A collaborative control method for drone-robotic arm system for precision spraying

Abstract

The present invention discloses a collaborative control method for a UAV-robotic arm system for precision spraying. The method realizes three-dimensional modeling of the target surface and semantic recognition of the spraying area through a depth camera and a point cloud reconstruction algorithm, and generates spraying path and spraying posture data. Based on the requirements of the spraying task, an empirical estimation model for the spraying film thickness is established to obtain the spraying parameters at the path point level. The load mass change, the center of mass change of the robotic arm and the spraying reaction force of each path point are calculated, and a coupled dynamic model is constructed in combination with the system state. The dynamic compensation control input is generated by using an adaptive sliding mode control algorithm. After the spraying is completed, the actual spraying effect is collected through image and point cloud detection, and the color, film thickness and texture features are extracted. After comparison with the task requirements, error feedback is generated, and the control parameters are adjusted online. The present invention effectively solves the problems of dynamic changes and error compensation in the spraying process, and is suitable for high-precision spraying tasks in complex environments.

Description

Technical Field

[0001] The present invention belongs to the technical field of spraying UAV control, and specifically relates to a UAV-mechanical arm system collaborative control method for precision spraying. Background Art

[0002] With the advancement of industrial automation, spray painting operations are gradually evolving from traditional manual operations to intelligent, unmanned operations. This is particularly true in areas such as architectural decoration, precision manufacturing, and surface treatment of complex structures, which place higher demands on spray painting quality, consistency, and efficiency. Drones equipped with multi-degree-of-freedom robotic arms, with their flexibility and wide coverage, offer promising applications in spray painting tasks in complex environments.

[0003] Some existing solutions have attempted to integrate multi-rotor drones with multi-jointed robotic arms to achieve automated spraying of specific target areas. Most of these systems utilize preset paths and fixed-value control strategies to complete the spraying task. However, due to the highly nonlinear and strongly coupled dynamic characteristics of the drone platform itself, coupled with the additional loads and disturbances introduced by the robotic arm's motion, the system faces high uncertainty during flight and operation. Furthermore, the spraying process involves complex physical processes such as paint flow, air pressure fluctuations, and nozzle reaction forces, further complicating system modeling and control.

[0004] In the prior art, Chinese patent CN116786326A discloses a method and system for controlling the operation of a spraying robot. The method includes: obtaining multiple degree-of-freedom nodes of the spraying robot; when the spraying robot receives a preset spraying path, splitting the preset spraying path to obtain multiple path types; obtaining the robot arm operation status of multiple degree-of-freedom nodes identified on the multiple path types, and performing model training on each of them to obtain multiple degree-of-freedom control models corresponding to the multiple path types and perform collaborative analysis, outputting multiple collaborative control parameters, and controlling the spraying operation. However, this type of method is still mainly based on offline or approximate models and fails to fully address the real-time dynamic disturbances of the spraying operation.

[0005] Existing control methods are usually based on simplified dynamic models and do not fully consider the real-time impact of the robot arm movement on the system's center of mass position, inertia distribution, and external disturbance torque. At the same time, dynamic factors such as airflow disturbances during the spraying process, the reaction force generated by the nozzle, and mass changes during the paint spraying process are often ignored. In high-precision spraying tasks, these neglected factors can easily cause spray trajectory deviation, uneven spray thickness, or exceed the target area, seriously affecting the spraying quality. In addition, the internal dynamic parameters of the system are constantly changing during the execution of the task, and fixed control parameters are difficult to adapt to the complex and dynamic operating environment, which limits the control robustness and operational stability of the system.

[0006] In summary, the existing drone-robotic arm collaborative spraying system still faces great challenges in dealing with dynamic changes such as multi-source disturbances, load fluctuations and posture coupling. The lack of efficient control strategies with good real-time performance, adaptability and robustness has become a key technical bottleneck restricting the application of high-quality unmanned spraying operations. Summary of the Invention

[0007] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a UAV-robotic arm system collaborative control method for precision spraying.

[0008] The purpose of the present invention can be achieved by the following technical solutions:

[0009] The present invention provides a collaborative control method for a UAV-mechanical arm system for precision spraying, comprising the following steps:

[0010] Use depth cameras and point cloud reconstruction algorithms to perform 3D modeling of the target surface, combine semantic recognition to determine the spraying area, and generate spraying path and spraying posture data;

[0011] Based on the spraying task requirements, including the target film thickness, an empirical estimation model for spraying film thickness is established to obtain spraying parameter data at the path point level, including spraying flow, spraying speed, and spraying distance;

[0012] Based on the spraying parameter data and spraying posture data, the load mass change at each point of the spraying path, the center of mass change caused by the robot arm actuation, and the spraying reaction force are calculated. Combined with the basic state data of the UAV-robot arm system, a coupled dynamics model of the coordinated motion of the UAV-robot arm system is constructed;

[0013] Based on the coupled dynamics model, the adaptive sliding mode control algorithm is used to generate dynamic compensation control quantity data, obtain the compensation control input of the UAV-manipulator system, and adjust the working state of the manipulator accordingly;

[0014] By measuring the surface image and point cloud after spraying, the actual spraying effect data is collected, the color distribution, film thickness distribution and texture characteristics are extracted, and compared with the spraying task requirements, error feedback data is generated, and the parameters of the adaptive sliding mode control algorithm are updated online based on the error feedback data.

[0015] Furthermore, the drone-robotic arm system includes a drone body, robotic arms arranged around the drone body, and a spraying device arranged at the end of the robotic arms. There are multiple robotic arms, which are evenly distributed on the outside of the drone body and have a radially expanded structure. Each robotic arm includes multiple rotatable joints and connecting rods connected in series.

