A building robot trajectory planning method for building component installation

Abstract

The application discloses a kind of building component installation-oriented building robot trajectory planning method, belong to building robot trajectory planning technical field.The method includes: obtaining the three-dimensional point cloud data of the work space of building component installation site, and identifies installation datum plane;According to installation datum plane, determine the spatial pose constraint condition of building component in installation process;According to the spatial pose constraint condition, the path search space of collision-free motion path is constructed;In the path search space, the initial motion path is obtained by using the rapid extension random tree algorithm search;The initial motion path is smoothed, and the target trajectory planning result of building robot end effector is obtained.The method extracts installation datum plane from point cloud automatically and determines pose constraint, plans and smooths collision-free path in joint space, so that the path planning is more in line with actual installation demand and movement is more stable.

Description

Technical Field

[0001] This invention relates to the field of construction robot trajectory planning technology, specifically a construction robot trajectory planning method for the installation of building components. Background Technology

[0002] When construction robots install components, they need to generate motion paths for the end effector from the gripping point to the target installation position. Early trajectory planning methods relied on manual teaching or offline programming, which is inefficient and difficult to guarantee installation accuracy when facing changing working environments and diverse components. Existing model-based planning methods usually require the pre-establishment of an accurate environmental model and path search in Cartesian space. However, construction sites often have a lot of dust and obstructions, making pre-modeling difficult and lacking adaptability to environmental changes.

[0003] In determining the target installation pose constraints, existing methods often assume the installation surface is an ideal plane or that pose parameters are obtained through manual measurement, failing to automatically extract the true geometry and pose information of the installation reference surface from the actual collected environmental point cloud data. This approach, detached from the actual on-site conditions, easily leads to deviations between the planned theoretical target pose and the executable pose on-site, affecting the installation accuracy and fit of the components. Furthermore, during path search, if end-effector pose planning is performed only in Cartesian space, it is necessary to repeatedly solve inverse kinematics to verify joint space feasibility, increasing computational overhead and potentially causing sudden joint velocity changes near singular configurations, resulting in discontinuous end-effector motion. Conversely, when directly performing joint space planning, existing methods struggle to effectively incorporate the end-effector pose constraints derived from the installation reference surface, easily generating path nodes that do not match the target installation requirements. In addition, the initial path obtained through the search is usually composed of straight line segments, resulting in an insufficiently smooth trajectory. Direct execution of this path can cause impacts and vibrations to the robot, reducing the stability of the installation process. How to efficiently generate a smooth trajectory that satisfies the end-effector posture constraints directly in the joint space, while taking into account the constraints of the actual installation reference plane, is a problem that needs to be solved in the current planning of installation trajectories for building components. Summary of the Invention

[0004] This invention aims to provide a trajectory planning method for building robots used in the installation of building components. Under unstructured site conditions, it can automatically extract the installation reference surface from a 3D point cloud and construct accurate spatial pose constraints. Then, it can directly plan a collision-free and smooth end effector motion trajectory in the joint space to solve the problems of inaccurate installation pose constraints and unsmooth motion trajectories in traditional methods, thereby improving the accuracy of building component installation and the stability of robot motion.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a trajectory planning method for construction robots used in the installation of building components. This method includes the following steps: acquiring three-dimensional point cloud data of the work space at the construction component installation site, and identifying the installation reference plane at the target installation location of the building component based on the three-dimensional point cloud data; determining the spatial pose constraints that the building component must satisfy during installation based on the installation reference plane; constructing a path search space corresponding to a collision-free motion path from the initial grasping pose of the building component to the target installation pose based on the spatial pose constraints; performing path search in the path search space using a fast expanding random tree algorithm to obtain an initial motion path; and smoothing the initial motion path to obtain the target trajectory planning result for the end effector of the construction robot. Through the above process, the installation reference features can be autonomously identified for the actual work site, and a smooth and usable installation trajectory can be planned under the premise of fully considering the end effector pose constraints and obstacle environment, ensuring the accuracy and safety of component installation and improving the adaptability of the construction robot to complex installation tasks.

[0006] As a technical solution of this invention, the process of acquiring three-dimensional point cloud data of the work space at the construction component installation site and identifying the installation reference surface specifically involves: using a depth camera installed at the end effector of a construction robot to collect three-dimensional point cloud data of the work space surrounding the target installation location of the construction component; performing planar fitting processing on the three-dimensional point cloud data of the work space to obtain multiple candidate planes; and selecting the candidate plane with the smallest angle between the normal vector and the gravity direction from among the multiple candidate planes as the installation reference surface. This accurately extracts a reference surface that meets the actual installation requirements, providing a reliable basis for subsequent pose constraint construction and reducing manual calibration work.

[0007] Based on this technical solution, the process of determining spatial pose constraints according to the installation reference plane includes: determining the end effector attitude constraint range of the building component in the target installation pose based on the normal vector of the installation reference plane; this end effector attitude constraint range is the allowable attitude angle range of the end effector of the construction robot at the target installation position; simultaneously, combining the geometric dimensions of the building component and the boundary of the installation reference plane, determining the end effector position constraint range of the building component in the target installation pose; this end effector position constraint range is the allowable position coordinate range of the end effector of the construction robot at the target installation position. Thus, the installation process requirements are quantified into the spatial pose constraint range of the robot end effector, enabling subsequent motion planning to directly match the on-site installation accuracy requirements.

[0008] Preferably, the path search space for constructing a collision-free motion path includes: acquiring a static 3D model of obstacles at the construction component installation site; determining the joint angle limits of each joint of the construction robot during its movement based on the kinematic model of the construction robot; sampling in the joint space of the construction robot, combining the static obstacle 3D model, joint angle limits, and spatial pose constraints, to obtain multiple collision-free joint space nodes; and constructing the path search space from these collision-free joint space nodes. Specifically, during sampling, candidate joint angle combinations are randomly generated within the joint angle limits of each joint of the construction robot; based on the forward kinematic model of the construction robot, the candidate joint angle combinations are converted into candidate poses of the end effector in Cartesian space; it is determined whether the candidate pose satisfies the spatial pose constraints, and whether the construction robot collides with the static obstacle 3D model under the candidate joint angle combination; if the candidate pose satisfies the spatial pose constraints and does not collide, the candidate joint angle combination is marked as a collision-free joint space node. By using the above method, environmental obstacle information and end-mount constraints are uniformly mapped into the joint space to form an effective sampling area. This facilitates the rapid expansion of the random tree algorithm to efficiently complete path exploration within the constrained joint space, reducing invalid sampling and collision risks.

