Remote assistance positioning assembly system and method for aircraft panel based on human-robot collaborative teleoperation system
By combining a human-machine collaborative teleoperation system with a nonlinear mapping neural network, efficient positioning and clamping of aircraft panel stringers and skin is achieved, solving the problems of manual dependence and large errors in the assembly process, and improving the degree of automation and assembly accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NORTHWESTERN POLYTECHNICAL UNIV
- Filing Date
- 2023-11-30
- Publication Date
- 2026-07-21
AI Technical Summary
During the assembly of aircraft panels, the positioning and clamping of the stringer and skin rely on manual operation, resulting in low automation, large errors, and affecting assembly accuracy and overall aircraft performance.
The system employs a human-machine collaborative teleoperation system, utilizing the collaborative work of the execution unit, measurement unit, control unit, and data processing unit. It uses a nonlinear mapping neural network for error prediction and compensation to achieve efficient positioning and clamping of the stringer and skin.
It has increased the automation level of aircraft panel assembly, reduced positional deviations during the manufacturing process, improved assembly accuracy and safety, and reduced the labor intensity of operators.
Smart Images
Figure CN117799850B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of aircraft assembly technology, and relates to an auxiliary positioning and assembly system and method for stringers on the skin in a digital unmanned assembly scenario of aircraft. Technical Background
[0002] Aircraft assembly is a crucial part of the overall aircraft manufacturing process, and aircraft panels are vital aerodynamic components. Their assembly quality directly impacts the assembly precision of subsequent parts and the entire aircraft, which is essential for ensuring flight performance. Currently, digital and intelligent manufacturing is gradually becoming the future trend of aircraft manufacturing. Achieving digital and intelligent manufacturing in factories has proven to effectively improve production efficiency and quality. Simultaneously, liberating workers' labor productivity, minimizing repetitive heavy manual labor, and investing limited human resources in more meaningful tasks are also major aspirations. The degree of automation and unmanned operation in factories is considered an important indicator of digital and intelligent manufacturing; however, the current assembly process for aircraft panels still heavily relies on manual labor, with a low level of automation. For example... Figure 6 As shown, the current assembly process in the factory typically begins with multi-point support flexible tooling for the definitive positioning and clamping of the skin at the positioning points. Then, with worker assistance, the stringer is positioned and clamped, using the stringer axis and the inner shape of the skin as a reference, and positioned according to pre-drilled positioning holes on the skin and stringer, and then clamped. Next, the skin and stringer are manually riveted together, that is, the stringer is riveted to the skin in a specific sequence to form a skin-stringer assembly. The frame and corner pieces are then positioned, clamped, assembled, and riveted in sequence, ultimately forming the aircraft panel. Throughout the entire aircraft panel assembly process, various deviations can occur due to factors such as equipment precision, production capacity, and environment, such as… Figure 7 As shown, this includes: part deviation, fixture deviation, positioning deviation, deviation caused by assembly process, and deformation deviation. It is difficult to compensate for this series of errors by manual operation based solely on worker experience, which will eventually lead to a large assembly deviation. Subsequently, forced assembly methods may be required, which will reduce the performance of the aircraft. Summary of the Invention
[0003] Based on the above background technology, this invention proposes a remote assisted positioning and assembly system and method for aircraft panels based on a human-machine collaborative teleoperation system, aiming to solve the technical problem of the auxiliary positioning and clamping of fuselage skin and stringer, which is highly dependent on manual labor and has a low degree of automation in the assembly of stringer and aircraft skin.
[0004] This invention can also predict and compensate for multi-source errors during the assembly process, further solving the technical problem of large errors in manual assembly.
[0005] The technical solution of this invention is as follows:
[0006] A remote-assisted positioning and assembly system for aircraft panels based on a human-machine collaborative teleoperation system, wherein the aircraft panels include a stringer and skin to be assembled as a single unit; its special feature is that the remote-assisted positioning and assembly system includes...
[0007] An execution unit is used to clamp the stringer and, under the control commands of the operation and control unit, move it to the correct assembly position and then gradually press it against the skin;
[0008] The measurement unit is used to acquire spatial coordinate point cloud data of multiple key measurement points on the skin and send it to the data processing unit, and to acquire the force information of the stringer as it is gradually pressed onto the skin and send it to the control unit.
