An assembly unit based on a wire-controlled robot

Abstract

The application discloses an assembly unit based on a wire-controlled robot, which comprises a wire-controlled execution system, a guiding assembly system and a ground execution system. The wire-controlled execution system can be flexibly arranged according to the size of a working space and comprises a control system, a truss, a rope, a base and an end effector. The end effector is connected with a tooling fixture, a camera and a sensor, can execute linear motion and circumferential rotation motion, and utilizes a deep SARSA algorithm to perform path planning on the wire-controlled execution system and solve a motion trajectory. The guiding assembly system comprises a control system and a laser projection device, can provide accurate positioning, guiding and verification of parts, and the ground execution system comprises a control system, an intelligent carrying robot and an assembly robot. The assembly robot is equipped with a camera and a sensor and can provide intelligent auxiliary assembly. The combined operation of the three systems is suitable for most assembly scenes and can improve the automation and intelligent level of an assembly process.

Description

Technical Field

[0001] This invention belongs to the field of complex product manufacturing and assembly technology, specifically relating to an assembly unit based on a wire-controlled robot. Background Technology

[0002] Complex products, such as communication and navigation equipment, radar, satellites, and spacecraft, are characterized by complex structures, short development cycles, rapid replacement, diverse varieties, and high reliability requirements. Traditional assembly methods are no longer sufficient to meet the increasingly complex assembly needs of military and civilian products. The gradual application of emerging intelligent assembly technologies is replacing traditional manual and semi-automated assembly techniques. Although a small number of tooling devices can replace workers in completing simple and tedious operations, they still suffer from high costs, insufficient large-scale assembly capacity, and difficulty in quickly responding to changes in component types and processes. Furthermore, the installation of fasteners for large components (such as aircraft wings and satellite bodies), such as the assembly of satellite body corner pieces, brackets, and supports, requires precise positions, sizes, hole positions, and other auxiliary assembly information. However, these assembly operations still rely on manual measurement, reference datum establishment, and marking. Operators also need to consult a large amount of documentation for assembly guidance, resulting in high labor intensity, low assembly quality, accuracy, and efficiency. Therefore, traditional assembly guidance methods can no longer guarantee the requirements of high-intensity development and high-frequency iteration of complex products. Summary of the Invention

[0003] Purpose of the invention: To address the problems existing in the above-mentioned background technology, the present invention provides an assembly unit based on a wire-controlled robot.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: an assembly unit based on a wire-controlled robot, comprising the following steps:

[0005] (1) Based on the size of the workspace, the wire control execution system is flexibly arranged to accurately control the part position, assembly path and assembly sequence;

[0006] (2) Based on the installation scenario and the requirements for fastener assembly, the mobile deployment of the assembly system is used to accurately complete product positioning and installation.

[0007] (3) Based on the requirements of material demand time and delivery route planning, the ground execution system is precisely deployed to realize timely material delivery and intelligent assisted assembly;

[0008] (4) Based on the requirements of complex product assembly tasks, determine the automated assembly tasks of the wire-controlled execution system, the guided assembly system, and the ground execution system.

[0009] Furthermore, the flexible arrangement of the wire-controlled execution system in step (1) is specifically implemented as follows:

[0010] (11) The wire-controlled execution system includes a control system, truss, pulleys, ropes, base, end effector, etc. The end effector is connected to tooling fixtures, cameras and sensors through an adapter or standard interface;

[0011] (12) The wire-controlled execution system transmits instructions through the control system to control the rope to move up and down. The rope drives the end effector to achieve balance under the combined action of the rope traction force, driving the parts to translate and rotate on the x, y, and z axes. The end effector moves point to point in the spatial coordinate system, which can be decomposed into linear motion along a specific straight line and rotational motion around a circle.

[0012] Linear motion utilizes assembly path information for linear interpolation, moving along one or more axes, or determining the movement from the starting point to the target point using absolute or incremental dimensions. Assume the assembly path distance of the end effector at a certain moment is X. c(i) feed rate F c(i) The acceleration is a c Run ΔT c After a period of time, X c(i+1) and F c(i+1) These represent the assembly path distance and feed rate at the next moment, respectively.