[0009] For the path search of the fast expanding random tree algorithm, a preferred implementation is as follows: The joint angles of the construction robot in its initial grasping pose are set as the root node of the fast expanding random tree; collision-free joint space nodes in the path search space are used as the expansion targets of the fast expanding random tree; random sampling is performed in the joint space to generate random sampling nodes; the node closest to the random sampling node is selected from the current nodes of the fast expanding random tree as the nearest node; a preset step size is extended from the nearest node towards the random sampling node to generate a new node; if the joint motion trajectory corresponding to the line connecting the new node and the nearest node does not collide with static obstacles, the new node is added to the fast expanding random tree; the above random sampling and expansion steps are repeated until the nodes in the fast expanding random tree reach the joint angles of the construction component in the target installation pose, and the path from the root node to the node corresponding to the target installation pose in the fast expanding random tree is used as the initial motion path. During the expansion process, to reliably detect collisions, the preferred method is to calculate the joint space distance between the randomly sampled node and the nearest node; determine the number of interpolation steps based on the ratio of the joint space distance to a preset step size; perform linear interpolation between the nearest node and the randomly sampled node according to the number of interpolation steps to generate an intermediate interpolation node sequence; sequentially determine whether the pose of the construction robot corresponding to each interpolation node in the sequence collides with static obstacles, and only add the new node to the fast expansion random tree if no collision occurs. This gives the path search reliable obstacle avoidance capabilities, and the generated initial motion path can effectively bypass obstacles on site while extending towards the target installation pose.

[0010] The preferred scheme for smoothing the initial motion path is as follows: Perform cubic spline interpolation on the joint angle sequence in the initial motion path to obtain the initial joint trajectory where the joint angles change continuously over time; adjust the initial joint trajectory in time according to the angular velocity and angular acceleration limits of each joint of the construction robot to obtain the smooth motion trajectory of each joint; perform trajectory synchronization processing on the smooth motion trajectory of each joint of the construction robot in a time synchronization manner to obtain the target trajectory planning result. When performing cubic spline interpolation on the joint angle sequence: arrange the path nodes in the initial motion path according to the path order to obtain the joint angle sequence; assign an initial time interval between every two adjacent joint angles in the joint angle sequence, the initial time interval being proportional to the joint space distance between the two adjacent joint angles; construct a cubic spline interpolation function using each joint angle in the joint angle sequence as an interpolation node and the initial time interval as the interpolation interval length; generate the initial joint trajectory where the joint angles change continuously over time using the cubic spline interpolation function. During time-optimized adjustment, the angular velocity and angular acceleration of each joint at each moment are calculated based on the initial joint trajectory. The angular velocity of each joint is compared with its corresponding angular velocity limit, and the angular acceleration of each joint is compared with its corresponding angular acceleration limit. If the angular velocity or angular acceleration of a joint exceeds its limit, the initial time interval corresponding to that joint is increased, and cubic spline interpolation is performed again until both the angular velocity and angular acceleration of each joint do not exceed their respective limits. Through this smoothing and time synchronization process, jitter and discontinuities in the path are effectively eliminated. A kinematically feasible continuous trajectory is generated while meeting robot dynamics constraints, ensuring smooth installation movements and accurate positioning. Simultaneously, the overall motion time is shortened, improving the efficiency and quality of building component installation.

[0011] The technical effects and advantages provided by the present invention in the above technical solution are as follows: By acquiring 3D point cloud data of the workspace using a depth camera mounted on the end effector of a construction robot, planar fitting is performed on the point cloud to obtain multiple candidate planes. The candidate plane with the smallest angle between its normal vector and the direction of gravity is selected as the installation reference plane. Based on the normal vector of this installation reference plane, the allowable range of attitude angles of the end effector in the target installation pose is determined, forming an end effector attitude constraint interval. Simultaneously, using the geometric dimensions of the building components and the boundary of the installation reference plane, the allowable range of position coordinates of the end effector at the target installation position is determined, forming an end effector position constraint interval. This method of obtaining the installation reference plane based on real-world perception data allows the spatial pose constraints to directly reflect the actual conditions of the construction site, avoiding deviations caused by manual calibration or ideal assumptions. This ensures that the planned target pose can adapt to the actual tilt and positional changes of the installation surface, improving the fitting accuracy between the component and the installation surface.

[0012] When constructing the path search space, a 3D model of static obstacles at the construction component installation site is obtained. Based on the kinematic model of the construction robot, the joint angle limits for each joint are determined. Then, based on the static obstacle model, joint angle limits, and spatial pose constraints determined by the installation reference plane, sampling is directly performed in the construction robot's joint space. Specifically, candidate joint angle combinations are randomly generated within the limits of each joint. These combinations are converted into candidate poses in Cartesian space through forward kinematics. It is then determined whether the pose simultaneously satisfies the end-effector posture constraint range and the end-effector position constraint range, and whether the robot does not collide with the static obstacle under this joint angle combination. If so, the joint angle combination is marked as a collision-free joint space node, thus forming the path search space. This approach of directly generating collision-free nodes that satisfy the end-effector pose constraints in the joint space eliminates the need for repeated inverse kinematics solutions in subsequent path planning, avoiding path infeasibility problems caused by non-unique inverse solutions or singular configurations. It also ensures that nodes within the search space inherently possess joint space feasibility. A fast expanding random tree algorithm is employed to plan the initial motion path in the joint space search space. The joint angles of the initial grasping pose are used as the root node, and collision-free nodes in the space are the expansion targets. Through random sampling, nearest node selection, and expansion of new nodes with a preset step size, and by generating an intermediate node sequence through linear interpolation during the expansion process for continuous collision detection, it is ensured that each expansion step does not interfere with obstacles, ultimately resulting in a collision-free initial path in the joint space from the initial grasping pose to the target mounting pose. Based on this, cubic spline interpolation is applied to the joint angle sequence in the initial path, assigning an initial time interval proportional to the joint space distance to adjacent joint angles. A cubic spline interpolation function is constructed to generate the initial joint trajectory where the joint angles change continuously over time. Subsequently, based on the angular velocity and angular acceleration limits of each joint, the initial joint trajectory is time-optimally adjusted. When the angular velocity or angular acceleration of a joint exceeds the limit, the corresponding time interval for that joint is increased and re-interpolated until the motion parameters of all joints are within the limit range. Finally, the smooth motion trajectories of each joint are synchronized in a time-synchronized manner to obtain the final target trajectory planning result. The smoothed and time-optimized trajectory ensures continuous joint movement with no abrupt changes in angular velocity and angular acceleration. This guarantees a collision-free environment and allows the construction robot to operate smoothly during installation, reducing component positional shifts caused by motion impacts and further ensuring installation quality. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0014] Figure 1 This is a flowchart of a trajectory planning method for construction robots used in the installation of building components; Figure 2 This is a flowchart illustrating the method for acquiring 3D point cloud data of the workspace and identifying the installation reference surface; Figure 3 This is a flowchart for determining the installation posture constraint range of building components; Figure 4 This is a graph showing the change in the number of points within the random sampling consensus algorithm; Figure 5 This is a comparison chart of the smooth trajectories of joint angles. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of 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, 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.

