Self-adaptive construction process for hoisting and positioning spherical curved glass of glass curtain wall

Abstract

The invention relates to the technical field of glass curtain wall construction, and discloses a glass curtain wall spherical curved glass hoisting and positioning self-adaptive construction process which comprises the following steps: firstly, acquiring a dynamic state of a to-be-mounted glass panel in a hoisting process and environmental parameters of a construction area in real time; then predicting a future movement track of the glass panel in a future prediction time window based on the obtained dynamic state and the environment parameters according to a preset coupling dynamics model; and comparing the predicted future motion trail with a preset target installation pose. The accurate six-degree-of-freedom dynamic state of the glass panel is obtained in real time by deploying a multi-mode sensor system, and a high-precision current state and a high-precision target reference are provided for subsequent control in combination with a target mounting pose after deviation correction of a high-precision main body structure, so that manual positioning errors and precision limitation in the traditional hoisting process are effectively overcome.

Description

Technical Field

[0001] This invention relates to the field of glass curtain wall construction technology, specifically to an adaptive construction process for hoisting and positioning spherical curved glass in glass curtain walls. Background Technology

[0002] Glass curtain walls, especially those with complex geometric shapes such as spherical surfaces, are becoming increasingly popular in modern high-rise buildings and landmark projects due to their unique visual aesthetics and architectural expressiveness. However, compared with traditional flat glass curtain walls, the installation and construction of spherical curved glass curtain walls face a series of more severe technical challenges.

[0003] In existing technologies, the hoisting and positioning of spherical curved glass mainly relies on experienced on-site construction personnel. Typically, the glass panels are connected to the crane using traditional lifting equipment. During the hoisting process, the operators perform initial positioning and attitude adjustment by manually operating the crane or using simple traction ropes. However, this operation method based on human experience and visual judgment has inherent limitations. First, the geometry of spherical curved glass panels is complex, and the size, curvature, and installation angle of each panel vary significantly, making accurate target pose recognition and manual alignment extremely difficult. Simultaneously controlling the position and attitude of the glass panels in three-dimensional space places extremely high demands on the operator's spatial perception ability and operational precision.

[0004] Secondly, the high-altitude working environment is complex and changeable. External disturbances pose a serious threat to hoisting accuracy and safety. Wind is one of the most significant challenges. Even moderate gusts can cause unpredictable swaying and swinging of glass panels, causing them to deviate from their intended path or even collide with surrounding structures or already installed panels. Under such dynamic loads, traditional manual operations struggle to compensate for these external disturbances in real time and accurately, leading to a significant decrease in the positioning accuracy of the glass panels, extended installation time, and a substantial increase in construction risks. Furthermore, the slight swaying of the crane itself, the elastic deformation of the slings, and changes in on-site lighting conditions all further exacerbate the difficulty of positioning. Moreover, existing manual methods... Semi-automatic hoisting methods often lack consideration for the flexible deformation of the glass panels themselves. Especially for large, ultra-thin, or irregularly shaped glass panels, they may undergo a certain degree of elastic deformation under hoisting and wind loads. If the impact of this deformation on the final posture cannot be detected and predicted in real time, it will lead to stress concentration after installation or failure to accurately match the main structure, thereby affecting the overall performance and service life of the curtain wall. Currently, the ability to acquire the real-time, complete six-degree-of-freedom state (including position, attitude, velocity, and angular velocity) of the glass panels during hoisting, as well as the ability to predict future motion trends, is relatively lacking. This makes it impossible for traditional hoisting methods to achieve true active prevention and adaptive control. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an adaptive construction process for the hoisting and positioning of spherical curved glass in glass curtain walls. This process solves the problems of difficulty in ensuring installation accuracy, efficiency, and safety in existing hoisting and construction of spherical curved glass curtain walls, due to the difficulty in accurately controlling the precise posture of complex curved glass panels, the significant impact of external dynamic environmental disturbances, and the lack of real-time prediction and adaptive adjustment capabilities.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an adaptive construction process for the hoisting and positioning of spherical curved glass in glass curtain walls, comprising the following steps:

[0007] S1. First, the dynamic state of the glass panel to be installed during the hoisting process and the environmental parameters of the construction area are acquired in real time; then, based on the acquired dynamic state and environmental parameters, and according to a preset coupled dynamics model, the future motion trajectory of the glass panel within a future prediction time window is predicted.