[0016] Furthermore, the spraying path includes the boundary outline of the target spraying area, a set of path points and their spatial order, which are used to describe the motion trajectory of the drone-robotic arm system in the spraying task; the path point set contains a series of three-dimensional coordinate points and corresponding speed planning, which are used to guide the spraying operation to proceed smoothly along the set trajectory.

[0017] Furthermore, the spraying posture data includes the nozzle posture direction of the spraying device at the end of the robot arm at each spraying path point, which is defined as direction information represented by an equivalent quaternion, as well as the nozzle posture angle and the nozzle end spraying point position.

[0018] Furthermore, the method of using a depth camera and a point cloud reconstruction algorithm to perform three-dimensional modeling of the target surface, combining semantic recognition to determine the spraying area, and generating spraying path and spraying posture data specifically includes:

[0019] Use a depth camera to collect multi-view depth images of the target surface, and generate a three-dimensional point cloud model of the target surface through point cloud registration and fusion algorithms;

[0020] Based on the pre-trained convolutional neural network, the 3D point cloud model of the target surface is semantically segmented to extract the outline of the spraying area. , and perform grid path point sampling within the spraying area outline to obtain the path point set , where m is the number of path points in the path point set, Indicates the i The three-dimensional coordinates of the path points;

[0021] Generate corresponding spraying posture data for each path point, including nozzle posture direction, nozzle posture angle, and nozzle end spraying point position.

[0022] Furthermore, the generation of corresponding spraying posture data for each path point specifically includes:

[0023] For each path point in the path point set The local neighborhood point cloud at Perform principal component analysis to extract the surface unit normal vector at that point :

[0024]

[0025] in, Indicates a waypoint Local neighborhood point cloud The covariance matrix of Indicates a waypoint The estimated unit normal vector of the target surface at , is the unit vector candidate set;

[0026] Set the nozzle to the unit vector , the default selection is to align with the target surface normal, that is: ;

[0027] According to the spraying distance set in the spraying task , calculate the spray point position at the end of the nozzle :

[0028]

[0029] in, The end of the spraying device is at the path point The corresponding expected position;

[0030] Construct the attitude quaternion of the nozzle end , so that in the coordinate system of the aircraft or robotic arm, the default Z axis is aligned with the spray direction vector , expressed as:

[0031]

[0032] in, is a posture alignment function used to calculate the reference vector , that is, the nozzle default direction, such as the Z axis, rotates to the target vector The minimum rotation quaternion required, Indicates a waypoint The rotation quaternion of the nozzle attitude at ;

[0033] Get complete spraying posture data, including: nozzle posture direction , nozzle attitude angle , spray point position at the end of the nozzle .

[0034] Furthermore, the establishment of an empirical estimation model for spray film thickness to obtain spray parameter data at the path point level specifically includes:

[0035] For a set of waypoints And its corresponding spraying posture data, establish the spray film thickness experience estimation model, calculate the first i Predicted film thickness at each path point:

[0036]

[0037] in, is the predicted spray film thickness value, is the spraying pressure, is the spraying flow rate, is the spraying speed, is the spray angle correction function, is the nozzle attitude angle, is the spraying distance correction function, is the spray distance attenuation coefficient, obtained by experimental fitting, is the spraying distance;

[0038] According to the target film thickness in the spraying task , by adjusting the spraying parameters , so that the film thickness error satisfies the following constraints:

[0039]

[0040] in, is the film thickness error tolerance;

[0041] The final output path point-level spray parameter dataset is:

[0042]

[0043] in, For waypoints i Spraying parameter data.

[0044] Furthermore, the calculation of the load mass change at each point of the spraying path, the center of mass change caused by the actuation of the robotic arm, and the spraying reaction force based on the spraying parameter data and the spraying posture data specifically includes:

[0045] Calculate the mass of the spray load at each path point based on the path point spray parameter data set, spray posture data and spray parameter data and the center of mass position , the load mass changes dynamically with the spray flow rate and pressure:

[0046]

[0047] in, is the conversion coefficient between spray material density and flow rate, For waypoints i The spray flow rate, The time interval for spraying the waypoints, For waypoints i Location, is the offset of the center of mass of the spraying material relative to the path point position, For waypoints i The nozzle is oriented towards the unit vector;

[0048] Based on the mass of the spray load at each path point and the center of mass position Calculate the overall load mass of the spray path and center of mass changes :

[0049]

[0050] Where m is the number of path points;

[0051] Calculate the center of mass position of each link of the robotic arm based on the spraying posture data and corresponding quality , solve the overall center of mass position of the drone-robotic arm system:

[0052]

[0053] in, 、 is the mass and center of mass position of the UAV body, For the j The mass of each robot link and its center of mass position, is the number of links of the robotic arm; is the mass and position of the spray load, which changes dynamically during the spraying process;

[0054] Calculate the spraying reaction force based on the spraying parameter data of the path point:

[0055]

[0056] in, is the spraying reaction force, is the empirical coefficient, is the spraying pressure, is the spraying flow rate, is the spraying distance, is the unit vector of the spraying direction.