[0016] See Figure 1 This invention provides a trajectory planning method for construction robots for the installation of building components. The method includes: acquiring three-dimensional point cloud data of the work space at the construction component installation site, and identifying the installation reference plane at the target installation location of the construction component based on the three-dimensional point cloud data of the work space; determining the spatial pose constraints of the construction component during the installation process based on the installation reference plane; constructing a path search space for a collision-free motion path of the construction component from the initial grasping pose to the target installation pose based on the spatial pose constraints; performing path search in the path search space using a fast expanding random tree algorithm to obtain an initial motion path; and performing path smoothing processing on the initial motion path to obtain the target trajectory planning result of the end effector of the construction robot.

[0017] In specific implementation, please refer to Figure 2 The process of acquiring three-dimensional point cloud data of the work space at the construction component installation site and identifying the installation reference surface at the target location of the construction component based on the three-dimensional point cloud data of the work space includes the following implementation methods.

[0018] A depth camera mounted on the end effector of a construction robot acquires 3D point cloud data of the workspace surrounding the target installation location of a building component. The construction robot controls the end effector to move to a predetermined acquisition position in front of the target installation location. The distance between the predetermined acquisition position and the target installation location is set according to the field of view and depth parameters of the depth camera. A time-of-flight (TOF) depth camera is used, with a depth image acquisition resolution of 640 pixels × 480 pixels and a depth range of 0.5 meters to 8.0 meters. The TOF depth camera acquires a depth image frame at the predetermined acquisition position. Using camera intrinsic parameters, the 2D image coordinates and depth values ​​of each pixel in the depth image are converted into 3D spatial point coordinates, generating a frame of 3D point cloud data for the workspace. Preprocessing is performed on the generated 3D point cloud data, including voxel grid downsampling and statistical outlier removal. In the voxel grid downsampling process, the cube side length of the voxel grid is set to 0.01 meters. The centroid of all 3D spatial points falling within the same cubic grid is calculated, and these centroids are used to replace all 3D spatial points within the grid, resulting in the downsampled 3D point cloud data of the workspace. In the statistical outlier removal process, for each 3D spatial point in the downsampled workspace 3D point cloud data, the average distance from that point to its K nearest neighbors is calculated. K is set to 50, and assuming the average distance follows a Gaussian distribution, the mean μ and standard deviation σ of the average distances of all 3D spatial points are calculated. 3D spatial points with average distances exceeding μ + ασ are considered outliers and removed. α is set to 1.0, resulting in the filtered workspace 3D point cloud data. This filtered workspace 3D point cloud data is used for subsequent plane fitting processing.

[0019] The filtered 3D point cloud data of the workspace is subjected to plane fitting processing to obtain multiple candidate planes. The plane fitting processing employs a multi-plane iterative extraction strategy based on the random sample consensus algorithm. In one random sample consensus plane fitting process, three non-collinear 3D spatial points are randomly selected from the filtered 3D point cloud data of the workspace. A plane model is determined by the coordinates of these three 3D spatial points, and the plane model is defined by the plane equations. It means that among them These are the three components of the normal vector of the planar model. This refers to the intercept parameter of the planar model. The planar distance from each 3D point in the filtered workspace 3D point cloud data to the planar model is calculated using the following method: ,in This represents the first point in the filtered 3D point cloud data of the workspace. The three-dimensional coordinates of each three-dimensional point are determined. Three-dimensional points with a planar distance less than a threshold are marked as interior points of the planar model, and the number of interior points is counted. The planar distance threshold is set to 0.02 meters, based on the fact that the depth measurement repeatability of a time-of-flight depth camera at a distance of 1 meter is approximately 0.01 meters; twice the depth measurement repeatability is used as the planar distance threshold. The above random sampling, planar model establishment, and interior point counting steps are repeated. The number of repetitions is determined by the random sampling consensus algorithm theory based on the expected confidence level and the estimated interior point ratio. The number of repetitions is set to 1000 times, corresponding to an expected confidence level of 99% and an estimated interior point ratio of 10%. The planar model with the most interior points in each iteration, and whose interior point number is greater than the minimum interior point number threshold, is selected as a valid candidate plane. The minimum interior point number threshold is set to 10% of the total number of three-dimensional points contained in the filtered work space three-dimensional point cloud data. The planar model parameters and their normal vectors of the valid candidate planes are recorded. Then, all three-dimensional points classified as interior points of the valid candidate plane are removed from the filtered work space three-dimensional point cloud data to obtain the remaining point cloud data. The above random sampling consistency plane fitting and point cloud removal process is performed again on the remaining point cloud data to extract the next valid candidate plane. This process is repeated until the number of three-dimensional spatial points in the remaining point cloud data is less than 5% of the total number of three-dimensional point cloud data in the filtered working space. The iteration is then terminated to obtain all the extracted valid candidate planes, i.e., multiple candidate planes.

[0020] After obtaining multiple candidate planes, the candidate plane with the smallest angle between its normal vector and the gravity direction is selected as the installation reference plane. The gravity direction is obtained through an inertial measurement unit fixed to the base of the construction robot. The inertial measurement unit outputs the three-axis components of gravitational acceleration in the base coordinate system of the construction robot in real time, denoted as... ,in This represents the component of gravitational acceleration output by the inertial measurement unit along the X-axis of the base coordinate system. This represents the component of gravitational acceleration output by the inertial measurement unit along the Y-axis in the base coordinate system. This represents the component of gravitational acceleration output by the inertial measurement unit (IMU) along the Z-axis of the base coordinate system. Since the construction robot is placed on a horizontal surface, the direction of gravitational acceleration is approximately along the negative Z-axis of the base coordinate system. By reversing the direction of the gravitational acceleration component output by the IMU and normalizing it, we obtain the unit gravity direction vector. The expression for calculating each component of the unit gravity direction vector is:

[0021] in, Let X be the component of the unit gravity direction vector along the X-axis of the base coordinate system. Let Y be the component of the unit gravity direction vector along the Y-axis in the base coordinate system. The denominator represents the component of the unit gravity direction vector along the Z-axis in the base coordinate system. This is the magnitude of the gravitational acceleration vector output by the inertial measurement unit.

[0022] For each candidate plane, the unit normal vector of the candidate plane is determined by the corresponding effective candidate plane model parameters, denoted as […]. ,in Let be the component of the candidate plane unit normal vector along the X-axis in the base coordinate system. Let be the component of the candidate plane unit normal vector along the Y-axis in the base coordinate system. Let be the component of the candidate plane's unit normal vector along the Z-axis in the base coordinate system. Calculate the angle between the candidate plane's unit normal vector and the unit gravity direction vector. The calculation formula is:

[0023] in, This represents taking the absolute value, where the numerator is the absolute value of the dot product of the unit normal vector of the candidate plane and the unit gravity direction vector, and the denominator contains... The magnitude of the unit normal vector of the candidate plane is always equal to 1. The magnitude of the unit gravity direction vector is always equal to 1.