[0009] The main force feedback teleoperated handle is used to receive assembly operation commands from the operator and send motion signals to the control unit;
[0010] The control unit is used to control the action of the execution unit according to the received motion signal; it is used to generate a force feedback command according to the received force information and transmit it to the main force feedback teleoperating handle. The main force feedback teleoperating handle applies resistance to the operator according to the force feedback command to simulate the force situation of the clamping part on the execution unit used to clamp the stringer, so as to provide the operator with tactile perception of the assembly status.
[0011] The data processing unit is used to process the measurement data of the measurement unit, including: visualizing the spatial coordinate point cloud data into video data for reference by the operator when inputting the assembly operation command through the main force feedback remote control handle.
[0012] Furthermore, the measuring unit is also used to obtain the actual spatial position of the end effector of the execution unit;
[0013] The data processing unit is also used to acquire the position, velocity, and spatial pose transformation of the master force feedback teleoperated handle, as well as the theoretical position transformation and velocity information of the execution unit, and send them to the control unit;
[0014] The control unit is also used to establish a nonlinear mapping neural network for predicting the position error of the upper end actuation part of the execution unit, and to compensate the position of the upper end actuation part of the execution unit in real time according to the output of the nonlinear mapping neural network; the nonlinear mapping neural network takes the spatial position change of the current action of the main end force feedback teleoperating handle as input and the position error of the current action of the upper end actuation part of the execution unit as output.
[0015] Furthermore, the nonlinear mapping neural network is a three-layer BP neural network, a radial basis function network, or a recurrent neural network.
[0016] Furthermore, the execution unit includes a slave robot and a clamping fixture; the end of the slave robot has an end effector, and the clamping fixture is fixed to the end effector of the slave robot for fixing and clamping the stringer.
[0017] Furthermore, the measurement unit includes coded markers, a binocular camera, a target ball, a laser tracker, and a six-dimensional force sensor. The coded markers are uniformly affixed to the skin, serving as key measurement points for the skin. The binocular camera is positioned directly opposite the assembly stringer of the skin and remains in a fixed position, used to photograph the skin to measure the spatial coordinate point cloud data of the key measurement points. The target ball is positioned on the clamping fixture and located on the central axis of the end effector. The laser tracker is positioned next to the slave robot, maintaining a fixed offset from it. The laser tracker tracks the spatial coordinates of the target ball in real time to obtain the actual position change of the end effector. The six-dimensional force sensor is installed between the end effector and the clamping fixture, used to acquire force information during the force application process as the stringer gradually presses against the skin and send it to the control unit.
[0018] Furthermore, the control unit includes an integrated control box.
[0019] Furthermore, the data processing unit includes a host computer.
[0020] Furthermore, it also includes an operating platform located away from the assembly workshop, on which the main force feedback remote control handle and the host computer are mounted.
[0021] This invention also provides a method for auxiliary positioning and assembly of aircraft panels using the aforementioned aircraft panel remote assisted positioning and assembly system based on a human-machine collaborative teleoperation system, characterized by the following steps:
[0022] Step 1: Fix the stringer to the clamping fixture, turn on the laser tracker, binocular camera and slave robot, and the laser tracker will start tracking the target ball;
[0023] Step 2: Identify the positional relationship between the stringer and the skin based on the video screen displayed on the host computer. Using this as a reference, manipulate the master end force feedback remote control handle to send motion signals to the integrated control box. The integrated control box drives the slave robot to complete the corresponding spatial motion, so that the end effector reaches the expected position, lifts the stringer and moves it to the vicinity of the target assembly position on the skin.
[0024] Step 3: Manipulate the main force feedback remote control handle to gradually press the stringer onto the skin. During the force application process, the six-dimensional sensor sends the force information to the integrated control box. The integrated control box generates a force feedback command based on the force information and sends it to the main force feedback remote control handle. The main force feedback remote control handle applies resistance to the operator's fingers based on the force feedback command to simulate the force situation and provide tactile perception of the assembly status.