[0057] Furthermore, the coupled dynamics model of the coordinated motion of the UAV-manipulator system is:

[0058]

[0059] in, is the system generalized coordinate vector, including the position and posture of the UAV Joint angle with the robotic arm ; 、 are the first-order derivative and second-order derivative of the system's generalized coordinate vector respectively; is the mass inertia matrix of the drone-manipulator system, which explicitly depends on the dynamic center of mass position , reflecting the change in inertia caused by load change; is the Coriolis and centrifugal force matrix, including the additional coupling torque term caused by the center of mass offset, is the gravity term, which takes into account the effect of the change in the center of mass position on the gravitational torque;

[0060] For control input, it includes drone thrust and robotic arm drive torque; is the spraying reaction torque, the formula is:

[0061]

[0062] in, is the Jacobian matrix of the spraying point, is the spraying reaction force;

[0063] is the overall center of mass of the UAV-manipulator system caused by the mass change of the spraying load The coupling disturbance torque caused by the change is expressed as:

[0064]

[0065] in, is the overall center of mass of the UAV-robotic arm system About generalized coordinates The Jacobian matrix of is the total mass of the UAV-robotic arm system, is the system center of mass acceleration, which indicates the dynamic change of the center of mass caused by the movement of the robot arm or the change of the spray load. is the angular momentum of the robotic arm to the drone:

[0066]

[0067] in, Indicates the j The inertia matrix of the robot arm links around the center of mass, For the j The angular velocity of the robot arm link, For the j The linear velocity of the center of mass of the robot arm link, is the number of links of the robotic arm;

[0068] is the first-order derivative of the manipulator’s angular momentum, is the angular velocity of the UAV body.

[0069] Furthermore, the method of generating dynamic compensation control quantity data by using an adaptive sliding mode control algorithm based on the coupled dynamics model to obtain the compensation control input of the UAV-manipulator system specifically includes:

[0070] Define the system error vector :

[0071]

[0072] in, 、 are the current generalized coordinate vector of the UAV-manipulator system and the generalized coordinate vector corresponding to the desired trajectory;

[0073] Based on the system error vector Construct sliding surface:

[0074]

[0075] in, is the sliding surface, is the error vector The first derivative of is a positive definite diagonal matrix;

[0076] Based on coupled dynamic model and sliding surface and error vector Design an adaptive sliding mode control law:

[0077]

[0078] in, is the control input of the drone-robotic arm system, is the expected acceleration, is the positive definite control gain matrix, is the sliding surface sign function, is an adaptive disturbance estimation term used to compensate for model uncertainty and external disturbances, and is expressed as:

[0079]

[0080] in, is a positive definite learning rate matrix;

[0081] The compensation control input is obtained according to the adaptive sliding mode control law, the spraying reaction torque and the coupling disturbance torque:

[0082]

[0083] in, is the compensation control input;

[0084] The drone thrust and robotic arm joint torque are corrected in real time according to the compensation control input to achieve precise control and dynamic compensation of the system attitude and trajectory.

[0085] Compared with the prior art, the present invention has the following advantages:

[0086] (1) The present invention solves the problems of platform dynamic parameters changing with time, disturbance sources being diverse, and significant coupling effects of the robot arm motion on the platform flight attitude during the spraying process in a multi-rotor UAV-robot collaborative spraying system by establishing a dynamic coupling model that includes factors such as the platform flight attitude, the robot arm motion load, and the spraying reaction force. The traditional control strategy based on simplified models and fixed control parameters is difficult to cope with the complex dynamic environment during operation, resulting in decreased system stability and insufficient spraying accuracy. The present invention introduces a system multi-source dynamic disturbance identification mechanism at the modeling level, integrates factors such as load changes, center of mass drift, spraying reaction force, and airflow disturbances caused by the robot arm motion, and establishes a coupled dynamic model that reflects the dynamic characteristics of the system. On this basis, a control architecture based on disturbance observation and adaptive control is constructed, which can dynamically perceive the system state and disturbance changes, realize accurate compensation and robust control of the system nonlinearity and uncertainty, and effectively improve the operational stability and control accuracy of the unmanned spraying system in complex environments.

[0087] (2) The present invention uses a depth camera combined with a point cloud reconstruction algorithm and semantic recognition technology to perform three-dimensional modeling of the target surface and identify the spraying area, thereby solving the problem that traditional spraying path planning relies on manual or rule templates and is difficult to adapt to complex curved surfaces. It realizes automatic recognition of the spraying area, three-dimensional planning of path points, and accurate generation of spraying postures.

[0088] (3) The present invention constructs an empirical estimation model for spray film thickness based on the target film thickness and outputs spray parameter data at the path point level. This solves the technical problem of the existing method in which the spray parameters are uniformly set and cannot adapt to spatial changes, resulting in uneven film thickness. It realizes the dynamic adjustment and optimization of spray parameters (flow rate, speed, distance) at the path point level, significantly improving the spray quality control capability.

[0089] (4) The present invention combines spraying parameters with spraying posture data to systematically calculate the load change, center of mass offset and spraying reaction force of each path point during the spraying process, thereby solving the problem of difficult estimation of dynamic disturbances caused by changes in robot arm movements and spraying working conditions, and realizing accurate modeling of the overall center of mass and dynamic load of the system, providing reliable input for dynamic compensation of the control system.

[0090] (5) This invention effectively improves the robustness of the system to external disturbances and model uncertainty by designing an adaptive sliding mode control law based on system error and sliding surface, and introduces a disturbance estimation term for online compensation, thus achieving rapid response and high-precision tracking control in complex dynamic environments. This control strategy can dynamically adjust control parameters according to the system operating state, suppress chattering, and improve the stability and adaptability of the controller, thereby ensuring precise control of the attitude and position of the robot arm end effector during the spraying operation, and improving the spraying quality and operation consistency.

[0091] (6) The present invention establishes a spray quality error feedback mechanism by comparing the image and point cloud detection data after spraying with the model prediction results, achieving a detailed assessment of the spray color distribution, film thickness distribution, and texture characteristics, and obtaining information on the deviation between the spray effect and the expected effect. Based on the error feedback, the parameters of the adaptive sliding mode control algorithm are updated online, thereby achieving adaptive optimization and closed-loop correction of the spray control strategy, further improving the intelligence level of the spray operation and the consistency of film formation, and significantly enhancing the system's adaptability to complex target surfaces and changing environmental conditions.