[0024] Calculate the included angles of all candidate planes. Then, select The candidate plane with the smallest numerical value is used as the installation reference plane. If there are multiple candidate planes, the included angle... If the angles are equal and both have the smallest included angle, then the candidate plane with the most interior points during the random sampling consistency fitting process is selected as the installation reference plane.

[0025] See Figure 4 In the figure, the horizontal axis represents the number of random sampling iterations, ranging from 0 to 1000; the vertical axis represents the number of interior points obtained in each random sampling, in units of individual points. The blue solid line curve depicts the changing trend of the number of interior points in each random sampling. This curve shows a relatively stable fluctuation, with the number of interior points mostly concentrated between 100 and 300, occasionally showing brief peaks in extreme values, indicating that there are natural fluctuations in the identification of interior points in the point cloud plane during the random sampling process. The red dashed line is the minimum interior point number threshold line, with a value of approximately 100, representing the lower limit of the interior point number calculated based on 10% of the total number of points in the filtered 3D point cloud set in Example 1. All planar models below this threshold are judged as invalid. The green dotted line is the marker line for the maximum number of interior points of the effective candidate plane, with a value of approximately 840, representing the maximum number of effective interior points detected in all iterations, serving as one of the criteria for selecting the optimal mounting reference plane.

[0026] The data illustrated reflects the statistical distribution characteristics of the number of interior points when performing planar fitting on the filtered 3D point cloud data of the workspace based on the random sampling consensus algorithm in Example 1. Planar models are continuously generated during the random sampling process. The fluctuations in the blue interior point count curve reflect the differences in the number of interior points corresponding to different candidate planar models. Furthermore, the number of interior points in all samples is higher than the set minimum interior point count threshold (red line), ensuring the effectiveness of the fitted plane. The iterative result where the number of interior points reaches the green maximum value line is the effective candidate plane with the most interior points. This plane is further used as the preferred target for the installation reference plane, and the final determination is made based on the principle of minimum included angle.

[0027] In specific implementation, please refer to Figure 3 The process of determining the spatial posture constraints of building components during installation based on the installation reference plane includes two parts: determining the end posture constraint range and determining the end position constraint range.

[0028] Based on the normal vector of the mounting reference plane, the end effector attitude constraint range of the building component in the target mounting pose is determined. The end effector attitude constraint range is the range of allowed attitude angles of the building robot's end effector at the target mounting position. In some embodiments, the nominal end effector attitude of the building component in the target mounting pose is calculated from the normal vector of the mounting reference plane. The unit normal vector of the mounting reference plane in the building robot's base coordinate system is denoted as... ,in The component of the unit normal vector along the X-axis of the base coordinate system. The component of the unit normal vector along the Y-axis in the base coordinate system. Let be the component of the unit normal vector along the Z-axis of the base coordinate system. The direction of the bottom surface of a building component towards the installation reference plane is defined as opposite to the direction of the normal vector of the installation reference plane. Therefore, the bottom normal vector of a building component in its correct installation state should be . The direction of this vector is taken as the Z-axis direction of the end effector's attitude coordinate system for the construction robot. A side reference edge of the construction component is selected to determine the Y-axis direction. This side reference edge is specified by the geometric features of the construction component, such as the direction of the long side of a rectangular component, obtained from the baseline at the installation site or the edge direction of an adjacent installed component. The end effector's attitude coordinate system is represented by ZYX Euler angles, where the three Euler angles are the roll angles about the X-axis of the base coordinate system. Pitch angle about the Y-axis of the base coordinate system and yaw angle about the Z-axis of the base coordinate system By converting the direction vector to Euler angles, the nominal roll angle corresponding to the nominal end-effector attitude is obtained. nominal pitch angle and nominal yaw angle .

[0029] After determining the nominal end-effector attitude, the allowable range of attitude angles for the construction robot's end effector at the target installation position is set, i.e., the end-effector attitude constraint range. The end-effector attitude constraint range is described by the permissible deviation range around the nominal attitude, and the roll angle constraint range is... The pitch angle constraint range is The constraint range for the yaw angle is: Among them, the deviation threshold , and The setting method considers the assembly tolerance between the building component and the mounting reference plane, as well as the geometric dimensions of the building component. For the two attitude angle deviations (i.e., roll and pitch angle deviations) related to the normal vector of the mounting reference plane, their permissible deviation angles are calculated using the maximum allowable clearance between the bottom surface of the component and the mounting reference plane, and the diagonal length of the building component's projection in a plane parallel to the mounting reference plane. Optionally, the upper limit of the permissible angle deviation is determined based on the geometric relationship using the following formula:

[0030] in, This indicates the maximum allowable gap distance between the bottom edge of a building component and the mounting reference surface. The installation gap is set according to the installation accuracy requirements of building components, and is generally taken as the allowable installation gap value specified in the building component installation specification, for example, 0.005 meters; This represents the length of the diagonal of the rectangle circumscribed on the base of a building component. Calculated directly from the geometric dimensions of the building components, using the longer side of the circumscribed rectangle. and short side dimensions Substitution get, and All units are meters. Roll angle deviation threshold. and pitch angle deviation threshold The calculation result is used to limit the bottom surface contact angle. Yaw angle deviation threshold. The rotational positioning tolerance of the building component around the normal vector of the installation reference plane is set, and is generally taken as a fixed angle value, such as... This corresponds to the allowable deviation of the azimuth angle of component installation during building construction.

[0031] After obtaining the constraint ranges for roll, pitch, and yaw angles, the combined angle range is used as the end effector attitude constraint range. The attitude angles of the construction robot's end effector at the target installation position must fall within the above range to meet the installation fit requirements.

[0032] Based on the geometric dimensions of the building component and the boundary of the mounting reference surface, the end-effector position constraint range of the building component in the target installation pose is determined. The end-effector position constraint range is the range of position coordinates allowed for the end effector of the construction robot at the target installation position. In some embodiments, the boundary information of the mounting reference surface is obtained. The boundary of the mounting reference surface is represented by the convex hull or polygon boundary of the point cloud data corresponding to the mounting reference surface identified in Embodiment 1 on the reference surface plane. The convex hull boundary is projected onto the two-dimensional local coordinate system where the mounting reference surface is located. The local coordinate system has a point on the mounting reference surface as the origin and two orthogonal directions within the mounting reference surface as the horizontal axis u and the vertical axis v. The bottom geometric contour of the building component is obtained. The bottom geometric contour is the projected shape on the mounting reference surface plane, and the length of the circumscribed rectangle of the bottom geometric contour is... Width is During installation, the bottom surface of the building component must be completely within the boundary of the installation reference plane, and an edge gap must be maintained between the edge of the bottom surface of the building component and the boundary of the installation reference plane. , A value of 0.01 meters is used to ensure that the component does not exceed the boundary of the reference plane and to prevent edge collisions. Based on this, the allowable position range of the center point of the bottom surface of the building component in the local coordinate system uv plane is calculated.