[0025] Furthermore, in the initial stage of assembly, the slave robot predicts the position error of its end effector every time it moves using a nonlinear mapping neural network, and compensates for the position of the end effector in real time based on the prediction results, and performs secondary compensation for the position of the end effector based on the measured data of the laser tracker; in the middle and later stages of assembly, the slave robot predicts the position error of its end effector every time it moves using a nonlinear mapping neural network, and compensates for the position of the end effector in real time based on the prediction results.
[0026] The beneficial effects of this invention are:
[0027] 1. This invention makes full use of the accurate measurement data provided by the measurement unit to provide a reference for the main force feedback remote control handle in the manual operation and control unit. This enables the execution unit to move the stringer to the correct assembly position and gradually press it onto the skin according to the control commands output by the operation and control unit, thereby realizing the positioning and assembly of the two and improving the degree of automation of aircraft panel assembly.
[0028] 2. This invention constructs a nonlinear mapping neural network, which is used to predict the position error of the end-effector of the execution unit before and after each action of the execution unit, and uses the predicted position error as a compensation amount to perform real-time error compensation for the position of the end-effector of the execution unit. This minimizes the position error of the execution unit caused by a series of factors such as manufacturing geometric errors, actual working conditions, signal transmission problems, its own weight, load weight, motion acceleration function settings, and internal mechanical working conditions. This can greatly improve the positioning and assembly efficiency and positioning and assembly accuracy.
[0029] 3. This invention supports remote control, which can effectively prevent operators from approaching the relatively dangerous assembly site and ensure the safety of production operations. Since it adopts a remote operation control method instead of pre-input control commands, it does not require operators to learn programming. Instead, it remotely controls assembly through spatial movements similar to manual assembly, which is more flexible and easier to learn.
[0030] 4. This invention has strong versatility and can be applied to auxiliary positioning and assembly in multiple models and occasions. Most of the equipment and components included in the positioning and assembly system of this invention are general-purpose equipment, so there is no need to purchase too many new equipment. Furthermore, it is relatively easy to assemble and debug, occupies a small area, and does not require replanning of the factory workshop or redesign of the assembly process.
[0031] 5. This invention can make full use of the operator's knowledge and experience, avoiding the waste of talent, and is more convenient and versatile than computers.
[0032] 6. Most of the devices in this invention possess a high degree of flexibility, allowing for the replacement of devices with different functions according to actual needs. By changing the end-effector of the slave robot, various assembly scenarios can be achieved, and various assembly objects can be freely interchanged. The slave robot can also be equipped with a suitable robotic arm based on the usage scenario and cost. Attached Figure Description
[0033] Figure 1 This is a schematic diagram of the structure of the aircraft panel remote assisted positioning and assembly system of the present invention;
[0034] Figure 2 This is a schematic diagram of the end effector of the slave robot in the remote assisted positioning and assembly system for aircraft panels of the present invention.
[0035] Figure 3 This is a schematic diagram of the clamping fixture of the present invention;
[0036] Figure 4 This is a system architecture diagram of the present invention;
[0037] Figure 5 This is a system workflow diagram of the present invention;
[0038] Figure 6 This is a schematic diagram of a wall panel being clamped in a multi-point support flexible fixture;
[0039] Figure 7 This is a fishbone diagram showing the sources of error during the panel assembly process.
[0040] Icon labels:
[0041] 1-Slave robot; 2-Binocular camera; 3-Stringer; 4-Aircraft panel; 5-Host computer; 6-Master force feedback remote control handle; 7-Operating table; 8-Integrated control box; 9-Adjustable base; 10-Laser tracker; 11-Clamping fixture; 12-End effector; 13-End flange; 14-Six-dimensional force sensor; 15-Target ball holder; 16-Target ball. Detailed Implementation
[0042] The present invention will now be described in further detail with reference to the accompanying drawings. This description is exemplary and is only used to explain the embodiments of the present invention, and should not be construed as limiting the present invention.
[0043] like Figures 1 to 3 As shown, the aircraft panel remote assisted positioning and assembly system based on human-machine collaborative teleoperation system provided by the present invention includes an execution unit, a measurement unit, a main force feedback teleoperation handle 6, a control unit, and a data processing unit.