[0092] (7) When calculating the compensation control input, the present invention takes into account the load mass change, the center of mass change caused by the movement of the manipulator, and the spraying reaction force, thus solving the problem of decreased control accuracy and stability caused by the traditional UAV-manipulator system ignoring the dynamic load change and reaction force during the spraying process. Since the reduction of the spraying material during the spraying process causes the system load mass to change continuously, the movement of the manipulator causes the overall center of mass position to shift, and the spraying operation itself generates a reaction force. These factors will cause the system dynamic characteristics to change. If not compensated, it will cause the attitude control error to increase, affecting the accurate execution of the spraying path and the spraying quality. By incorporating the load mass change, the dynamic adjustment of the center of mass position, and the spraying reaction force into the coupled dynamic model and control input calculation, the present invention realizes the accurate modeling and real-time compensation control of the dynamic characteristics of the UAV-manipulator system. The technical effects include significantly improving the stability and accuracy of the system attitude and trajectory control, effectively suppressing the disturbance caused by load fluctuations and reaction forces, and ensuring the path tracking accuracy and spraying uniformity during the spraying process. In addition, this technical feature enhances the robustness and adaptability of the system in complex spraying tasks, improves the overall spraying efficiency and quality, and meets the application requirements of high-precision precision spraying. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 This is a flow chart of a method for collaborative control of a UAV-manipulator system according to an embodiment of the present invention;

[0094] Figure 2 A flow chart of a method for spraying path and spraying posture data according to an embodiment of the present invention;

[0095] Figure 3 Flowchart of a method for constructing a coupled dynamics model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0096] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0097] Example 1:

[0098] The drone-robotic arm system of this embodiment includes a drone body, robotic arms arranged around the drone body, and a spraying device arranged at the end of the robotic arms. There are multiple robotic arms, which are evenly distributed on the outside of the drone body and have a radially expanded structure. Each robotic arm includes multiple rotatable joints and connecting rods connected in series, and spray paint is suspended under the drone.

[0099] This embodiment provides a UAV-manipulator system collaborative control method for precision spraying. Figure 1 As shown, the following steps are included:

[0100] Step S1: Use a depth camera and point cloud reconstruction algorithm to perform three-dimensional modeling of the target surface, combine semantic recognition to determine the spraying area, and generate spraying path and spraying posture data; wherein, the spraying path includes the boundary contour of the target spraying area, the path point set and its spatial order, which is used to describe the motion trajectory of the drone-robotic arm system in the spraying task; the path point set contains a series of three-dimensional coordinate points and corresponding speed planning, which is used to guide the spraying operation to proceed smoothly along the set trajectory; the spraying posture data includes the nozzle posture direction of the spraying device at the end of the robotic arm at each spraying path point, which is defined as the direction information represented by the equivalent quaternion, as well as the nozzle posture angle and the spraying point position at the end of the nozzle.

[0101] like Figure 2 As shown, step S1 specifically includes:

[0102] Step S101: using a depth camera to collect multi-view depth images of the target surface, and generating a three-dimensional point cloud model of the target surface through point cloud registration and fusion algorithms;

[0103] Step S102: Perform semantic segmentation on the target surface 3D point cloud model based on the pre-trained convolutional neural network to extract the spraying area contour ;

[0104] Step S103: Grid path point sampling is performed within the spraying area outline to obtain a path point set , where m is the number of path points in the path point set, Indicates the i The three-dimensional coordinates of the path points;

[0105] Step S104: Generate corresponding spraying posture data for each path point, including nozzle posture direction, nozzle posture angle, and nozzle end spraying point position, specifically including:

[0106] Step S105: For each path point in the path point set The local neighborhood point cloud at Perform principal component analysis to extract the surface unit normal vector at that point :

[0107]

[0108] in, Indicates a waypoint Local neighborhood point cloud The covariance matrix of Indicates a waypoint The estimated unit normal vector of the target surface at , is the unit vector candidate set;

[0109] Step S106: Set the nozzle direction unit vector , the default selection is to align with the target surface normal, that is: ;

[0110] Step S107: According to the spraying distance set in the spraying task , calculate the spray point position at the end of the nozzle :

[0111]

[0112] in, The end of the spraying device is at the path point The corresponding expected position;

[0113] Step S108: Construct the attitude quaternion of the nozzle end , so that in the coordinate system of the aircraft or robotic arm, the default Z axis is aligned with the spray direction vector , expressed as:

[0114]

[0115] in, is a posture alignment function used to calculate the reference vector , that is, the nozzle default direction, such as the Z axis, rotates to the target vector The minimum rotation quaternion required, Indicates a waypoint The rotation quaternion of the nozzle attitude at ;

[0116] Get complete spraying posture data, including: nozzle posture direction , nozzle attitude angle , spray point position at the end of the nozzle .