[0033] The polygon of the installation reference plane boundary is offset inwards at equal intervals, with an offset distance of [missing value]. or The corresponding half-size is selected based on the offset direction. Specifically: for a boundary parallel to the horizontal axis u, the offset distance is... For a boundary parallel to the vertical axis v, the offset distance is... The interior region of the polygon obtained after offsetting represents the permissible position area of ​​the center point of the bottom surface of the building component in the local coordinate system uv plane. The upper bound of this permissible position area in the u direction is... The lower bound is The upper bound in the v direction is The lower bound is By combining the position and orientation of the mounting reference plane in the base coordinate system, the position region in the UV coordinate system is transformed to the base coordinate system, resulting in the position coordinate constraint range of the construction robot's end effector on the X, Y, and Z axes of the base coordinate system. Specifically, the coordinates along the normal vector direction of the mounting reference plane are limited by the plane position of the mounting reference plane, while the height coordinates are allowed only slight variations. The center of the height coordinate range is the coordinate value determined by the plane equation of the mounting reference plane in the base coordinate system, and the upper and lower deviations are taken as the clearance. Values ​​of the same order of magnitude, for example Meters. Ultimately, the end-effector position constraint interval consists of the coordinate range of the three axes of the base coordinate system. The position coordinates of the construction robot's end effector at the target installation position must fall within the range of the above three axes.

[0034] In practice, a static 3D model of obstacles at the building component installation site is acquired. This static obstacle 3D model is obtained through a pre-imported Building Information Model (BIM), which contains the 3D geometric representations of existing walls, columns, floors, and other fixed equipment at the installation site. The static obstacle 3D model is stored using a triangular mesh model. Each triangular facet in the mesh model consists of three vertex coordinates and a normal vector. The 3D geometric representation of each obstacle includes its position coordinates and dimensions in the building's base coordinate system. This static obstacle 3D model is used as an impassable environmental constraint in subsequent collision detection.

[0035] Based on the kinematic model of the construction robot, the joint angle limits of each joint during its movement are determined. The kinematic model of the construction robot is a link transformation model established based on the standard DH parameter method. The DH parameter table includes the link length, link offset, joint torsion angle, and joint angle variables for each joint. The joint angle limits of each joint of the construction robot include mechanical limit angles and soft limit angles. The mechanical limit angles are the maximum and minimum angles allowed by the joint's physical structure. The soft limit angles are the angle working range set by the control system after reducing the mechanical limit angles by a certain safety margin. The safety margin is 5% of the mechanical limit range to prevent collisions caused by the joints moving to their physical limit positions. The joint angle limits of each joint are represented as intervals. ,in For joint indexing, , This represents the total number of joints in a construction robot. For the first The minimum value of each joint angle limit range For the first The maximum value of each joint angle limit range. For each joint, Take the lower boundary value of the soft limit. Take the upper boundary value of the soft limit, the lower boundary value of the soft limit is the lower boundary value of the mechanical limit plus 5% of the mechanical limit range, and the upper boundary value of the soft limit is the upper boundary value of the mechanical limit minus 5% of the mechanical limit range.

[0036] Based on the static obstacle 3D model, joint angle limits, and spatial pose constraints, sampling is performed in the joint space of the construction robot to obtain multiple collision-free joint space nodes. The sampling process is performed iteratively, and each iteration includes the following processing.

[0037] Within the joint angle limits of each joint of the construction robot, candidate joint angle combinations are randomly generated. Random generation uses a uniform sampling method. The randomly generated candidate joint angle combinations are represented as a vector. ,in For the first In the candidate joint angle combinations of the nth sampling, the th The angle values ​​of each joint. This represents the vector transpose operation. Each component is calculated one by one using the following formula:

[0038] in, In order to target the The first sampling Uniformly distributed random numbers for each joint. The range of values ​​for is a closed interval. It is generated independently in each sample using the MersenneTwister pseudo-random number generator. For the first The minimum value of each joint angle limit range For the first The maximum value of each joint angle limit interval, and both are pre-set fixed constants. The above formula ensures that each joint angle value within the candidate joint angle combination is within the joint angle limit interval of the corresponding joint.

[0039] Based on the forward kinematics model of the construction robot, the candidate joint angles are combined. This is converted into the candidate pose of the construction robot's end effector in Cartesian space. The calculation process of the forward kinematics model is as follows: Substituting the angle values ​​of each joint into the homogeneous transformation matrices established by the DH parameters, and multiplying them along the kinematic chain up to the last link, we obtain the homogeneous transformation matrix of the end effector coordinate system relative to the building base coordinate system, from which the position coordinates are extracted. Attitude representation. Attitude representation uses ZYX Euler angles to extract the roll angle. Pitch angle and yaw angle Location coordinates Euler angles Together they constitute the candidate pose.

[0040] Determine whether the candidate pose satisfies the spatial pose constraints. The spatial pose constraints consist of the end-effector position constraint interval and the end-effector attitude constraint interval determined in Example 2. The end-effector position constraint interval includes the allowable interval of the X-axis coordinate in the base coordinate system. Y-axis coordinate allowable range and Z-axis coordinate allowable range The end attitude constraint range includes the roll angle allowance range. Pitch angle allowable range and yaw angle allowable range When the candidate pose simultaneously satisfies , , , , as well as When considering the membership conditions of these six intervals, the candidate pose is determined to satisfy the spatial pose constraints. If any interval membership condition is not met, the candidate pose is determined to not satisfy the spatial pose constraints, and further processing of the current candidate joint angle combination is abandoned, directly proceeding to the next iteration.

[0041] When the candidate poses satisfy the spatial pose constraints, further determine the construction robot's position in the candidate joint angle combinations. The collision detection process involves checking whether the link of the construction robot collides with the 3D model of the static obstacle. The simplified geometric model of each link is represented by a cylindrical bounding body. Each link's cylindrical bounding body consists of its central axis segment and radius. The radius of the cylindrical bounding body is the maximum outer contour radius of the link, and the central axis segment is determined by the spatial position of the two joints connected to the link under forward kinematics. The 3D model of the static obstacle is represented by a triangular mesh set. The interference check uses a separation axis theorem algorithm based on hierarchical bounding boxes. Specifically, it first calculates whether the cylindrical bounding body of each link intersects with the axis-aligned bounding box of the static obstacle's triangular mesh. If they do not intersect, the link is determined not to collide with the 3D model of the static obstacle. If the axis-aligned bounding boxes intersect, the cylindrical bounding body is further converted into a polygon approximation body, and a precise separation axis test is performed with the triangular facets in the 3D model of the static obstacle. The construction robot is judged to be in a good position in the candidate joint angle combination if and only if the cylindrical enclosure of all links does not collide with the static obstacle 3D model. No collision occurs. If any link collides with the 3D model of a static obstacle, a collision is determined, and the candidate joint angle combination is discarded.