[0044] The actuator is used to clamp the stringer 3 and, under the control commands of the operation and control unit, move the stringer 3 to the correct assembly position and gradually press it onto the skin. The actuator includes a slave robot 1 and a clamping fixture 11. The slave robot 1 has an end effector 12 at its end, and the clamping fixture 11 is fixed to the end effector 12 of the slave robot 1 for fixing and clamping the stringer 3; as a preferred embodiment, such as Figure 3 As shown, the clamping fixture 11 in this embodiment is fork-shaped to prevent the long stringer 3 from bending due to its excessive length and poor rigidity during clamping. Of course, the structure of the clamping fixture 11 is not limited to... Figure 3 As shown, the specific structure of the clamping fixture 11 can also be adjusted according to the actual operation requirements, as long as it can reliably clamp the stringer 3 and ensure that the stringer 3 does not bend as much as possible during clamping.
[0045] The measurement unit includes coded markers, a binocular camera 2, a target ball 16, a laser tracker 10, and a six-dimensional force sensor 14. The coded markers are uniformly affixed to the skin of the panel 4, serving as key measurement points (KMPs). It is important to avoid affixing the coded markers to the mounting position of the stringer 3 to prevent obstruction during assembly. The binocular camera 2 is positioned directly opposite the panel 4 and fixed in place, used to photograph the skin of the panel 4 to measure the spatial coordinate point cloud data of the key measurement points. Specifically, the binocular camera 2 can be supported and fixed on a camera bracket. The target ball 16 is located on the central axis of the end effector 12 of the slave robot 1, specifically mounted on the clamping fixture 11 via a target ball holder 15. The center of the target ball 16 must pass through the rotation centerline of the end effector 12 to facilitate the calculation of the spatial position of the end effector 12. A laser tracker 10 is positioned next to the slave robot 1, maintaining a fixed offset from it. The laser tracker 10 tracks the spatial coordinates of the target ball 16 in real time to obtain the actual position changes of the end effector 12 of the slave robot 1. A six-dimensional force sensor 14 is installed between the end effector 12 of the slave robot 1 and the clamping fixture 11. It is used to obtain the force information during the force application process and send it to the control unit when the stringer 3 is gradually pressed onto the skin.
[0046] The main force feedback teleoperated handle 6 is used to receive assembly operation commands from the operator and send motion signals to the control unit. The operator can input assembly operation commands by gripping the main force feedback teleoperated handle 6 and making specific movements.
[0047] The control unit controls the actions of the slave robot 1 based on motion signals from an external command input unit. The control unit includes an integrated control box 8. During actual assembly, the operator manually controls the master force feedback teleoperated handle 6 to perform specific movements. The integrated control box 8 receives the motion signals from the master force feedback teleoperated handle 6, driving the end effector 12 of the slave robot 1 to the desired position to clamp and transport the stringer 3 to the assembly target location. The integrated control box 8 can also receive force information from the six-dimensional force sensor 14 regarding the end effector 12, process it, generate force feedback commands, and send them to the master force feedback teleoperated handle 6. The master force feedback teleoperated handle 6 can apply resistance to the operator's fingers based on the received force feedback commands to simulate the force on the end effector 12.
[0048] The data processing unit, including the host computer 5, processes the measurement data from the measurement unit. The host computer 5 is used for: 1) converting the spatial coordinate point cloud data of key measurement points measured by the binocular camera 2 into video data using publicly available visualization processing methods, so that workers can observe and adjust the master force feedback teleoperation handle 6 in real time to eliminate the error between the actual position of the stringer 3 and the target assembly position, ultimately enabling the stringer 3 to move to the target assembly position; 2) utilizing the teleoperation control software installed on the host computer 5 to obtain the velocity and spatial pose transformation information of the master force feedback teleoperation handle 6, and performing smooth mapping processing on the velocity and spatial pose transformation information of the master force feedback teleoperation handle 6, combined with motion space scaling to calculate the theoretical position transformation of the end effector 12 of the slave robot 1; 3) calculating the position response of each axis of the slave robot 1 and sending it as a control quantity to the integrated control box 8, so that the integrated control box 8 can drive the slave robot 1 to move according to the control quantity.