[0117] Step S1 of this embodiment solves the key technical problems in the automatic recognition, precise modeling and path planning of the UAV-mechanical arm spraying system in complex three-dimensional target surfaces. Traditional spraying methods usually rely on manual setting of the spraying area and path, which is difficult to adapt to target surfaces with complex shapes and variable structures, and the spraying posture control is imprecise, resulting in unstable spraying quality. By using a depth camera to collect multi-perspective depth images and combining point cloud registration and fusion algorithms, the system can reconstruct the three-dimensional point cloud model of the target surface with high precision, and achieve a comprehensive and detailed spatial perception of the spraying object. Semantic segmentation based on pre-trained convolutional neural networks effectively distinguishes and extracts the spraying area, ensures the accuracy and pertinence of path planning, and avoids the problem of blind spots or missed areas in spraying. By generating a set of path points through grid sampling, the system obtains a series of path points with three-dimensional coordinates and speed planning, ensuring that the spraying process proceeds smoothly along the predetermined trajectory, solving the problem of uneven spraying quality caused by discontinuous paths or uneven speeds. Principal component analysis is performed on the local neighborhood point cloud of each path point to extract the surface unit normal, enabling precise calculation of the nozzle attitude. This ensures alignment of the nozzle direction with the target surface normal, resulting in uniform and complete spray coverage. Accurate calculation of the spray point position at the nozzle end ensures a constant spray distance, effectively controlling coating thickness and adhesion quality. The use of quaternions to represent the nozzle attitude avoids the Euler angle gimbal lock issue, improves the stability and continuity of attitude control, and ensures precise positioning and flexible adjustment of the robotic arm's spray device.

[0118] Step S2: Based on the spraying task requirements, including the target film thickness, an empirical estimation model for spraying film thickness is established to obtain spraying parameter data at the path point level, including spraying flow rate, spraying speed, and spraying distance, specifically including:

[0119] For a set of waypoints And its corresponding spraying posture data, establish the spray film thickness experience estimation model, calculate the first i Predicted film thickness at each path point:

[0120]

[0121] in, is the predicted spray film thickness value, is the spraying pressure, is the spraying flow rate, is the spraying speed, is the spray angle correction function, is the nozzle attitude angle, is the spraying distance correction function, is the spray distance attenuation coefficient, obtained by experimental fitting, is the spraying distance;

[0122] According to the target film thickness in the spraying task , by adjusting the spraying parameters , so that the film thickness error satisfies the following constraints:

[0123]

[0124] in, is the film thickness error tolerance;

[0125] The final output path point-level spray parameter dataset is:

[0126]

[0127] in, For waypoints i Spraying parameter data.

[0128] Step S2 solves the core technical problem of how to achieve precise film thickness control during the spraying process. Traditional spraying operations often find it difficult to adjust the spraying parameters in real time to match the film thickness requirements under complex surfaces and variable working conditions, which can easily lead to uneven coating thickness and affect the spraying quality and functional effects. By establishing an empirical estimation model for spray film thickness, key parameters such as spray pressure, flow, speed and spray distance are associated with the spray film thickness, thereby achieving accurate prediction of the film thickness at the path point level. The correction functions for spray angle and spray distance respectively consider the influence of the nozzle attitude angle on the spraying effect and the attenuation effect of the spray distance, making the model closer to the actual physical spraying process and improving the accuracy and reliability of film thickness estimation. Based on the error constraints of the target film thickness and the predicted film thickness, the system can dynamically adjust the spray pressure, flow, speed and spray distance to achieve the optimal configuration of the spray parameters, ensure that the thickness of the spray film layer is within the allowable error range, and solve the problems of uneven spray thickness, local over-thickness or over-thinness. The resulting pathpoint-level spray parameter dataset output by this method provides precise and detailed operational guidance for subsequent spray execution, effectively enhancing the automation and intelligence of the spray process. The overall technical effect is to achieve online controllable and high-precision adjustment of spray film thickness, ensuring the stability and consistency of coating quality, improving the efficiency and reliability of spray operations, and significantly reducing the complexity and trial-and-error costs of manually adjusting spray parameters.

[0129] Step S3: Based on the spraying parameter data and the spraying posture data, the load mass change at each point of the spraying path, the center of mass change caused by the manipulator actuation, and the spraying reaction force are calculated. Combined with the basic state data of the UAV-manipulator system, a coupled dynamics model of the coordinated motion of the UAV-manipulator system is constructed;

[0130] Based on the spraying parameter data and spraying posture data, the load mass change at each point of the spraying path, the center of mass change caused by the movement of the robot arm, and the spraying reaction force are calculated, such as Figure 3 As shown, specifically including:

[0131] Step S301: Calculate the mass of the spray load at each path point based on the path point spray parameter data set, spray posture data and spray parameter data and the center of mass position , the load mass changes dynamically with the spray flow rate and pressure:

[0132]

[0133] in, is the conversion coefficient between spray material density and flow rate, For waypoints i The spray flow rate, The time interval for spraying the waypoints, For waypoints i Location, is the offset of the center of mass of the spraying material relative to the path point position, For waypoints i The nozzle is oriented towards the unit vector;

[0134] Step S302: Based on the mass of the spray load at each path point and the center of mass position Calculate the overall load mass of the spray path and center of mass changes :

[0135]

[0136] Where m is the number of path points;

[0137] Step S303: Calculate the center of mass position of each link of the robot arm according to the spraying posture data and corresponding quality , solve the overall center of mass position of the drone-robotic arm system:

[0138]

[0139] in, 、 is the mass and center of mass position of the UAV body, For the j The mass of each robot link and its center of mass position, is the number of links of the robotic arm; is the mass and position of the spray load, which changes dynamically during the spraying process;

[0140] Step S304: Calculate the spraying reaction force according to the spraying parameter data of the path point:

[0141]

[0142] in, is the spraying reaction force, is the empirical coefficient, is the spraying pressure, is the spraying flow rate, is the spraying distance, is the unit vector of the spraying direction.