[0042] If the candidate pose satisfies the spatial pose constraints and the construction robot does not collide under the candidate joint angle combination, then the candidate joint angle combination will be... Joint space nodes are marked as collision-free. Collision-free joint space nodes are stored in a data structure containing a vector of candidate joint angle combinations. The corresponding end effector pose is obtained from the orthogonal kinematics calculation.

[0043] Repeat the above iterative process, generating a new candidate joint angle combination in each iteration and performing constraint condition judgment and collision detection until the number of collision-free joint space nodes collected reaches a preset number. The preset number is set according to the dimension of the construction robot's joint space and the path search complexity of the installation task; for a six-DOF construction robot, the preset number is set to 8000. All collected collision-free joint space nodes together constitute the path search space, which is stored in the path planner as an unordered set of points for subsequent fast-expanding random tree algorithm to perform random sampling during the search phase.

[0044] In practice, the process of using the fast expanding random tree algorithm to search for paths in the path search space and obtain the initial motion path includes operations such as root node initialization, iterative sampling expansion, and path extraction.

[0045] The joint angles of the construction robot in its initial grasping pose are set as the root node of a fast-expanding random tree. The initial grasping pose is the starting posture of the construction robot after grasping the building component, ready to begin the installation motion; the corresponding joint angle vector is denoted as... ,in This indicates the total number of joints in a construction robot. Indicates the first node in the root node The angle values ​​of each joint. , This represents the vector transpose operation. The joint angle vector and node indices are then stored in the node set of the fast expanding random tree.

[0046] Using collision-free joint space nodes in the path search space as the expansion targets for the fast-expanding random tree, random sampling is performed in the joint space to generate random sampling nodes. In each iteration, the random sampling nodes are generated as follows: a node is generated within a closed interval... Uniformly distributed random numbers ,like Less than the preset target bias probability Then the target joint angle combination of the building components in the target installation position. As a random sampling node ;like Greater than or equal to Then, within the joint angle limit range of each joint of the construction robot. Uniform random sampling within the area yields random sampling nodes. . Set as Target joint angle combination The end effector pose of the building component in the target installation pose is obtained through inverse kinematics. If multiple inverse kinematics solutions exist, the solution with the smallest spatial distance to the root node joint is selected as the final solution. .

[0047] From the current nodes of the fast-expanding random tree, select the node closest to the randomly sampled node as the nearest node. Calculate the distance between each node in the fast-expanding random tree and the randomly sampled node. The joint space distance is measured using Euclidean distance. For the first joint space distance in the tree... Nodes , and The distance calculation formula is:

[0048] in, Indicates the first The joint space Euclidean distance between each node and a randomly sampled node; Represents the first random sampling node Each joint angle value; Indicates the first The node of the first Each joint angle value; This represents the total number of joints in the construction robot. Traverse all nodes in the tree... Select the node with the minimum distance as the nearest node, denoted as . .

[0049] A new node is generated by expanding the node with a preset step size towards the randomly sampled node. The preset step size is denoted as... , The value is determined based on the joint motion granularity setting of the construction robot. For a six-DOF construction robot, Set as Radians. Calculate the nearest node. To random sampling node Joint space distance , The same Euclidean distance formula is used for calculation. If Then the new node Directly selected as random sampling nodes ;like The new node is calculated using the following formula:

[0050] in, This is the joint angle vector of the newly generated node. Let be the joint angle vector of the nearest node. For the joint angle vector of the randomly sampled nodes, To preset the step size, This represents the joint space distance between the nearest node and the randomly sampled node.

[0051] After generating a new node, it is necessary to determine whether the joint motion trajectory corresponding to the connection between the new node and the nearest node collides with a static obstacle. The collision detection process uses discretization interpolation, specifically including: calculating the joint space distance between a randomly sampled node and the nearest node. Based on joint space distance With preset step size The ratio determines the number of interpolation steps. , The calculation method is as follows ,in This indicates a round-up operation; at the nearest node With random sampling nodes According to the number of interpolation steps Perform linear interpolation to generate a sequence of intermediate interpolation nodes. ,in , .

[0052] Sequentially determine whether the pose of the construction robot corresponding to each interpolation node in the intermediate interpolation node sequence collides with a static obstacle. For each interpolation node... The poses of each link are calculated using the forward kinematics model of the construction robot. Interference checks are then performed between the cylindrical bounding boxes of each link and the 3D model of the static obstacle. The interference check employs an algorithm based on axis-aligned bounding boxes and the separation axis theorem. If the poses of the construction robot corresponding to all interpolated nodes do not collide with the 3D model of the static obstacle, then the joint motion trajectory corresponding to the entire link is determined not to collide with the static obstacle, and the new node is... Add nodes to the set of nodes in the fast-expanding random tree and record them. As The parent node. If any interpolation node collides, the new node is discarded, not added to the tree, and the process proceeds directly to the next iteration.

[0053] Repeat the above random sampling and expansion steps until a node in the rapidly expanded random tree is combined with the target joint angle. The distance between them is less than the preset connection threshold, and the node is... The joint motion trajectories corresponding to the connecting lines do not collide with static obstacles. The connection threshold is set to... radians. At this point, Add it to the tree as a child node of this node. From the root node Begin by tracing back along the parent node relationship to... Extract path node sequence This sequence of path nodes constitutes the initial motion path. The initial motion path is stored as a sequence of joint angles, containing combinations of joint angles along the path from the initial grasping pose to the target mounting pose.

[0054] In practice, the process of smoothing the initial motion path to obtain the target trajectory planning result of the end effector of the construction robot includes three stages: cubic spline interpolation, time-optimal adjustment, and trajectory synchronization.

[0055] Cubic spline interpolation is performed on the joint angle sequence in the initial motion path to obtain the initial joint trajectory in which the joint angles change continuously over time.

[0056] The path nodes in the initial motion path are arranged according to the path order to obtain the joint angle sequence. The initial motion path, obtained from Example 4, is a node sequence from the root node to the target joint angle combination, and the number of nodes is denoted as . ,Include Each motion segment. The sequence of nodes is represented as... ,in Corresponding to the root node, Corresponding target joint angle combination. For the first... 1 node , This represents the total number of joints in a construction robot. Indicates the first The node in the node The angle values ​​of each joint. , .

[0057] An initial time interval is assigned between every two adjacent joint angles in the joint angle sequence. This initial time interval is proportional to the joint space distance between the two adjacent joint angles. For the ... Nodes With the Nodes Between the movement segments, joint space distance The Euclidean distance is used for calculation, and the expression is:

[0058] in, Indicates the first The joint space distance corresponding to each movement segment For the first The node in the node The angle values ​​of each joint. For the first The node in the node The angle values ​​of each joint. This represents the total number of joints in a construction robot.