[0008] S2. The predicted future motion trajectory is then compared with a preset target installation pose to generate control commands for driving a multi-degree-of-freedom intelligent lifting tool.

[0009] S3. Finally, the control command is executed by the intelligent hoist to achieve adaptive adjustment of the glass panel's posture, making it approach the target installation posture.

[0010] Preferably, the method for obtaining the dynamic state includes: acquiring multi-source data, including visual images and inertial measurement data, through a multi-modal sensor system deployed on the glass panel or the smart hanger; and using a filtering algorithm to fuse the multi-source data to estimate the six degrees of freedom state of the glass panel in real time as the dynamic state.

[0011] Preferably, the method for obtaining the environmental parameters includes: collecting real-time meteorological data from multiple spatial nodes through a distributed sensor network deployed in the construction area; and based on the real-time meteorological data, constructing a three-dimensional local wind field model of the construction area in real time using a spatial interpolation algorithm to output the environmental parameters.

[0012] Preferably, the coupled dynamics model integrates the flexible mechanical properties of the glass panel, the dynamic properties of the hoisting system, and the external dynamic loads derived from the environmental parameters.

[0013] Preferably, the method for generating the control command includes: using a model predictive control method to solve an online optimization problem within a rolling time window, with the objective of minimizing the deviation between the future motion trajectory of the glass panel and the target installation pose, thereby obtaining the optimal control sequence as the control command.

[0014] Preferably, before solving the online optimization problem, a pre-trained reinforcement learning model is used to generate macroscopic policy parameters for adjusting the weight coefficients in the online optimization problem based on the dynamic state and the future trajectory.

[0015] Preferably, during the hoisting and construction process, the verified macroscopic strategy parameters and corresponding working condition information are stored in an adaptive knowledge base, and the reinforcement learning model is continuously optimized using the adaptive knowledge base.

[0016] Preferably, the adaptive adjustment includes: predicting the pose deviation trend caused by external disturbances based on the future motion trajectory, and driving the intelligent lifting device to perform active compensation operations in order to suppress or offset the impact of the external disturbances in advance.

[0017] Preferably, before the process is executed, the method further includes: performing a three-dimensional scan of the main building structure to obtain its as-built point cloud model, comparing the as-built point cloud model with the design model to determine the actual construction deviation; and correcting the target installation pose based on the actual construction deviation.

[0018] Preferably, the multi-degree-of-freedom intelligent lifting device has the ability to actively adjust in three-dimensional translation and three-dimensional rotation directions.

[0019] This invention provides an adaptive construction process for hoisting and positioning spherical curved glass in glass curtain walls.

[0020] It has the following beneficial effects:

[0021] 1. This invention acquires the precise six-degree-of-freedom dynamic state of the glass panel in real time by deploying a multimodal sensor system (vision, IMU, radar, etc.), and combines it with the target installation pose after high-precision main structure deviation correction, providing a high-precision current state and target reference for subsequent control. At the same time, through the model predictive control (MPC) algorithm, the dynamic state is combined with the predicted future motion trajectory to continuously optimize the control commands, enabling the intelligent lifting device to accurately guide the glass panel to approach and stabilize at the final target installation pose, effectively overcoming the manual positioning error and accuracy limitations in the traditional lifting process.