[0013] When the end effector is on a straight line with a constant velocity, the assembly path distance can be expressed as:

[0014] X c(i+1) =X c(i) +F c(i) ΔT c (1)

[0015] When the end effector is on a uniformly accelerated straight segment, the interpolation equations for the assembly path distance and velocity can be expressed as:

[0016]

[0017] F c(i+1) =F c(i) +a c ΔT c (3)

[0018] When the end effector is on a uniformly decelerated straight segment, the velocity interpolation equation can be expressed as:

[0019]

[0020] The rotational motion is a circular motion at a certain speed on a specified plane. Clockwise or counterclockwise circular interpolation is performed on the selected xoy, xoz, yoz principal planes. Assume the arc length of the end effector at a certain moment is l. c(i) radian H c(i) , lc(i+1) H c(i+1) Given the arc length and radians for the next moment, run ΔT. c Time, arc length l c(i+1) The linear velocity is v c(i+1) Arc length l c(i) The linear velocity is v c(i) angular velocity is w c(i) Angular acceleration is a c(i) radius R c For the end effector on a uniformly accelerated straight segment, the interpolation equations for arc length and velocity can be expressed as:

[0021]

[0022] When the end effector is on a uniformly decelerated straight segment, the velocity interpolation equation can be expressed as:

[0023]

[0024] (13) By designing the assembly task state and action space, end effector position, obstacle initialization position and sparse reward function, the deep SARSA algorithm in reinforcement learning is used to perform path planning and solve the corresponding motion trajectory for the wire-controlled execution system, so as to realize the intelligent decision-making of the wire-controlled execution system according to different environments and control the assembly path and assembly sequence of parts.

[0025] The control system is configured to control the rotation angle β and speed of the eight motors at any given time. 8 rope lengths l and elongation End effector spatial position x c ,y c ,z c and attitude information and obstacle information x d ,y d ,z d ,R d , where R d The radius of the obstacle, together with the other two elements, constitutes the current state of the drive-by-wire system in a 42-dimensional real space, namely:

[0026]

[0027] By randomly defining the workspace of the online control execution system, the end effector is set to reach the target spatial position point o = [x]. o ,y o ,z o ] T The threshold ε is used to determine whether the end effector has reached the target position.

[0028] f o (s)=bool(||opc ||) (9)

[0029] At time s, the actuator position coordinate is p. c =[x c ,y c ,z c When the Euclidean distance between two points Complete the obstacle avoidance task;

[0030] The obstacle is set as a cuboid placed on the spatial surface, with a height H∈[0m,2m], length and width L∈[0m,0.5m], and W∈[0m,0.5m]. During the movement of the wire-controlled execution system, interference from obstacles, ropes, and the end position of the end effector must be avoided.

[0031] The obstacle avoidance action is based on the motor response within the control system, thus adapting to different operating conditions. The output of the algorithm is the sum of the elongations of the eight motors (Δβ) and the eight cables (Δl).

[0032] a t =[Δβ1,Δβ2,Δβ3,Δβ4,Δβ5,Δβ6,Δβ7,Δβ8,Δl1,Δl2,Δl3,Δl4,Δl5,Δl6,Δl7,Δl8] (10)

[0033] Set the output range of 16 real numbers to [-1, 1], and the shift increment to k. β The rope increment k controls the movement of obstacles that are actually acted upon by the wire-controlled actuator. l Controlling rope changes;

[0034] The drive-by-wire actuator is a redundant constraint positioning mechanism, suitable for both high-speed and low-speed assembly operations. The state transition function of the drive-by-wire system can be described by kinematic and dynamic equations.

[0035] s t+1 =f kinematic (s t +a t )+f Dynamic (s t +a t (11)

[0036] The time step of one movement of the drive-by-wire system is set to t. step ∈[0,0.005s], meaning the motor rotates once within any time interval from 0 to 0.05 milliseconds. The deep learning policy function maintains the same action for 30 consecutive time steps, maintaining the stability of the motion estimate, i.e., maintaining the motion for n steps. step =30, maximum allowed number of steps n max=200, meaning the end effector reaches the target position, ending the current motion assembly trajectory planning. Therefore, the maximum allowable time for the wire-controlled actuator is t = t_wire. step ·n keep ·n max .

[0037] When the end effector reaches the target point and completes the task, the reward is set to 1; otherwise, the reward is -1. If interference occurs with a collision object, the reward is 100. The assembly trajectory planning ends when the maximum allowable duration or when rope interference occurs.