[0049] It should be noted that the structural form of the master force feedback teleoperated handle 6 and the slave robot 1 used in this invention is not limited, and can be any force feedback teleoperated handle and robot with the above-mentioned corresponding functions in the prior art.
[0050] Furthermore, for ease of operation, in this embodiment, the host computer 5 and the main force feedback remote control handle 6 are mounted on the control panel 7. Alternatively, the host computer 5, the main force feedback remote control handle 6, and the control panel 7 can be located in the control room, while the execution unit and measurement unit are located in the assembly workshop, which is far from the control room. This allows operators to remotely control the assembly of aircraft panels from away from the assembly site.
[0051] In addition, considering that during the actual assembly process, after each action of the slave robot 1 is completed, there may be a certain positional deviation between the actual position and the expected position of its end effector 12, in order to improve the positioning accuracy of the end effector 12 during the assembly process and improve the assembly quality and efficiency, the integrated control box 8 of the present invention will also receive the actual position change of the end effector 12 acquired and sent by the laser tracker 10, and receive the theoretical position change of the end effector 12 obtained based on the spatial position change of the master force feedback teleoperating handle 6, and compare the actual position change of the end effector 12 with the theoretical position change, output a compensation command after obtaining the deviation, and record the historical data of the master and slave position change errors; the integrated control box 8 constructs a nonlinear mapping neural network based on the recorded historical data of the master and slave position changes to predict the position error of the end effector 12 before and after each action of the end effector 12. The input of the neural network is the position change data of the current action of the master force feedback teleoperating handle 6, and the output is the predicted position deviation of the current action of the end effector 12. The integrated control box 8 uses the predicted position deviation of the end effector 12 output by the nonlinear mapping neural network as a compensation amount to perform real-time error compensation for the position of the end effector 12, thereby improving assembly precision and accuracy. Furthermore, considering that the nonlinear mapping neural network is trained while in use, in the initial stage of assembly, after each action performed by the end effector 12, this invention can first use the predicted value of the nonlinear mapping neural network to perform error compensation for the position of the end effector 12, and then perform secondary compensation based on the measured data from the laser tracker. This can prevent the error compensation predicted by the nonlinear mapping neural network from failing to meet requirements at the beginning. As the nonlinear mapping neural network is used more frequently, its prediction accuracy becomes increasingly higher, and only one error compensation based on its predicted value is needed to meet the requirements. At this point, there is no need to use the measured data from the laser tracker for secondary compensation.
[0052] like Figure 1 , 2 As shown in Figures 4 and 5, the method for auxiliary positioning and assembly of aircraft panels using the aircraft panel remote assisted positioning and assembly system of the present invention is as follows:
[0053] First, in the assembly workshop, the stringer 3 is clamped and fixed on the clamping fixture 11 located at the end of the slave robot. The laser tracker 10, the binocular camera 2 and the slave robot 1 are turned on, and the laser tracker 10 begins to track the target ball 16 located on the center line of the clamping fixture 11.
[0054] Subsequently, the assembly workshop is cleared, and workers observe the assembly site video sent from the binocular camera 2 through the host computer 5 at the operation platform 7 in the control room. The operators identify the positional relationship between the stringer 3 and the skin based on the video, and then manually manipulate the master force feedback remote control handle 6 to make specific spatial movements. The integrated control box 8 receives the handle movement signal sent by the master force feedback remote control handle 6, and drives the slave robot 1 to complete the corresponding spatial movements, so that the end effector 12 of the slave robot 1 reaches the expected position and completes the action of lifting the stringer 3 and transferring it to the vicinity of the target assembly position on the wall panel 4.
[0055] The specific data transfer process during this assembly phase is as follows:
[0056] 1) Use the remote operation control software installed in the host computer 5 to collect the speed and spatial pose transformation information of the master force feedback remote operation handle 6, and extract the spatial position transformation information from it.
[0057] 2) The remote control software performs smooth mapping processing on the collected speed and spatial pose information of the master end force feedback remote control handle 6, and calculates the theoretical position change and speed information of the end effector 12 of the slave robot 1 by combining the motion space scaling.