[0022] 2. This invention utilizes a three-dimensional local wind field model constructed in real time through a distributed microclimate sensor network. This model incorporates rapidly changing external environmental parameters into the system considerations. The coupled dynamic model further integrates these environmental parameters (such as wind load) with the mechanical properties of the glass panel itself, predicting its future trajectory under dynamic loads. The intelligent lifting device can proactively perform predictive compensation operations based on this prediction information, offsetting or suppressing external disturbances in advance, rather than responding passively. This significantly improves the robustness and adaptive adjustment capability of the system under dynamic conditions such as complex wind fields and crane swaying.

[0023] 3. This invention uses a reinforcement learning model as a high-level strategy planner to generate macro-strategy parameters based on real-time dynamic states and future motion trajectories, thereby optimizing and guiding the underlying control of the MPC. This hierarchical intelligent decision-making mechanism enables the system to intelligently adjust the control center of gravity according to the current working conditions, prioritizing attitude stability at critical moments, thus effectively improving the intelligence level and construction efficiency of the entire hoisting process.

[0024] 4. This invention uses a multimodal sensing system to monitor the glass panel and its surrounding environment in real time with high precision. Combined with a coupled dynamics model, it predicts the future trajectory of potential collision risks and unstable movements, which can detect and avoid potential dangers in advance. The intelligent lifting device can then actively perform obstacle avoidance or stabilization operations based on these predictions, preventing the glass panel from being damaged or causing personal injury due to accidental shaking, collisions, or other factors. This forward-looking control strategy greatly reduces the safety risks in the lifting of complex curved glass surfaces.

[0025] 5. Throughout the entire hoisting process, this invention comprehensively records all key data. This data is used to build and continuously update an adaptive knowledge base, forming valuable digital construction experience. This data closed-loop mechanism enables the system to continuously learn, iterate, and optimize its control strategies, which not only improves the construction quality and efficiency of the current project but also promotes the continuous development of construction technology. Attached Figure Description

[0026] Figure 1 This is a flowchart of the method steps of the present invention;

[0027] Figure 2 This is a flowchart illustrating the installation posture operation of the present invention. Detailed Implementation

[0028] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Please see the appendix Figure 1 -Appendix Figure 2 This invention provides an adaptive construction process for the hoisting and positioning of spherical curved glass in glass curtain walls, including the following steps:

[0030] S1. First, the dynamic state of the glass panel to be installed and the environmental parameters of the construction area during the hoisting process are acquired in real time. The environmental parameters are acquired by means of: collecting real-time meteorological data from multiple spatial nodes through a distributed sensor network deployed in the construction area; and using a spatial interpolation algorithm based on the real-time meteorological data to construct a three-dimensional local wind field model of the construction area in real time to output the environmental parameters. Then, based on the acquired dynamic state and environmental parameters, and according to a preset coupled dynamic model, which integrates the flexible mechanical properties of the glass panel, the dynamic properties of the hoisting system, and the external dynamic loads converted from the environmental parameters, the future motion trajectory of the glass panel within a future prediction time window is predicted. The dynamic state is acquired by means of: collecting multi-source data, including visual images and inertial measurement data, through a multi-modal sensor system deployed on the glass panel or intelligent hoist; and using a filtering algorithm to fuse the multi-source data to estimate the six degrees of freedom state of the glass panel as the dynamic state in real time.