[0038]

[0039] p e For time s t+1 End effector position coordinates, p e =[x e ,y e ,z e ].

[0040] (14) Tooling fixtures lift parts from the material area to the assembly area through actions such as adsorption, picking up, clamping, handling, and placing;

[0041] (15) The camera identifies the assembled parts and guides them to the assembly position. The sensor detects data such as position, pressure and force, and achieves hole alignment and mounting surface fit through impedance control.

[0042] As one of the preferred embodiments of the present invention, in step (1), since the rope can be freely wound and released and the pulley has a large rotation range, the working space of the wire-controlled execution system ranges from a few meters to hundreds of meters or even thousands of meters; since the rope and truss are detachable and reconfigurable, the wire-controlled execution system has flexible characteristics; since the wire-controlled execution system has low inertial force and the rope is lightweight, the end effector can obtain a large speed and a high load-bearing capacity.

[0043] Furthermore, the mobile deployment and guidance assembly system in step (2) is specifically implemented as follows:

[0044] (21) The guided assembly system includes a control system and a laser projection device, etc.

[0045] (22) The laser projection equipment has corresponding customized software that can import constructed CAD three-dimensional models, such as typical components such as cabins and bottoms of complex products;

[0046] (23) The projection information includes reference points, reference holes and assembly guidance information on typical component features. The projection information is used to construct a physical reference. The physical reference is compared with the CAD model to determine the reference error between the two. If the error exceeds the limit, the model is refitted.

[0047] (24) The control system projects the feature information of the part onto the surface of the assembled parts, positions the part so that it coincides with the laser beam, or guides the assembly of the part according to the assembly auxiliary information, and verifies the accuracy of the projected position by changing the coordinate system or by using traditional measurement methods.

[0048] Furthermore, the precise deployment of the ground execution system in step (3) is specifically implemented as follows:

[0049] (31) The ground execution system includes a control system, an intelligent handling robot, and an assembly robot. The assembly robot is connected to a camera and sensors.

[0050] (32) The ground execution system sends instructions through the control system to ensure the correct placement of parts. The intelligent handling robot picks up parts from the tray, transfers items between different workstations, and moves finished products to the next production stage according to the process completion time and material requirements list, delivering the required materials on time and to the correct location.

[0051] (33) The assembly robot is loaded by the intelligent assembly robot and moved to the designated assembly location to perform actions such as screw tightening, welding, bonding, and riveting.

[0052] (34) Use a camera to identify moving parts and projection information, and use sensors to feed back force, displacement and other data to detect parts, check assembly quality and measure dimensions to ensure that the assembly process meets quality standards.

[0053] The present invention also discloses the relationship between the wire-controlled execution system, the guided assembly system, and the ground execution system.

[0054] The guided assembly system uses key technologies such as projection system calibration and simulation to project brackets, supports and parts of complex products. The wired execution system identifies the laser light source through the camera, and the control system controls the motor to pull the rope. The rope controls the end effector to transport the parts to the laser projection location for assembly. The ground execution mechanism moves to the designated location through the control system, and the camera performs repairs or assists in assembly according to the laser indication mark.

[0055] Compared with the prior art, the present invention has the following beneficial effects: (1) The flexible wire-controlled execution system of the present invention does not require a large assembly space, is easy to disassemble and install, and can be flexibly arranged in the work site. Combined with a specially designed end effector, it can adapt to parts of different sizes and shapes, accurately control the position of the hoisted parts, and realize flexible assembly; (2) The laser projection system can complete the product positioning and installation with high precision, reduce the error caused by manual marking and benchmarking, and reduce the labor intensity of operators and inspectors; (3) The ground execution system can deliver the assembly parts on time, ensure that there is no collision on the path, and accurately complete the assembly auxiliary actions; (4) The combined operation of the three systems is suitable for most assembly operation scenarios, effectively reducing labor costs and improving the automation and intelligence level of the assembly process. Attached Figure Description

[0056] Figure 1 This invention relates to an assembly unit based on a wire-controlled robot.

[0057] Figure 2 A schematic diagram of the structure of a wire-controlled actuator for hoisting parts in the material area;

[0058] Figure 3 A schematic diagram of a wire-controlled actuator hoisting parts in the assembly area;

[0059] Figure 4 A schematic diagram of the laser projection-guided assembly system.