[0058] 3) The host computer sends the theoretical position change and speed information of the end effector 12 as a control signal to the integrated control box 8, and the integrated control box 8 drives the slave robot 1 to move according to the received control signal;
[0059] 4) The slave robot 1 is connected to the integrated control box 8 via a cable. The slave robot 1 receives the control signals sent by the host computer 5, interprets the timing signals (including the speed, direction and trajectory of the specified motion), and makes a motion response.
[0060] After the stringer 3 is transferred to the target assembly position, the worker manipulates the master-end force feedback teleoperation handle 6 to gradually press the stringer 3 onto the skin of the wall panel 4. During the application of force, the six-dimensional force sensor 14 sends the force information transmitted to the end effector 12 through the clamping fixture 11 to the integrated control box 8 for processing and generating force feedback commands. Then, the integrated control box 8 sends the force feedback commands to the master-end force feedback teleoperation handle 6. The master-end force feedback teleoperation handle 6 applies resistance to the worker's fingers according to the force feedback commands to simulate the force situation of the end effector 12 of the slave robot 1, allowing the worker to use touch to perceive the assembly status of the stringer 3.
[0061] In the final stage of assembly positioning, the binocular camera 2 acquires spatial coordinate point cloud data of key measurement points on the skin by photographing the coded marks on the skin, achieving the effect of "points representing surfaces". After the spatial coordinate point cloud data of the key measurement points is visualized by the host computer 5, the operator visually compares it with the standard position data required by production on the host computer 5. Based on the comparison results, the operator manipulates the main force feedback remote control handle 6 to fine-tune the position of the stringer 3. Since the skin is relatively thin and easily deformed, the stringer 3 is squeezed and deformed during fine-tuning, thereby achieving the purpose of fine-tuning the position and shape of the skin. When the visual comparison results are within the acceptable deviation range required by production, the positioning and assembly process is completed. The final assembly quality inspection is passed by the "average" deviation of the key measurement points. +standard deviation "To measure assembly accuracy, where the mean is..." The standard deviation reflects the accuracy of the dimensions. This indicates dimensional stability. The standard positional data required for production are measured data from pre-assembled wall panels that meet production requirements via manual assembly.
[0062] Considering that in actual operation, during the movement of the slave robot 1, a series of factors may occur, such as manufacturing geometric errors, actual working conditions, signal transmission problems, its own weight, load weight, motion acceleration function settings, and internal mechanical working conditions, causing the end effector position of the slave robot 1 to deviate from the expected position by a certain distance after completing a movement action, i.e., absolute positioning accuracy. We use the magnitude of the position error of the end effector 12 to represent the level of absolute positioning accuracy of the end effector 12. The smaller the position error of the end effector 12, the higher its absolute positioning accuracy; the larger the position error, the lower its absolute positioning accuracy. To eliminate this kind of deviation as much as possible, this invention adopts a slave robot deviation compensation method based on neural network optimization, which transforms the spatial position of the master end force feedback teleoperated handle 6 obtained by the host computer 5. The calculated theoretical position transformation of the end effector 12 of the slave robot 1 Together, they are sent to the integrated control box 8. The integrated control box 8 constructs a nonlinear mapping neural network to predict the position transformation error of the master-slave end effector 12, thereby predicting the position deviation of the end effector 12. Then, the integrated control box 8 compensates for the position of the end effector 12 based on the prediction result, specifically as follows:
[0063] a) After the worker manipulates the master force feedback remote control handle 6 to perform a specific movement, the remote control software on the host computer 5 obtains the spatial pose transformation information of the master force feedback remote control handle 6 and extracts the spatial position transformation from it. ;
[0064] b) The remote control software will change the spatial position of the master force feedback remote control handle 6. A smooth mapping process is performed, and the theoretical position transformation of the end effector 12 of the slave robot 1 is calculated by combining the motion space scaling. According to the theoretical error model (i.e. , For the actual position change of the end effector 12, The theoretical position transformation of the end effector 12 after error compensation. (For the position transformation error of the end effector 12), the integrated control box 8 acquires historical position transformation data of the master and slave ends, and establishes the spatial position transformation of the master end force feedback remote control handle 6. As input, the position deviation is calculated based on the historical transformation error of the end effector 12. For the nonlinear mapping neural network output, the integrated control box 8 controls the position deviation of the neural network output. As a compensation amount, error compensation is performed on the end effector 12. At this time, the theoretical position change of the end effector 12 after error compensation is... .