[0031] Specifically, parametric design software (e.g., but not limited to Rhino / Grasshopper) is first used to create a precise geometric model of the spherical curved surface building. The entire curved surface is then subdivided into several manufacturable and installable glass panel units. The geometric information, material type, surface texture features, spatial coordinates of connection nodes, and suggested installation sequences of these units are all integrated into the BIM platform. For each glass panel unit, the initial theoretical target installation pose T in the global coordinate system {W} is extracted. target This pose is determined by the position of its center of mass. attitude q represented by quaternions target ∈SO(3) is composed of, in addition, the physical properties of each glass panel are extracted, including its mass m, inertial tensor I in the object coordinate system {B}. body For glass panels of different specifications, a detailed mechanical model is established through finite element analysis (FEM), including the stiffness matrix K characterizing its elastic deformation properties in the object coordinate system {B}. d and the mass matrix describing its mass distribution and inertial properties These matrices will be used in subsequent coupled dynamics predictions. To address potential cumulative errors during actual construction, a high-precision 3D laser scanner will be used to comprehensively scan the completed main structure after the curtain wall frame construction is finished, acquiring high-density "as-built" point cloud data P. as-builtUsing point cloud registration algorithms (e.g., the improved Iterative Closest Point (ICP) algorithm), P as-built With the design structural model M in BIM design High-precision alignment is achieved by analyzing the rigid transformation matrix between the two components to accurately calculate the geometric deviation of the main structure relative to the design baseline. Based on the identified deviations, the coordinates of the theoretical connection nodes and final installation positions of each glass panel unit are corrected and optimized, thus resolving the original T-axis in the BIM model. target Updated to the corrected target mounting pose T′ target , including the corrected position p′ target and posture q′ target This ensures the final precision of the glass panel installation.

[0032] At key locations on the construction site, such as the end of the crane boom, high points of the main structure, the surface of installed glass panels, and the periphery of glass panels to be installed, a multimodal sensor system is strategically deployed. This system includes: a multi-view stereo vision camera array for high-frame-rate, high-resolution image data acquisition of surface features and edge contours of the glass panels to be hoisted; a high-precision inertial measurement unit (IMU) integrated into specific locations on the multi-degree-of-freedom intelligent hoist and glass panels to measure their angular velocity, linear acceleration, and attitude angle in real time; millimeter-wave radar / liDAR to provide precise distance and relative velocity information between the glass panels and their surrounding environment (such as installed structures and other equipment) to enhance perception robustness; and a distributed microclimate sensor network containing multiple high-precision wind speed and direction sensors, temperature sensors, humidity sensors, etc., for real-time acquisition of local meteorological data in the construction area. All deployed sensors undergo rigorous geometric and temporal calibration to ensure that their measurement references are consistent with the global coordinate system of the BIM model and are time-synchronized, which provides a foundation for subsequent multi-source data fusion.

[0033] S2. The predicted future motion trajectory is then compared with a preset target installation pose to generate control commands for driving a multi-degree-of-freedom intelligent lifting tool.

[0034] S3. Finally, the intelligent hoist executes control commands to achieve adaptive adjustment of the glass panel's posture, making it approach the target installation posture;

[0035] Specifically, the glass panel (flexible body), multi-degree-of-freedom intelligent lifting device, sling (elastic body), and crane (rigid body / flexible boom) are considered as a unified coupled dynamic system. Its equations of motion can be established using the Lagrange equations or the Newton-Euler equations, with generalized coordinates q. gen The dynamic equations describing the system state can be expressed as follows:

[0036]

[0037] q gen The system's generalized coordinate vector includes the six-degree-of-freedom pose of the glass panel, its elastic deformation mode coordinates, the joint angles of the intelligent lifting device, and the extension, contraction, and swing states of the sling.

[0038] M sys (q gen ): The generalized mass matrix of the system.

[0039] The system's generalized Coriolis force, centrifugal force, and damping matrix.

[0040] K d,sys (q gen ): The generalized stiffness matrix of the system includes the elastic properties of the glass panel and the elasticity of the suspension cable established in S1.

[0041] F ext External generalized force / moment vector, mainly including gravity and instantaneous wind load calculated from the real-time three-dimensional local wind field model in S2, wind load force F wind and torque T wind It can be obtained by integrating the panel surface.

[0042] F ctrl : The generalized control force / torque vector applied by a multi-degree-of-freedom intelligent spreader;

[0043] This coupled dynamics model fully integrates the flexible mechanical properties of the glass panel (through K... d,sys and M sys Indirectly reflected), the dynamic characteristics of the hoisting system (through M) sys C sys (embodied), and external dynamic loads derived from environmental parameters (through F) ext reflect).