[0059] Initial time interval Assigned to , This is the distance-time ratio coefficient. The value is determined as follows: obtain the angular velocity limits of each joint of the construction robot, and select the minimum value as . ,Will Multiply by speed utilization factor Obtain the reference average angular velocity. Set as This ensures that the average running speed of the initial trajectory is within the middle range of the capabilities of each joint, leaving room for subsequent time adjustments. Then, based on the relationship... get In a set of settings, if Values radians per second Pick ,but Pick Seconds per radian. In this way, each motion segment obtains an initial time interval proportional to the joint space distance.

[0060] Using each joint angle in the joint angle sequence as an interpolation node and the initial time interval as the interpolation interval length, a cubic spline interpolation function is constructed. The time points corresponding to each interpolation node are obtained by accumulating the initial time intervals: Time Points , , Time point series This serves as the interpolation time node. For each joint of the construction robot... Construct a cubic spline interpolation function separately. , making , The cubic spline interpolation function in each subinterval The above is a cubic polynomial. The polynomials in adjacent subintervals satisfy the boundary conditions of continuous function values, continuous first derivatives, and continuous second derivatives at common nodes. Additional natural boundary conditions are set as follows: and By solving the three-moment equation system, the coefficients of the cubic polynomial in each subinterval are determined. The cubic spline interpolation functions of all joints constitute a continuous set of functions relating joint angles to time, which is the initial joint trajectory.

[0061] An initial joint trajectory, in which the joint angles change continuously over time, is generated using a cubic spline interpolation function. This initial joint trajectory is stored in the form of an analytical function and can be accessed at any given time. Calculate the angle values ​​of each joint and their first and second derivatives.

[0062] Based on the angular velocity and angular acceleration limits of each joint of the construction robot, the initial joint trajectory is adjusted in time to obtain a smooth motion trajectory for each joint. The angular velocity limits of each joint of the construction robot are denoted as... , indicating the first The maximum absolute value of the permissible angular velocity of each joint; the angular acceleration limit of each joint is denoted as... , indicating the first The absolute value of the maximum permissible angular acceleration for each joint. Calculate the angular velocity and angular acceleration of each joint at each moment corresponding to the initial joint trajectory. (The interval is missing in the original text.) With a fixed sampling period Discretize the time by seconds to obtain a densely sampled time series. ,in This represents the total number of sampling points. For each sampling time... and each joint Calculate angular velocity and angular acceleration .

[0063] The angular velocities of each joint are compared with their respective angular velocity limits, and the angular accelerations of each joint are compared with their respective angular acceleration limits. This process is repeated across all sampling times and all joints to determine the percentage of the maximum angular velocity exceeding the limit. and the proportion of maximum angular acceleration exceeding the limit The calculation method for the over-limit ratio is as follows: For angular velocity, if... Then calculate the ratio. For angular acceleration, if Then calculate the ratio. The maximum value among all out-of-limit ratios is taken as the global maximum out-of-limit factor. , If there are no exceedances, then .

[0064] like If at least one joint exceeds the limit in angular velocity or angular acceleration at some moment, then the initial time intervals for all motion segments are increased by multiplying all initial time intervals by a scaling factor. ,in , The margin factor is set to . To ensure sufficient margin after adjustment, update the new initial time interval to... The time points are re-accumulated and the cubic spline interpolation function is reconstructed to obtain the updated joint trajectory. The process of calculating the over-limit ratio and increasing the time interval is repeated until... That is, the angular velocity of each joint does not exceed the angular velocity limit and the angular acceleration of each joint does not exceed the angular acceleration limit. The resulting joint trajectories are smooth motion trajectories, and the total motion time of each smooth joint trajectory is... ,in This refers to the initial time interval of the motion segment, which is finally determined after time-optimized adjustment. Optionally, during the time-optimized adjustment process, the time interval of each motion segment is scaled uniformly, preserving the geometry of the original path and only changing the execution speed, thereby satisfying the driving capability constraints without changing the path safety.

[0065] Following a time synchronization approach, the smooth motion trajectories of each joint of the construction robot are processed for trajectory synchronization to obtain the target trajectory planning result. The purpose of time synchronization processing is to ensure that all joints complete their movements synchronously within the same total motion time. Because a uniform scaling is applied to each motion segment during the time-optimization adjustment phase, the total motion time of each joint is automatically kept consistent. In implementations requiring individual time-based fine-tuning of the trajectory of each joint, if the total motion time differs between different joints after adjustment, the maximum value of the adjusted motion time for each joint is taken as the unified synchronization time. Furthermore, the trajectory of each joint is stretched proportionally along the time axis according to a uniform synchronization time, so that the motion time of each joint is extended to... The specific stretching method is based on the time intervals of each movement segment determined at the end of the adjustment phase. Re-accumulate each time point Total time If the movement time of all joints is the same, then If there are differences, then ,in For the first Optimal adjustment of total time for each joint. The smooth motion trajectory will change the original time variable. Mapped to The cubic spline interpolation function is then resampled using the mapped time to generate a time length of [missing information]. The new trajectory. After time synchronization processing, the trajectories of all joints are synchronized at a unified time. The motion is completed synchronously, resulting in a final output consisting of a continuous function of the joint angles changing over time. This output is the target trajectory planning result for the end effector of the construction robot. The target trajectory planning result can be used for trajectory tracking execution by the construction robot controller, driving the robot's end effector to install building components along the planned path.

[0066] See Figure 5In the figure, the horizontal axis represents time in seconds, and the vertical axis represents the joint angle of a certain joint of the construction robot in radians. The blue dashed curve in the figure represents the initial joint trajectory generated by cubic spline interpolation in Example 4, and the red solid curve represents the smooth motion trajectory obtained after time-optimized adjustment and smoothing of the initial trajectory in Example 5.

[0067] As shown in the figure, the initial joint trajectory exhibits a relatively obvious continuous change within the time interval from 0 to approximately 2.3 seconds, with the angle gradually increasing from approximately 0.2 radians to a peak of approximately 1.28 radians. During this period, the curve shows some high-frequency jitter and non-smooth fluctuations, reflecting local oscillations in the unsmoothed trajectory. Within the time interval from 2.3 seconds to 3 seconds, the initial trajectory angle fluctuates slightly, generally remaining around 1.2 radians.

[0068] In contrast, the smooth trajectory achieves a rapid and stable angle increase within the time interval of 0 to 0.5 seconds, with the angle increasing uniformly from about 0.2 radians to 1.2 radians. In the subsequent time interval of 0.5 to 3 seconds, the angle of the smooth trajectory remains constant at 1.2 radians without significant fluctuations, demonstrating the stability and continuity of the trajectory after optimal time adjustment.