[0060] Figure 5 This is a schematic diagram of the ground execution system structure. Detailed Implementation

[0061] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way.

[0062] Example:

[0063] like Figure 1 As shown, an assembly unit based on a wire-controlled robot includes the following steps:

[0064] (1) Based on the workspace size of 8m×8m×8m (length×width×height), the wire-controlled execution system 1 is flexibly arranged to precisely control the part's pose, movement path and assembly sequence.

[0065] (2) Based on the installation scenario and the requirements for fastener assembly, the mobile deployment of the assembly system 2 is used to accurately complete product positioning and installation.

[0066] (3) Based on the requirements of material demand time and delivery route planning, the ground execution system 3 is precisely deployed to realize timely material delivery and intelligent auxiliary assembly;

[0067] (4) Based on the assembly task requirements of complex product 4, the automated assembly tasks of wire-controlled execution system 1, guided assembly system 2 and ground execution system 3 are determined as follows: wire-controlled execution system 1 hoists large structural plate test pieces, guided assembly system 2 projects support information, and ground execution system 3 performs screw tightening operations to complete the assembly of complex product 4 cabin.

[0068] Furthermore, in step (1), the wire-controlled execution system 1 is flexibly arranged, such as... Figure 2 As shown, the specific implementation is as follows:

[0069] (11) The wire-controlled execution system 1 includes a control system 1-1, a truss 1-2, a rope 1-3, a base 1-4, an end effector 1-5, etc. The end effector 1-5 is connected to the tooling fixture 1-6, the camera 1-7 and the sensor 1-8 through an adapter or a standard interface;

[0070] (12) The wire-controlled execution system 1 transmits instructions through the control system 1-1 to control the rope to move up and down. The rope 1-3 drives the end effector 1-5 to achieve balance under the combined action of the rope traction force, which drives the part 4-1-1 to translate and rotate on the x, y, and z axes, and controls the movement trajectory and assembly sequence of the part 4-1-1. The end effector 1-5 moves point-to-point in the spatial coordinate system, which can be decomposed into translational motion along a specific straight line and rotational motion around a circle.

[0071] Linear motion utilizes assembly path information for linear interpolation, moving along one or more axes, or determining the movement from the starting point to the target point using absolute or incremental dimensions, such as... Figure 3 As shown. Assume the assembly path distance of end effectors 1-5 at a certain moment is X. c(i) feed rate F c(i) The acceleration is a c Run ΔT c After a period of time, X c(i+1) and F c(i+1) These represent the assembly path distance and feed rate at the next moment, respectively.

[0072] When the end effectors 1-5 are on a straight line with uniform velocity, the assembly path distance can be expressed as:

[0073] X c(i+1) =X c(i) +F c(i) ΔT c (1)

[0074] When the end effectors 1-5 are on a uniformly accelerated straight segment, the interpolation equations for the assembly path distance and velocity can be expressed as:

[0075]

[0076] F c(i+1) =F c(i) +a c ΔT c (3)

[0077] When the end effectors 1-5 are on a uniformly decelerated straight segment, the velocity interpolation equation can be expressed as:

[0078]

[0079] The rotational motion is a circular motion at a certain speed on a specified plane. Clockwise or counterclockwise circular interpolation is performed on the selected xoy, xoz, yoz principal planes. Assume the arc length of the end effector 1-5 at a certain moment is l. c(i) radian H c(i) , l c(i+1) H c(i+1) Given the arc length and radians for the next moment, run ΔT. c Time, arc length l c(i+1) The linear velocity is v c(i+1) Arc length l c(i) The linear velocity is v c(i) angular velocity is w c(i) Angular acceleration is a c(i) When the end effectors 1-5 are on a uniformly accelerated straight segment, the interpolation equations for arc length and velocity can be expressed as:

[0080]

[0081] When the end effectors 1-5 are on a uniformly decelerated straight segment, the velocity interpolation equation can be expressed as:

[0082]

[0083] (13) By designing the assembly task state and action space, the position of the end effector 1-5, the initial position of the obstacle and the sparse reward function, the deep SARSA algorithm in reinforcement learning is used to perform path planning for the wire-controlled execution system 1 and solve the corresponding motion trajectory, so as to realize the intelligent decision-making of the wire-controlled execution system 1 according to different environments and control the assembly path and assembly sequence of the parts.