[0044] Future motion trajectory prediction:

[0045] The precise dynamic state of the glass panel is estimated by S2 based on the current time k. As initial conditions, and with the future wind load predictions provided by the real-time three-dimensional local wind field model in S2, the above coupled dynamic equations are solved by numerical integration methods (e.g., higher-order Runge-Kutta method).

[0046] In a very short future time window (prediction time domain N) p Within this range, predict the future motion trajectory and attitude evolution sequence of the glass panel. This sequence of future motion trajectories details the glass panel in the future N p Position, attitude, velocity, and angular velocity within a time step.

[0047] The method for generating control commands includes: using model predictive control, within a rolling time window, solving an online optimization problem with the objective of minimizing the deviation between the future motion trajectory of the glass panel and the target installation pose, thereby obtaining the optimal control sequence as the control command; before solving the online optimization problem, using a pre-trained reinforcement learning model, based on the dynamic state and future motion trajectory, generating macroscopic strategy parameters for adjusting the weight coefficients in the online optimization problem; during the hoisting construction process, storing the verified effective macroscopic strategy parameters and corresponding working condition information into an adaptive knowledge base, and using the adaptive knowledge base to continuously optimize the reinforcement learning model;

[0048] Specifically, in the offline phase, a reinforcement learning (RL) model is deployed using the high-fidelity digital twin environment built in S1. This RL model interacts with the simulation environment to learn how to optimally adjust the control strategy to minimize positioning errors and hoisting time under different wind conditions, glass size, hoisting path, and initial deviations. The state space (S) of the RL agent includes the current dynamic state of the glass panel, the deviation from the target installation pose, real-time environmental parameters, etc. The action space (A) of the RL agent is defined as the adjustment of key parameters of the model predictive controller (MPC) in S5 (e.g., the weight coefficients of different error terms in the MPC cost function) through deep reinforcement learning algorithms (e.g., proximal policy optimization (PPO) or soft actor-critic (SAC)). The system performs large-scale training in an offline digital twin environment. During actual construction, this pre-trained reinforcement learning model acts as a high-level policy planner. Based on the current dynamic state and future trajectory provided by S2, it outputs a set of macro-policy parameters to adjust the weight coefficients in the online optimization problem of Model Predictive Control (MPC). For example, when the wind is strong, the RL model may generate a policy that instructs MPC to prioritize enhancing the control of attitude stability. In each hoisting operation, the system structurally stores the macro-policy parameters generated by the RL model and verified to be effective after actual execution, along with their corresponding dynamic states, environmental parameters, and future trajectories, into an adaptive knowledge base. This adaptive knowledge base can serve as an experience database, storing the 'best practice' control strategies under specific complex working conditions.

[0049] By utilizing an adaptive knowledge base, reinforcement learning models can be continuously optimized periodically or in real time (e.g., through experience replay, fine-tuning, etc.). Furthermore, when the system encounters a working condition similar to that recorded in the knowledge base, it can directly call or refer to the historical best strategy, thereby improving the efficiency, robustness, and generalization ability of hoisting. The proactive predictive adaptive control step is the physical execution layer of the system, responsible for translating high-level strategies into precise, real-time physical actions, driving the multi-degree-of-freedom intelligent lifting tool to perform precise pose adjustments.

[0050] Adaptive adjustment includes: predicting the pose deviation trend caused by external disturbances based on the future motion trajectory, and driving the intelligent lifting device to perform active compensation operations to suppress or offset the impact of external disturbances in advance. Before the process is executed, it also includes: performing a 3D scan of the main building structure to obtain its as-built point cloud model, and comparing the as-built point cloud model with the design model to determine the actual construction deviation; based on the actual construction deviation, correcting the target installation pose. The multi-degree-of-freedom intelligent lifting device has the ability to actively adjust in the three-dimensional translation and three-dimensional rotation directions.