[0069] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A trajectory planning method for construction robots used in the installation of building components, characterized in that, The method includes: Acquire three-dimensional point cloud data of the work space at the construction component installation site, and identify the installation reference surface at the target location of the construction component installation based on the three-dimensional point cloud data of the work space. Based on the installation reference plane, determine the spatial orientation constraints of the building components during the installation process; Based on the spatial pose constraints, a path search space is constructed for the collision-free motion path of the building component from the initial grasping pose to the target installation pose. In the path search space, a fast expanding random tree algorithm is used to search for paths and obtain the initial motion path; The initial motion path is smoothed to obtain the target trajectory planning result of the end effector of the construction robot.

2. The method for trajectory planning of a construction robot for installing building components according to claim 1, characterized in that, The process of acquiring three-dimensional point cloud data of the work space at the building component installation site, and identifying the installation reference surface at the target location of the building component installation based on the three-dimensional point cloud data of the work space, includes: A depth camera installed at the end of a construction robot collects 3D point cloud data of the work space around the target location for installing building components. The three-dimensional point cloud data of the work space is subjected to plane fitting processing to obtain multiple candidate planes; The candidate plane with the smallest angle between the normal vector and the direction of gravity is selected from the plurality of candidate planes and used as the installation reference plane.

3. The method for trajectory planning of a construction robot for installing building components according to claim 1, characterized in that, The step of determining the spatial orientation constraints of building components during installation based on the installation reference plane includes: Based on the normal vector of the installation reference plane, the end attitude constraint range of the building component under the target installation pose is determined. The end attitude constraint range is the range of attitude angles allowed for the end effector of the building robot at the target installation position. Based on the geometric dimensions of the building component and the boundary of the installation reference surface, the end position constraint range of the building component in the target installation pose is determined. The end position constraint range is the range of position coordinates allowed for the end effector of the building robot at the target installation position.

4. The method for trajectory planning of a construction robot for installing building components according to claim 1, characterized in that, The step of constructing a collision-free motion path search space for the building component from its initial grasping pose to its target installation pose based on the spatial pose constraints includes: Obtain a 3D model of static obstacles at the construction component installation site; Based on the kinematic model of the construction robot, determine the joint angle limits of each joint during the movement of the construction robot; Based on the static obstacle 3D model, the joint angle limit, and the spatial pose constraint, sampling is performed in the joint space of the construction robot to obtain multiple collision-free joint space nodes. The path search space is composed of the multiple collision-free joint space nodes.

5. The method for trajectory planning of a construction robot for installing building components according to claim 4, characterized in that, Based on the static obstacle 3D model, the joint angle limits, and the spatial pose constraints, sampling is performed in the joint space of the construction robot to obtain multiple collision-free joint space nodes, including: Within the joint angle limit range of each joint of the construction robot, candidate joint angle combinations are randomly generated; Based on the forward kinematics model of the construction robot, the candidate joint angle combinations are converted into candidate poses of the construction robot end effector in Cartesian space. Determine whether the candidate pose satisfies the spatial pose constraint conditions, and whether the construction robot collides with the static obstacle 3D model under the candidate joint angle combination; If the candidate pose satisfies the spatial pose constraint and no collision occurs, then the candidate joint angle combination is marked as the collision-free joint spatial node.

6. The method for trajectory planning of a construction robot for installing building components according to claim 1, characterized in that, The initial motion path is obtained by using a fast expanding random tree algorithm in the path search space, including: Set the joint angles of the construction robot in its initial grasping pose to the root node of a rapidly expanding random tree; Using the collision-free joint space nodes in the path search space as the expansion targets of the fast-expanding random tree, random sampling is performed in the joint space to generate random sampling nodes; Select the node that is closest to the random sampling node from the current nodes of the fast-expanding random tree, and designate it as the nearest node; Expand a new node by a preset step size from the nearest node to the random sampling node. If the joint motion trajectory corresponding to the connection between the new node and the nearest node does not collide with a static obstacle, then add the new node to the fast-expanding random tree. Repeat the above random sampling and expansion steps until the nodes in the rapidly expanded random tree reach the joint angle of the building component in the target installation pose. The path from the root node to the node corresponding to the target installation pose in the rapidly expanded random tree is taken as the initial motion path.

7. The method for trajectory planning of a construction robot for installing building components according to claim 6, characterized in that, The step of expanding a new node from the nearest node to the randomly sampled node by a preset step size, and if the joint motion trajectory corresponding to the line connecting the new node and the nearest node does not collide with a static obstacle, then the new node is added to the fast-expanding random tree, including: Calculate the joint space distance between the random sampling node and the nearest node; The number of interpolation steps is determined based on the ratio of the joint space distance to the preset step length; Linear interpolation is performed between the nearest node and the random sampling node according to the number of interpolation steps to generate an intermediate interpolation node sequence. Sequentially determine whether the pose of the construction robot corresponding to each interpolation node in the intermediate interpolation node sequence collides with a static obstacle. If no collision occurs, add the new node to the fast expanding random tree.

8. The method for trajectory planning of a construction robot for installing building components according to claim 1, characterized in that, The process of smoothing the initial motion path to obtain the target trajectory planning result of the end effector of the construction robot includes: The joint angle sequence in the initial motion path is subjected to cubic spline interpolation to obtain the initial joint trajectory in which the joint angles change continuously over time. Based on the angular velocity and angular acceleration limits of each joint of the construction robot, the initial joint trajectory is adjusted in time to obtain a smooth motion trajectory for each joint. The smooth motion trajectories of each joint of the construction robot are synchronized using a time synchronization method to obtain the target trajectory planning result.

9. A method for trajectory planning of a construction robot for installing building components according to claim 8, characterized in that, The step of performing cubic spline interpolation on the joint angle sequence in the initial motion path to obtain the initial joint trajectory in which the joint angles change continuously over time includes: The path nodes in the initial motion path are arranged in the path order to obtain the joint angle sequence; An initial time interval is assigned between every two adjacent joint angles in the joint angle sequence, and the initial time interval is proportional to the joint space distance between the two adjacent joint angles; A cubic spline interpolation function is constructed using each joint angle in the joint angle sequence as an interpolation node and the initial time interval as the length of the interpolation interval. The initial joint trajectory, in which the joint angle changes continuously over time, is generated using the cubic spline interpolation function.

10. A method for trajectory planning of a construction robot for installing building components according to claim 8, characterized in that, The step of performing time-optimized adjustment on the initial joint trajectory based on the angular velocity and angular acceleration limits of each joint of the construction robot to obtain a smooth motion trajectory for each joint includes: Calculate the angular velocity and angular acceleration of each joint at each moment corresponding to the initial joint trajectory; Compare the angular velocity of each joint with the angular velocity limit of each joint, and compare the angular acceleration of each joint with the angular acceleration limit of each joint; If the angular velocity of a joint exceeds the angular velocity limit of that joint, or the angular acceleration of a joint exceeds the angular acceleration limit of that joint, then the initial time interval corresponding to that joint is increased, and cubic spline interpolation is performed again until the angular velocities of each joint do not exceed the angular velocity limit and the angular accelerations of each joint do not exceed the angular acceleration limit.

Images