[0143] The coupled dynamics model of the coordinated motion of the UAV-manipulator system is:

[0144]

[0145] in, is the system generalized coordinate vector, including the position and posture of the UAV Joint angle with the robotic arm ; 、 are the first-order derivative and second-order derivative of the system's generalized coordinate vector respectively; is the mass inertia matrix of the drone-manipulator system, which explicitly depends on the dynamic center of mass position , reflecting the change in inertia caused by load change; is the Coriolis and centrifugal force matrix, including the additional coupling torque term caused by the center of mass offset, is the gravity term, which takes into account the effect of the change in the center of mass position on the gravitational torque;

[0146] For control input, it includes drone thrust and robotic arm drive torque; is the spraying reaction torque, the formula is:

[0147]

[0148] in, is the Jacobian matrix of the spraying point, is the spraying reaction force;

[0149] is the overall center of mass of the UAV-manipulator system caused by the mass change of the spraying load The coupling disturbance torque caused by the change is expressed as:

[0150]

[0151] in, is the overall center of mass of the UAV-robotic arm system About generalized coordinates The Jacobian matrix of is the total mass of the UAV-robotic arm system, is the system center of mass acceleration, which indicates the dynamic change of the center of mass caused by the movement of the robot arm or the change of the spray load. is the angular momentum of the robotic arm to the drone:

[0152]

[0153] in, Indicates the j The inertia matrix of the robot arm links around the center of mass, For the j The angular velocity of the robot arm link, For the j The linear velocity of the center of mass of the robot arm link, is the number of links of the robotic arm;

[0154] is the first-order derivative of the manipulator’s angular momentum, is the angular velocity of the UAV body.

[0155] Step S3 proposes a systematic modeling and calculation method to address the complex coupled dynamics issues caused by the dynamic load and center of mass changes of the drone-robotic arm system during spraying operations. When a conventional drone equipped with a robotic arm sprays, the load mass continuously changes with the consumption of spray material and changes in spraying parameters. The center of mass shift caused by the robotic arm's movement also affects the overall dynamic characteristics of the system, and the spraying reaction force also perturbs the system. Without accurate modeling of these factors, control performance can be degraded, flight and spraying processes can become unstable, and spraying accuracy and safety can be compromised.

[0156] By combining spraying parameters with spraying posture data, the load mass change and center of mass position at each point along the spraying path are calculated, achieving real-time quantification of the dynamic load during the spraying process. The dynamic position of the system's overall center of mass is then determined by combining the center of mass of the robotic arm's linkage and the parameters of the drone itself. The calculation of the spraying reaction force reflects the impact of nozzle recoil on the robotic arm and drone during the spraying process. Ultimately, a coupled dynamic model incorporating the dynamic changes of the center of mass was constructed. This model not only accounts for the load-dependent adjustment of the system's inertia matrix but also incorporates key dynamic terms such as the Coriolis force, centrifugal force, and gravitational torque. The spraying reaction force and the perturbation torque caused by the center of mass change are introduced into the dynamic equations via the Jacobian matrix, achieving a comprehensive description of the complex coupling effects of the system. This model accurately reflects the dynamic characteristics and load coupling changes of the drone-robotic arm system during the spraying operation, improving the realism and accuracy of the system dynamics model and significantly enhancing the targetedness and effectiveness of subsequent control strategy design. By taking into account the changes in load mass and spray reaction forces during the spraying process in real time, the system can better predict and compensate for motion disturbances, improving flight stability and spraying quality, and reducing the risk of vibration and flight jitter caused by the robot arm's movement. This method also helps achieve safe and reliable control of the spraying operation, ensuring the coordinated and efficient collaboration between the drone and the robot arm, and significantly improving the accuracy and efficiency of the overall spraying task.

[0157] Step S4: Based on the coupled dynamics model, an adaptive sliding mode control algorithm is used to generate dynamic compensation control quantity data to obtain the compensation control input of the UAV-manipulator system; specifically, the following steps are included:

[0158] Define the system error vector :

[0159]

[0160] in, 、 are the current generalized coordinate vector of the UAV-manipulator system and the generalized coordinate vector corresponding to the desired trajectory;

[0161] Based on the system error vector Construct sliding surface:

[0162]

[0163] in, is the sliding surface, is the error vector The first derivative of is a positive definite diagonal matrix;

[0164] Based on coupled dynamic model and sliding surface and error vector Design an adaptive sliding mode control law:

[0165]

[0166] in, is the control input of the drone-robotic arm system, is the expected acceleration, is the positive definite control gain matrix, is the sliding surface sign function, is an adaptive disturbance estimation term used to compensate for model uncertainty and external disturbances, and is expressed as:

[0167]

[0168] in, is a positive definite learning rate matrix;

[0169] The compensation control input is obtained according to the adaptive sliding mode control law, the spraying reaction torque and the coupling disturbance torque:

[0170]

[0171] in, is the compensation control input;

[0172] The drone thrust and robotic arm joint torque are corrected in real time according to the compensation control input to achieve precise control and dynamic compensation of the system attitude and trajectory.

[0173] Step S5: Through the surface image and point cloud measurement after spraying, the actual spraying effect data is collected, the color distribution, film thickness distribution and texture characteristics are extracted, and the data are compared with the spraying task requirements to generate error feedback data. Based on the error feedback data, the parameters of the adaptive sliding mode control algorithm are updated online.

[0174] Step S5 specifically includes:

[0175] A high-resolution camera is used to obtain images of the surface after spraying, and a depth camera is used simultaneously to collect point cloud data of the sprayed surface. By combining the image and point cloud information, the sprayed area is registered and reconstructed to obtain a 3D texture model of the sprayed surface.