[0084] The control system is configured to control the rotation angle β and speed of the eight motors at any given time. 8 ropes, length l and elongation of 1-3 End effector 1-5 spatial position x c ,y c ,z c and attitude information and obstacle information x d ,yd ,z d ,R d , where R d The radius of the obstacle, together forming the current state of the drive-by-wire system 1, is a 42-dimensional real space, namely:

[0085]

[0086] By randomly defining the workspace of the online control execution system 1, the target position point reached by the end effectors 1-5 is set to o = [x]. o ,y o ,z o ] T The threshold ε is used to determine whether the end effectors 1-5 have reached the target position.

[0087] f o (s)=bool(||op c ||) (9)

[0088] At time s, the coordinates of the end effector 1-5 are p. c =[x c ,y c ,z c When the Euclidean distance between two points Complete the obstacle avoidance task;

[0089] The obstacle is set as a cuboid placed on the spatial surface, with a randomly generated height H = 1.5m, length and width L = 0.5m, W = 0.5m. During the movement of the wire-controlled actuator 1, interference from the obstacle, the rope, and the end position of the end effector 1-5 should be avoided.

[0090] The obstacle avoidance action is based on the motor response within the control system, thus adapting to different working conditions. The output of the algorithm is the sum of the elongations Δβ of the eight motors and Δl of the eight ropes 1-3.

[0091] a t =[Δβ1,Δβ2,Δβ3,Δβ4,Δβ5,Δβ6,Δβ7,Δβ8,Δl1,Δl2,Δl3,Δl4,Δl5,Δl6,Δl7,Δl8] (10)

[0092] Set the output range of 16 real numbers to [-1, 1], and the shift increment to k. β The actual movement of the obstacle controlled by the wire-controlled actuator 1 is achieved by the rope increment k. l Controlling rope changes;

[0093] The wire-controlled actuator 1 is a redundant constraint positioning mechanism, applicable to both high-speed and low-speed assembly actions. The state transition function of the wire-controlled actuator system 1 can be described by kinematic and dynamic equations:

[0094] s t+1 =f kinematic (s t +a t )+f Dynamic (s t +a t (11)

[0095] The time step of one movement of the drive-by-wire system 1 is set to t. step ∈[0,0.005s], meaning the motor rotates once within any time interval from 0 to 0.05 milliseconds. The deep learning policy function maintains the same action for 30 consecutive time steps, maintaining the stability of the motion estimate, i.e., maintaining the motion for n steps. step =30, maximum allowed number of steps n max =200, meaning the end effector positions 1-5 reach the target position, ending the current motion assembly trajectory planning. Therefore, the maximum allowable time for the wire-controlled actuator is t = t step ·n keep ·n max .

[0096] When the end effector at positions 1-5 reaches the target point and completes the task, the reward is set to 1; otherwise, the reward is -1. If interference occurs with a collision object, the reward is 100. The assembly trajectory planning ends when the maximum allowable time or when rope interference occurs.

[0097]

[0098] p e For time s t+1 End effector position coordinates 1-5, p e =[x e ,y e ,z e ].

[0099] (14) Tooling fixtures 1-6 use actions such as adsorption, picking up, clamping, handling, and placement to lift part 4-1-1 from material area A to assembly area B. The assembly process is as follows: Figure 2 and 3 As shown;

[0100] (15) Camera 1-7 identifies assembly part 4-1-1 and visually guides assembly part 4-1-1 to the assembly position. Sensor 1-8 detects data such as force and realizes hole alignment through force feedback impedance control.

[0101] As one of the preferred embodiments of the present invention, in step (1), since the rope can be freely extended and retracted and the pulley has a large rotation angle range, the working space of the wire-controlled execution system is length × width × height = 8m × 8m × 8m; since the rope and truss are detachable and reconfigurable, the wire-controlled execution system has flexible characteristics; since the wire-controlled execution system has low inertial force and the rope is lightweight, the end effector can obtain a large speed and a high load-bearing capacity.

[0102] Furthermore, such as Figure 4 As shown, the mobile deployment and assembly system 2 in step (2) is specifically implemented as follows:

[0103] (21) The guided assembly system 2 includes a control system 2-1 and a laser projection device 2-2, etc.