[0065] c) After the slave robot 1 completes one movement, the laser tracker 10 tracks the spatial position of the target ball 16 and obtains the actual position change of the end effector 12 in real time. The integrated control box 8 receives the actual position change of the end effector 12 obtained by the laser tracker 10. The host computer 5 sends the main force feedback remote control 6 to change its spatial position. The theoretical position transformation of the end effector 12 after error compensation based on the neural network prediction results by the integrated control box 8. The integrated control box 8 utilizes the obtained... , and The position transformation error of the end effector 12 can be calculated. The integrated control box 8 adjusts according to the position change error. The position of the end effector 12 is compensated twice and the historical data is included. This historical data is used for the training and calculation of the neural network.
[0066] d) Due to different position changes made by the master force feedback teleoperated handle 6, the end effector 12 of the slave robot 1 will have different position error responses. Therefore, the position change error of the end effector 12 is... This can be viewed as a spatial position change of the main force feedback teleoperated handle 6. The function, i.e. ;
[0067] e) As the frequency of use of nonlinear mapping neural networks increases, the prediction bias of the nonlinear mapping neural network output increases. Increasingly accurate, while the position transformation error of the end effector 12... As the value gradually approaches zero, it indicates that the accuracy of the neural network's prediction is extremely high, and secondary compensation is no longer required with the help of the laser tracker 10.
[0068] For predicting the position transformation error between the master and slave ends, the nonlinear mapping neural network constructed in this embodiment is a three-layer BP neural network (which meets the functional requirements and consumes the least amount of computational resources). The input layer has three neurons, which respectively represent the spatial position transformation of the master force feedback teleoperated handle 6 along three coordinate directions. The output layer also has three parameters: the position transformation error of the end effector 12. =(ΔxΔyΔz), obtained through empirical formulas Calculate the number of nodes in the hidden layer ( It is the number of hidden nodes. (This refers to the number of inputs) to construct a 3*7*3 three-layer BP neural network. Based on the required business expression, the hidden layer activation function and the output layer activation function are respectively the tansigmoid sigmoid transfer function and the purelin linear transfer function. The training algorithm uses the Levenberg-Marquardt algorithm with the trainlm function, and the network performance function is mse (root mean square). The trained three-layer BP neural network predicts the master-slave position transformation error during assembly processing, and uses the predicted deviation output value as the system compensation amount to perform real-time error compensation for the position of the end effector 12, achieving human-machine collaborative control. It should be noted that in other embodiments, a radial basis function network or a recurrent neural network can also be constructed to predict the master-slave position transformation error.
[0069] The above description is merely an embodiment of the present invention and does not specifically limit the scope of protection of the present invention. Any equivalent or similar modifications made by those skilled in the art within the technical scope described in the present invention, based on the inventive content, are all within the scope of protection of the present invention.