[0176] On the reconstructed model, the color distribution information is extracted through image processing algorithms to identify the color uniformity and edge clarity within the spraying area;

[0177] Combining point cloud height information with the film thickness calculation model, the film thickness distribution of the spraying area is analyzed to obtain the estimated film thickness at each path point;

[0178] The texture feature parameters of the sprayed surface are extracted through image texture analysis algorithm to evaluate the consistency and detail fidelity of the spraying quality;

[0179] The actual color distribution, film thickness distribution and texture feature data are compared point by point with the target features set for the spraying task, and the color deviation, film thickness error and texture deviation are calculated to form a comprehensive error feedback data set;

[0180] The error feedback data is input into the adaptive sliding mode control algorithm, and the control parameters such as the sliding surface gain and the positive definite control gain matrix are corrected online through the parameter self-adjustment mechanism, so that the control strategy is more in line with the actual spraying error characteristics, thereby improving the control accuracy and system robustness.

[0181] Example 2:

[0182] This embodiment provides a UAV-robotic arm system collaborative control system for precision spraying, including a UAV platform, a robotic arm device, a depth camera, a control unit, and a spraying actuator. The system uses a depth camera to collect multi-view depth images of the target surface, and combines point cloud reconstruction algorithms and semantic segmentation technology to achieve three-dimensional modeling of the target surface and accurate identification of the spraying area; based on spray path planning and spray posture generation, it formulates spray parameters at the path point level, including spray flow, speed, and spray distance, to ensure that the film thickness is uniform and meets the task requirements; the control unit integrates the load mass change during the spraying process, the center of mass offset caused by the movement of the robotic arm, and the spraying reaction force, and constructs a system model including coupled dynamics to achieve collaborative control of the UAV and the robotic arm; the system dynamically adjusts the spray parameters and control strategy through real-time feedback of the spray quality error, effectively improving the spraying accuracy and system stability, and ensuring high-quality spraying operations in complex environments.

[0183] If the above functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0184] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and such modifications or substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A collaborative control method for a UAV-robotic arm system for precision spraying, characterized in that: The following steps are involved: Use depth cameras and point cloud reconstruction algorithms to perform 3D modeling of the target surface, combine semantic recognition to determine the spraying area, and generate spraying path and spraying posture data; Based on the spraying task requirements, including the target film thickness, an empirical estimation model for spraying film thickness is established to obtain spraying parameter data at the path point level, including spraying flow, spraying speed, and spraying distance; Based on the spraying parameter data and spraying posture data, the load mass change at each point of the spraying path, the center of mass change caused by the robot arm movement, and the spraying reaction force are calculated. Combined with the basic state data of the UAV-robot arm system, a coupled dynamics model of the coordinated motion of the UAV-robot arm system is constructed; Based on the coupled dynamics model, the adaptive sliding mode control algorithm is used to generate dynamic compensation control quantity data, obtain the compensation control input of the UAV-manipulator system, and adjust the working state of the manipulator accordingly; By measuring the surface image and point cloud after spraying, the actual spraying effect data is collected, the color distribution, film thickness distribution and texture characteristics are extracted, and compared with the spraying task requirements, error feedback data is generated, and the parameters of the adaptive sliding mode control algorithm are updated online based on the error feedback data.

2. The collaborative control method of a UAV-manipulator system for precision spraying according to claim 1, characterized in that: The drone-robotic arm system includes a drone body, robotic arms arranged around the drone body, and a spraying device arranged at the end of the robotic arms. There are multiple robotic arms, which are evenly distributed on the outside of the drone body and have a radially expanded structure. Each robotic arm includes multiple rotatable joints and connecting rods connected in series.

3. The collaborative control method of a UAV-manipulator system for precision spraying according to claim 1, characterized in that: The spray path includes the boundary outline of the target spraying area, a set of path points and their spatial order, which are used to describe the motion trajectory of the drone-robotic arm system in the spraying task; the path point set contains a series of three-dimensional coordinate points and corresponding speed planning, which are used to guide the spraying operation to proceed smoothly along the set trajectory.

4. The collaborative control method of a UAV-manipulator system for precision spraying according to claim 1, characterized in that: The spraying posture data includes the nozzle posture direction of the spraying device at the end of the robot arm at each spraying path point, which is defined as direction information represented by an equivalent quaternion, as well as the nozzle posture angle and the nozzle end spraying point position.

5. The collaborative control method of a UAV-manipulator system for precision spraying according to claim 1, characterized in that: The method uses a depth camera and a point cloud reconstruction algorithm to perform three-dimensional modeling of the target surface, combines semantic recognition to determine the spraying area, and generates spraying path and spraying posture data, specifically including: Use a depth camera to collect multi-view depth images of the target surface, and generate a three-dimensional point cloud model of the target surface through point cloud registration and fusion algorithms; Based on the pre-trained convolutional neural network, the 3D point cloud model of the target surface is semantically segmented to extract the outline of the spraying area. , and perform grid path point sampling within the spraying area outline to obtain the path point set , where m is the number of path points in the path point set, Indicates the i The three-dimensional coordinates of the path points; Generate corresponding spraying posture data for each path point, including nozzle posture direction, nozzle posture angle, and nozzle end spraying point position.

6. The method for cooperative control of a UAV-manipulator system for precision spraying according to claim 5, characterized in that: The generating of the corresponding spraying posture data for each path point specifically includes: For each path point in the path point set The local neighborhood point cloud at Perform principal component analysis to extract the surface unit normal vector at that point : in, Indicates a waypoint Local neighborhood point cloud The covariance matrix of Indicates a waypoint The estimated unit normal vector of the target surface at , is the unit vector candidate set; Set the nozzle to the unit vector , the default selection is to align with the target surface normal, that is: ; According to the spraying distance set in the spraying task , calculate the spray point position at the end of the nozzle : in, The end of the spraying device is at the path point The corresponding expected position; Construct the attitude quaternion of the nozzle end , so that in the coordinate system of the drone or robotic arm, the default Z axis is aligned with the spray direction vector , expressed as: in, is a posture alignment function used to calculate the reference vector , that is, the nozzle default direction, rotated to the target vector The minimum rotation quaternion required, Indicates a waypoint The rotation quaternion of the nozzle attitude at ; Get complete spraying posture data, including: nozzle posture direction , nozzle attitude angle , spray point position at the end of the nozzle .