[0104] (22) The laser projection equipment has corresponding customized software, which can import the constructed CAD three-dimensional model and select the typical component cabin 4-1 of the complex product;

[0105] (23) The projection information includes the reference points, reference holes and assembly guidance information on the support 4-1-2; the projection information is used to construct the physical reference of the support 4-1-2, and the physical reference is compared with the CAD model to determine the reference error between the two. The error does not exceed the limit range.

[0106] (24) The feature information of support 4-1-2 is projected onto the surface of assembly part 4-1-3 using control system 2-1. The position of support 4-1-2 is positioned to coincide with the laser beam. The accuracy of the projection position is verified by reprojecting using a different coordinate system.

[0107] Furthermore, such as Figure 5 As shown, the precise deployment of the ground execution system 3 in step (3) is specifically implemented as follows:

[0108] (31) The ground execution system 3 includes a control system 3-1, an intelligent handling robot 3-2, and an assembly robot 3-3 mounted on the intelligent handling robot. The assembly robot is connected to a camera 3-4 and a sensor 3-5;

[0109] (32) The ground execution system 3 sends instructions through the control system 3-1 to ensure the correct placement of parts. The intelligent handling robot picks up parts from the tray and transfers items between different workstations according to the process completion time and material requirements list, and moves the finished product to the next production stage, delivering the required materials on time and at the correct location.

[0110] (33) Assembly robot 3-3 is loaded by intelligent handling robot 3-2 and moved to the designated assembly location to achieve screw tightening action;

[0111] (34) Use camera 3-4 to identify hole position information and sensor 3-5 to provide force data, detect parts, check assembly quality, and measure dimensions to ensure that the assembly process meets quality standards.

[0112] The guided assembly system 2 utilizes key technologies such as projection system calibration and simulation to project the complex product 4's support 4-1-2 onto the laser projection. The wired execution system 1 identifies the laser source via camera 1-7, and the control system 1-1 controls the motor to pull the rope 1-3. The rope 1-3 controls the end effector 1-5 to transport part 4-1-1 to the laser projection location for assembly. The ground execution mechanism 3 moves to the designated location via control system 3-1, and the camera 3-4 assists in assembly based on the laser indicator. The specific implementation process is as follows: Figure 2 and 3 As shown.

[0113] It should be noted that the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Those skilled in the art can make various modifications and improvements without departing from the concept of the present invention. Therefore, any combination or equivalent transformation made based on the above embodiments falls within the scope of protection of the present invention.

Claims

1. An assembly unit based on a wire-controlled robot, characterized in that, The assembly unit includes a wire-controlled execution system, a guided assembly system, and a ground execution system. The guided assembly system uses projection system calibration and simulation key technologies to project the brackets, supports, and parts of complex products. The wire-controlled execution system identifies the laser light source through a camera, controls the motor to pull the rope, and the rope controls the end effector to transport the parts to the laser projection location for assembly. The ground execution mechanism moves to the designated location through the control system, and the camera performs repairs or assists in assembly based on the laser indication mark. The specific assembly includes the following steps: (1) Based on the size of the workspace, the wire control execution system is flexibly arranged to accurately control the part position, assembly path and assembly sequence; (2) Based on the installation scenario and the assembly requirements of the fasteners, the mobile deployment of the assembly system is used to accurately complete the product positioning and installation; (3) Based on the material demand time and delivery route planning requirements, accurately deploy the ground execution system to achieve timely material delivery and intelligent assisted assembly; (4) Based on the requirements of complex product assembly tasks, determine the automated assembly tasks of the wire-controlled execution system, the guided assembly system, and the ground execution system; The precise deployment of the ground execution system in step (3) is specifically implemented as follows: (31) The ground execution system includes a control system, an intelligent handling robot, and an assembly robot. The assembly robot is equipped with cameras and sensors. (32) The ground execution system sends instructions through the control system to ensure the correct placement of parts. The intelligent handling robot picks up parts from the trays and transfers items between different workstations according to the process completion time and material requirements list, and moves the finished products to the next production stage, delivering the required materials on time and to the correct location. (33) The assembly robot is loaded by the intelligent assembly robot and moved to the designated assembly location to perform screw tightening, welding, bonding, and riveting actions. (34) Use a camera to identify moving parts and projection information, and use sensors to feed back force and displacement data to detect parts, check assembly quality and measure dimensions to ensure that the assembly process meets quality standards.