[0051] Specifically, a multi-degree-of-freedom intelligent lifting device is a precision robotic arm or parallel mechanism capable of active adjustment in three-dimensional translation and three-dimensional rotation (i.e., six degrees of freedom). This device integrates high-precision servo motors, reducers, attitude sensors, and torque sensors, enabling it to accurately respond to control commands issued by the MPC. The MPC executes only the first command of the calculated optimal control sequence. The intelligent lifting device is driven to adjust its posture. Through this control mechanism, the system can predict the deviation trend of the glass panel's posture caused by external disturbances (such as wind load or crane sway) based on the predicted future motion trajectory. The intelligent lifting device is driven to perform active compensation operations. For example, before the wind load effect manifests, a reverse torque or displacement is applied in advance to suppress or offset the influence of external disturbances, thereby achieving adaptive adjustment of the glass panel's posture. The model predictive controller (MPC) acts as the underlying execution controller, receiving macroscopic strategy parameters (such as adjusted weight coefficients), the future motion trajectory of the glass panel, and the corrected target installation posture T in S1. target ′.

[0052] MPC solves an optimization problem online within a rolling time window, calculating the optimal control sequence that enables the future motion trajectory of the glass panel to approach the target installation pose as quickly and stably as possible while satisfying the kinematic and dynamic constraints of the smart lifting device (e.g., joint torque, velocity, and travel limits).

[0053] The online optimization problem of MPC can be described as follows:

[0054]

[0055] Constraints:

[0056] (System dynamics constraints, based on the S3 model) min ≤u k+j ≤u max (Control input constraints) (State constraints, such as obstacle avoidance and safe range)

[0057] Where, N p To predict the step size, N c To control the step size, Q mpc ,R mpc ,P mpc It is a positive semi-definite weight matrix, whose coefficients are adjusted by the S4 macroeconomic policy parameters. dyn It is a discrete coupled dynamics model established in S3.

[0058] The torque sensor integrated into the intelligent lifting device measures the actual force information between the glass panel and the lifting device in real time during the lifting process. This force feedback data is transmitted in real time to the MPC controller and the coupled dynamics model to further correct the parameters and states of the dynamics model, thereby improving the robustness and accuracy of the control.

[0059] The precise positioning and installation of the glass panel is the physical endpoint of the hoisting process. It achieves high-precision installation of the glass panel and completes the data closed loop. Under continuous closed-loop control, the swaying of the glass panel is effectively suppressed and it is precisely and smoothly guided to the corrected target installation posture Ttarget' in S1. When the system detects that the deviation between the actual posture of the glass panel and the target installation posture is continuously and stably within the preset tolerance range (e.g., position error is less than a specific threshold, attitude error is less than a specific angle), the system issues a "positioned" signal, and on-site technicians perform the final mechanical fixing and connection work. All key data in the entire hoisting process, including sensor readings, dynamic state estimation, real-time wind field model, predicted trajectory, macroscopic strategy parameters, control commands, and actual execution actions, are completely recorded. This data is transmitted back to the central control system for the updating and learning of the adaptive knowledge base, thereby achieving continuous improvement and optimization of the system.

[0060] In summary, this invention provides an adaptive construction process for the hoisting and positioning of spherical curved glass in glass curtain walls. By deploying a multimodal sensor system (vision, IMU, radar, etc.), the precise six-degree-of-freedom dynamic state of the glass panel is acquired in real time. Combined with the target installation posture after high-precision main structure deviation correction, a high-precision current state and target reference are provided for subsequent control. At the same time, through the model predictive control (MPC) algorithm, the dynamic state is combined with the predicted future motion trajectory to continuously optimize control commands, enabling the intelligent hoist to accurately guide the glass panel to approach and stabilize at the final target installation posture. Furthermore, through a three-dimensional local wind field model constructed in real time by a distributed microclimate sensor network, the rapidly changing external environmental parameters are incorporated into the system consideration. The coupled dynamic model further closely integrates these environmental parameters (such as wind load) with the mechanical properties of the glass panel itself, predicting its future motion trajectory under dynamic loads. This significantly improves the robustness and adaptive adjustment capability of the system under dynamic conditions such as complex wind fields and crane swaying.