7. The method for cooperative control of a UAV-manipulator system for precision spraying according to claim 6, characterized in that: Establish an empirical estimation model for spray film thickness and obtain spray parameter data at the path point level, including: For a set of waypoints And its corresponding spraying posture data, establish the spray film thickness experience estimation model, calculate the first i Predicted film thickness at each path point: in, is the predicted spray film thickness value, is the spraying pressure, is the spraying flow rate, is the spraying speed, is the spray angle correction function, is the nozzle attitude angle, is the spraying distance correction function, is the spray distance attenuation coefficient, obtained by experimental fitting, is the spraying distance; According to the target film thickness in the spraying task , by adjusting the spraying parameters , so that the film thickness error satisfies the following constraints: in, is the film thickness error tolerance; The final output path point spraying parameter data set is: in, For waypoints i Spraying parameter data.

8. The method for cooperative control of a UAV-manipulator system for precision spraying according to claim 7, characterized in that: Based on the spraying parameter data and spraying posture data, the load mass change at each point of the spraying path, the center of mass change caused by the robot arm movement, and the spraying reaction force are calculated, including: Calculate the mass of the spray load at each path point based on the path point spray parameter data set, spray posture data and spray parameter data and the center of mass position , the load mass changes dynamically with the spray flow rate and pressure: in, is the conversion coefficient between spray material density and flow rate, For waypoints i The spray flow rate, The time interval for spraying the waypoints, For waypoints i Location, is the offset of the center of mass of the spraying material relative to the path point position, For waypoints i The nozzle is oriented towards the unit vector; Based on the mass of the spray load at each path point and the center of mass position Calculate the overall load mass of the spray path and center of mass changes : Where m is the number of path points; Calculate the center of mass position of each link of the robotic arm based on the spraying posture data and corresponding quality , solve the overall center of mass position of the drone-robotic arm system: in, 、 is the mass and center of mass position of the UAV body, For the j The mass of each robot link and its center of mass position, is the number of links of the robotic arm; is the mass and position of the spray load, which changes dynamically during the spraying process; Calculate the spraying reaction force based on the spraying parameter data of the path point: in, is the spraying reaction force, is the empirical coefficient, is the spraying pressure, is the spraying flow rate, is the spraying distance, is the unit vector of the spraying direction.

9. The method for cooperative control of a UAV-manipulator system for precision spraying according to claim 8, characterized in that: The coupled dynamics model of the coordinated motion of the UAV-manipulator system is: in, is the system generalized coordinate vector, including the position and posture of the UAV Joint angle with the robotic arm ; 、 are the first-order derivative and second-order derivative of the system's generalized coordinate vector respectively; is the mass inertia matrix of the UAV-manipulator system, which depends on the dynamic center of mass position , reflecting the change in inertia caused by load change; is the Coriolis and centrifugal force matrix, including the additional coupling torque term caused by the center of mass offset, is the gravity term, which takes into account the effect of the change in the center of mass position on the gravitational torque; For control input, it includes drone thrust and robotic arm drive torque; is the spraying reaction torque, the formula is: in, is the Jacobian matrix of the spraying point, is the spraying reaction force; is the overall center of mass of the UAV-manipulator system caused by the mass change of the spraying load The coupling disturbance torque caused by the change is expressed as: in, is the overall center of mass of the UAV-robotic arm system About generalized coordinates The Jacobian matrix of is the total mass of the UAV-robotic arm system, is the system center of mass acceleration, which indicates the dynamic change of the center of mass caused by the movement of the robot arm or the change of the spray load. is the angular momentum of the robotic arm to the drone: in, Indicates the j The inertia matrix of the robot arm links around the center of mass, For the j The angular velocity of the robot arm link, For the j The linear velocity of the center of mass of the robot arm link, is the number of links of the robotic arm; is the first-order derivative of the manipulator’s angular momentum, is the angular velocity of the UAV body.

10. The method for cooperative control of a UAV-manipulator system for precision spraying according to claim 9, characterized in that: Based on the coupled dynamics model, the adaptive sliding mode control algorithm is used to generate dynamic compensation control data to obtain the compensation control input of the UAV-manipulator system, including: Define the system error vector : in, 、 are the current generalized coordinate vector of the UAV-manipulator system and the generalized coordinate vector corresponding to the desired trajectory; Based on the system error vector Construct sliding surface: in, is the sliding surface, is the error vector The first derivative of is a positive definite diagonal matrix; Based on coupled dynamic model and sliding surface and error vector Design an adaptive sliding mode control law: in, is the control input of the drone-robotic arm system, is the expected acceleration, is the positive definite control gain matrix, is the sliding surface sign function, is an adaptive disturbance estimation term used to compensate for model uncertainty and external disturbances, and is expressed as: in, is a positive definite learning rate matrix; The compensation control input is obtained according to the adaptive sliding mode control law, the spraying reaction torque and the coupling disturbance torque: in, is the compensation control input; The drone thrust and robotic arm joint torque are corrected in real time according to the compensation control input to achieve precise control and dynamic compensation of the system attitude and trajectory.

Images