based on the future time sequence position set to generate a dynamic obstacle space-time probability distribution; Step 33, generate an initial risk field based on the static obstacle occupancy area markers and the dynamic obstacle space-time probability distribution; perform normalization processing on the initial risk field to generate a spatial risk field.

4. The method of claim 3, wherein, Step 4, based on the spatial risk field and the predicted motion trajectory, select two spatially fixed calibration reference points in the target analysis domain, and generate a reference direction axis based on the spatial relationship of the two calibration reference points; Divide the spatial calibration interval along the reference direction axis; Extract historical positioning data within the spatial calibration interval, and calculate a set of position offset vectors between the predicted trajectory points and the actual trajectory points based on the trajectory points on the predicted motion trajectory; take minimizing the weighted sum of squares of the position offset vectors as the optimization objective to obtain a three-dimensional space transformation parameter optimization model; According to the optimization model, solve the optimization model using a quasi-Newton optimization algorithm to generate a final calibration parameter set, including: Step 41, based on the three-dimensional spatial distribution characteristics of the spatial risk field, select two spatially fixed calibration reference points in the target analysis domain; generate a reference direction axis based on the spatial coordinates of the two calibration reference points and calculate the direction cosine; perform equidistant segmentation along the reference direction axis to generate a continuous spatial calibration interval with the segmentation points as boundaries; Step 42, extract the historical positioning data of the device running in each spatial calibration interval to obtain the actual trajectory point set; extract the predicted trajectory point set matching the time stamp of the historical positioning data according to the predicted motion trajectory; calculate the spatial offset of the predicted trajectory point and the actual trajectory point point by point to generate a position offset vector set; Step 43, normalize and weight the position offset vector set to generate a weighted offset vector set; take the minimization of the weighted offset vector sum as the objective function, and combine the three-dimensional space transformation parameter constraint condition to obtain a three-dimensional space transformation parameter optimization model; Step 44, input the objective function into the optimization solver, initialize the parameter estimation value of the rotation matrix, translation vector and scaling factor; iterative execution using the quasi-Newton optimization algorithm, calculate the gradient vector of the objective function under the parameters, update the approximate value of the Hessian inverse matrix, and update the parameters in the opposite direction of the gradient with step adaptation; terminate the iteration when the change of the objective function is less than the convergence threshold, and output the final calibration parameter set composed of the rotation matrix, translation vector and scaling factor; terminate the iteration when the change of the objective function is less than the convergence threshold to obtain the final calibration parameter set.

5. The method of claim 4, wherein, Step 5, correct the predicted motion trajectory based on the final calibration parameter set, analyze the spatial relationship between the corrected trajectory and the spatial risk field, calculate the collision risk probability and accident severity level, generate a comprehensive risk score, and trigger a response action according to a preset threshold, including: Step 51, extract the rotation matrix, translation vector and scaling factor parameters from the final calibration parameter set; input the trajectory point set of the predicted motion trajectory into the space transformer, apply the parameters for coordinate transformation calculation to generate a preliminary corrected trajectory; perform time series filtering on the preliminary corrected trajectory to eliminate high-frequency dithering noise and generate a final corrected motion trajectory; Step 52, map the spatial position sequence of the final corrected motion trajectory to the grid coordinate system of the spatial risk field to generate a mapping relationship between the trajectory points and the grid elements; extract the real-time risk value of each trajectory point in the grid element based on the mapping relationship; Step 53, calculate the dynamic collision probability value based on the real-time risk value and the geometric distance from the trajectory point to the nearest obstacle; integrate the dynamic collision probability value along the trajectory time window to generate an overall collision risk probability value; Step 54, extract the pre-defined structure importance coefficient according to the grid element, query the pre-set accident level reference table combined with the overall collision risk probability value to generate the accident severity level; input the overall collision risk probability value and the accident severity level into the weighted fusion device to generate a comprehensive risk score; compare the comprehensive risk score with the preset multi-level risk threshold sequence to output a response instruction to trigger a response action. 6.The method of claim 5, wherein, Step 4, extract the historical positioning data in the spatial calibration interval, and calculate the position offset vector set between the predicted trajectory points and the actual trajectory points according to the trajectory points on the predicted motion trajectory; take the minimization of the weighted sum of the position offset vectors as the optimization objective to obtain a three-dimensional space transformation parameter optimization model; According to the optimization model, solve the optimization model using the quasi-Newton optimization algorithm to generate a final calibration parameter set, including: Step 61, based on the timestamp alignment result of the actual trajectory point set and the predicted trajectory point set, point-by-point spatial position matching is performed; the straight line distance difference of each matched point pair in the three-dimensional coordinate system is calculated and decomposed into X, Y and Z axial components; the single-point position offset vector is generated by combining the axial components, and the position offset vector set is formed by collecting the position offset vectors of all timestamps; Step 62, according to the time decay characteristic, an incremental weight coefficient is given to the position offset vector set; a square operation is performed on the weighted position offset vector and the sum is accumulated, and a three-dimensional space transformation parameter optimization model is constructed to minimize the weighted sum of squares; Step 63, the rotation angle, translation distance and scaling ratio parameters are initialized; based on the quasi-Newton optimization algorithm, the gradient vector of the objective function under the current parameters is calculated, the Hessian inverse matrix approximation is updated, and the parameter adjustment is performed in the opposite direction of the gradient with step adaptation; when the change of the objective function is less than the convergence threshold, the iteration is terminated, and the calibration parameter set composed of the final rotation angle, translation distance and scaling ratio is output.

7. A construction equipment operation data intelligent analysis system that implements the method according to any one of claims 1 to 6, characterized by, Comprise: The acquisition module is used for acquiring device state data and environment perception data; The input module is used for obtaining the device predicted motion trajectory according to the device state data and the environment perception data; The fusion module is used for fusing the environment perception data with the preset building information model and the real-time obstacle scanning data to generate a space risk field; The calculation module is used for selecting two spatially fixed calibration reference points in a target analysis domain based on the space risk field and the predicted motion trajectory, generating a reference direction axis according to the spatial relationship of the two calibration reference points; Divide the continuous spatial calibration interval along the reference direction axis; Extract the historical positioning data in the spatial calibration interval, calculate the position offset vector set between the predicted trajectory points and the actual trajectory points according to the trajectory points on the predicted motion trajectory, and obtain a three-dimensional space transformation parameter optimization model by minimizing the weighted sum of squares of the position offset vectors; according to the optimization model, the quasi-Newton optimization algorithm is used to solve the optimization model to generate a final calibration parameter set; The output module is used for correcting the predicted motion trajectory based on the final calibration parameter set, performing spatial relationship analysis on the corrected trajectory and the space risk field, calculating the collision risk probability and the accident severity level, generating a comprehensive risk score, and triggering a response action according to a preset threshold.

8. A computing device, comprising: Comprise: One or more processors; Storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, so that the one or more processors implement the method of any one of claims 1 to 6.

Images