Claims
1. A remote assisted positioning and assembly system for aircraft panels based on a human-machine collaborative teleoperation system, wherein the aircraft panels include a stringer and skin to be assembled as a single unit; characterized in that: The remote assisted positioning assembly system includes An execution unit is used to clamp the stringer and, under the control commands of the operation and control unit, move it to the correct assembly position and then gradually press it against the skin; The measurement unit is used to acquire spatial coordinate point cloud data of multiple key measurement points on the skin and send it to the data processing unit, acquire the force information of the stringer when it is gradually pressed onto the skin and send it to the control unit, and acquire the actual spatial position of the end execution part of the execution unit. The main force feedback teleoperated handle is used to receive assembly operation commands from the operator and send motion signals to the control unit; The control unit is used to control the action of the execution unit according to the received motion signal; it is used to generate a force feedback command according to the received force information and transmit it to the main force feedback teleoperating handle. The main force feedback teleoperating handle applies resistance to the operator according to the force feedback command to simulate the force on the clamping part of the execution unit used to clamp the stringer, so as to provide the operator with tactile perception of the assembly status; the control unit is also used to establish a nonlinear mapping neural network for predicting the position error of the upper end execution part of the execution unit, and to compensate the position of the upper end execution part of the execution unit in real time according to the output of the nonlinear mapping neural network; the nonlinear mapping neural network takes the spatial position transformation of the current action of the main force feedback teleoperating handle as input and the position error of the current action of the upper end execution part of the execution unit as output; The data processing unit is used to process the measurement data of the measurement unit, including: visualizing the spatial coordinate point cloud data into video data for reference by the operator when inputting the assembly operation command through the main force feedback remote control handle; The data processing unit is also used to acquire the position, velocity, and spatial pose transformation of the master force feedback teleoperated handle, as well as the theoretical position transformation and velocity information of the execution unit, and send them to the control unit; The execution unit includes a slave robot and a clamping fixture; the slave robot has an end effector at its end, and the clamping fixture is fixed to the end effector of the slave robot for fixing and clamping the stringer. The control unit includes an integrated control box, which is used to: in the initial stage of assembly, predict the position error of the end effector of the slave robot every time it moves, and compensate the position of the end effector in real time based on the prediction result, and perform secondary compensation of the position of the end effector based on the measured data of the laser tracker; in the middle and later stages of assembly, predict the position error of the end effector of the slave robot every time it moves, and compensate the position of the end effector in real time based on the prediction result.
2. The aircraft panel remote assisted positioning and assembly system based on a human-machine collaborative teleoperation system according to claim 1, characterized in that: The nonlinear mapping neural network is a three-layer BP neural network, a radial basis function network, or a recurrent neural network.
3. The aircraft panel remote assisted positioning and assembly system based on a human-machine collaborative teleoperation system according to claim 1 or 2, characterized in that: The measurement unit includes coded markers, a binocular camera, a target ball, a laser tracker, and a six-dimensional force sensor. The coded markers are uniformly affixed to the skin, serving as key measurement points. The binocular camera is positioned directly opposite the assembly stringer of the skin and remains fixed in position, used to photograph the skin to measure the spatial coordinate point cloud data of the key measurement points. The target ball is positioned on the clamping fixture and located on the central axis of the end effector. The laser tracker is positioned next to the slave robot, maintaining a fixed offset. The laser tracker tracks the spatial coordinates of the target ball in real time to obtain the actual position change of the end effector. The six-dimensional force sensor is installed between the end effector and the clamping fixture to acquire force information during the force application process as the stringer gradually presses against the skin and sends it to the control unit.
4. The aircraft panel remote assisted positioning and assembly system based on a human-machine collaborative teleoperation system according to claim 3, characterized in that: The data processing unit includes a host computer.
5. The aircraft panel remote assisted positioning and assembly system based on a human-machine collaborative teleoperation system according to claim 4, characterized in that: It also includes an operating platform located away from the assembly workshop, on which the main force feedback remote control handle and the host computer are mounted.
6. A method for auxiliary positioning and assembly of aircraft panels using the aircraft panel remote assisted positioning and assembly system based on a human-machine collaborative teleoperation system as described in claim 5, characterized in that, Includes the following steps: Step 1: Fix the stringer to the clamping fixture, turn on the laser tracker, binocular camera and slave robot, and the laser tracker will start tracking the target ball; Step 2: Identify the positional relationship between the stringer and the skin based on the video screen displayed on the host computer. Using this as a reference, manipulate the master end force feedback remote control handle to send motion signals to the integrated control box. The integrated control box drives the slave robot to complete the corresponding spatial motion, so that the end effector reaches the expected position, lifts the stringer and moves it to the vicinity of the target assembly position on the skin. Step 3: Manipulate the main force feedback remote control handle to gradually press the stringer onto the skin. During the force application process, the six-dimensional sensor sends the force information to the integrated control box. The integrated control box generates a force feedback command based on the force information and sends it to the main force feedback remote control handle. The main force feedback remote control handle applies resistance to the operator's fingers based on the force feedback command to simulate the force situation and provide tactile perception of the assembly status.