[0061] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive construction process for hoisting and positioning spherical curved glass in glass curtain walls, characterized in that: Includes the following steps: S1. First, the dynamic state of the glass panel to be installed during the hoisting process and the environmental parameters of the construction area are acquired in real time; then, based on the acquired dynamic state and environmental parameters, and according to a preset coupled dynamics model, the future motion trajectory of the glass panel within a future prediction time window is predicted. S2. The predicted future motion trajectory is then compared with a preset target installation pose to generate control commands for driving a multi-degree-of-freedom intelligent lifting tool. S3. Finally, the control command is executed by the intelligent hoist to achieve adaptive adjustment of the glass panel's posture, making it approach the target installation posture.

2. The adaptive construction process for hoisting and positioning spherical curved glass in glass curtain walls according to claim 1, characterized in that, The method for obtaining the dynamic state includes: collecting multi-source data, including visual images and inertial measurement data, through a multi-modal sensor system deployed on the glass panel or the smart hanger; and using a filtering algorithm to fuse the multi-source data to estimate the six degrees of freedom state of the glass panel in real time as the dynamic state.

3. The adaptive construction process for hoisting and positioning spherical curved glass in glass curtain walls according to claim 1, characterized in that, The method for obtaining the environmental parameters includes: collecting real-time meteorological data from multiple spatial nodes through a distributed sensor network deployed in the construction area; and based on the real-time meteorological data, using a spatial interpolation algorithm to construct a three-dimensional local wind field model of the construction area in real time to output the environmental parameters.

4. The adaptive construction process for hoisting and positioning spherical curved glass in glass curtain walls according to claim 1, characterized in that, The coupled dynamics model integrates the flexible mechanical properties of the glass panel, the dynamic properties of the hoisting system, and the external dynamic loads derived from the environmental parameters.

5. The adaptive construction process for hoisting and positioning spherical curved glass in glass curtain walls according to claim 1, characterized in that, The method for generating the control commands includes: using a model predictive control method to solve an online optimization problem within a rolling time window, with the objective of minimizing the deviation between the future motion trajectory of the glass panel and the target installation pose, thereby obtaining the optimal control sequence as the control commands.

6. The adaptive construction process for hoisting and positioning spherical curved glass in glass curtain walls according to claim 5, characterized in that, Before solving the online optimization problem, a pre-trained reinforcement learning model is used to generate macroscopic policy parameters for adjusting the weight coefficients in the online optimization problem based on the dynamic state and the future trajectory.

7. The adaptive construction process for hoisting and positioning spherical curved glass in glass curtain walls according to claim 6, characterized in that, During the hoisting and installation process, the verified macroscopic strategy parameters and corresponding working condition information are stored in an adaptive knowledge base, and the reinforcement learning model is continuously optimized using the adaptive knowledge base.

8. The adaptive construction process for hoisting and positioning spherical curved glass in glass curtain walls according to claim 1, characterized in that, The adaptive adjustment includes: predicting the pose deviation trend caused by external disturbances based on the future motion trajectory, and driving the intelligent lifting device to perform active compensation operations in order to suppress or offset the impact of the external disturbances in advance.

9. The adaptive construction process for hoisting and positioning spherical curved glass in glass curtain walls according to claim 1, characterized in that, Before the process is executed, the method further includes: performing a three-dimensional scan of the main building structure to obtain its as-built point cloud model, comparing the as-built point cloud model with the design model to determine the actual construction deviation; and correcting the target installation pose based on the actual construction deviation.

10. The adaptive construction process for hoisting and positioning spherical curved glass in glass curtain walls according to claim 1, characterized in that, The multi-degree-of-freedom intelligent lifting device has the ability to actively adjust in three-dimensional translation and three-dimensional rotation directions.

Images