2. The assembly unit based on a wire-controlled robot according to claim 1, characterized in that, The flexible arrangement of the wire-controlled execution system in step (1) is specifically implemented as follows: (11) The wired execution system includes a control system, truss, pulleys, ropes, base, and end effector, wherein the end effector is connected to tooling fixtures, cameras and sensors via adapters or standard interfaces; (12) The wire-controlled execution system transmits commands through the control system to control the rope to move up and down. The rope drives the end effector to achieve balance under the combined action of the rope traction force, driving the parts to translate and rotate on the x, y, and z axes. The end effector moves point-to-point in the spatial coordinate system, which can be decomposed into linear motion along a specific straight line and rotational motion around a circle. Linear motion utilizes assembly path information for linear interpolation, moving along one or more axes, or determining the movement from the starting point to the target point using absolute or incremental dimensions. Let the assembly path distance of the end effector at a certain moment be... feed rate acceleration is ,run After a period of time, and These represent the assembly path distance and feed rate at the next moment, respectively. When the end effector is on a straight line with a constant velocity, the assembly path distance can be expressed as: （1） When the end effector is on a uniformly accelerated straight segment, the interpolation equations for the assembly path distance and velocity can be expressed as: （2） （3） When the end effector is on a uniformly decelerated straight segment, the velocity interpolation equation can be expressed as: （4） Rotational motion involves circular motion at a certain speed on a specified plane. Perform clockwise or counterclockwise circular interpolation on the principal plane, and let the arc length of the end effector at a certain moment be... ,radian , , The arc length and radians are respectively for the next moment, and the operation is as follows. Time, arc length The linear velocity is arc length The linear velocity is angular velocity is angular acceleration is ,radius For the end effector on a uniformly accelerated straight segment, the interpolation equations for arc length and velocity are expressed as: （5） （6） When the end effector is on a uniformly decelerated straight segment, the velocity interpolation equation can be expressed as: （7） (13) By designing the assembly task state and action space, end effector position, obstacle initialization position and sparse reward function, the deep SARSA algorithm in reinforcement learning is used to perform path planning and solve the corresponding motion trajectory for the wire-controlled execution system, so as to realize the intelligent decision-making of the wire-controlled execution system according to different environments and control the assembly path and assembly sequence of parts; The control system is configured to rotate by 8 motors at any given time. and rotational speed 8 rope lengths and elongation Spatial position of the end effector and attitude information and obstacle information ,in The radius of the obstacle, together with the other two elements, constitutes the current state of the drive-by-wire system in a 42-dimensional real space, namely: （8） By randomly defining the workspace of the online control execution system, the target position point reached by the end effector is set as follows: Using threshold Determine whether the end effector has reached the target position, i.e. （9） At any moment The position coordinates of the end effector are When the Euclidean distance between two points Complete the obstacle avoidance task; The obstacle is set as a cuboid placed on the surface of space, with its height randomly generated. Length and width , During the movement of the wire-controlled actuator, interference from obstacles, ropes, and the end position of the end effector must be avoided; Obstacle avoidance actions are based on the response of the motors within the control system, thus adapting to different operating conditions. (8 motors) And the elongation of the 8 ropes As the output of the algorithm, that is: （10） Set the output range of 16 real numbers to [-1, 1], and move the increment... Controlling the movement of obstacles that actually affect the drive-by-wire system, cable increment Controlling rope changes; The drive-by-wire actuator is a redundant constraint positioning mechanism, suitable for both high-speed and low-speed assembly operations. The state transition function of the drive-by-wire system is described by kinematic and dynamic equations. （11） The time step of one movement of the drive-by-wire system is set to... This means that the motor rotates once within any time interval of 0-0.05 milliseconds, and the deep learning policy function maintains the same action for 30 consecutive time steps, maintaining the stability of the motion estimate, i.e., maintaining the number of steps is... Maximum number of steps allowed That is, the end effector reaches the target position, ending the current motion assembly trajectory planning. Therefore, the maximum allowable time for the wire-controlled actuator is... , When the end effector reaches the target point and completes the task, the reward is set to 1; otherwise, the reward is -1. If interference occurs with a collision object, the reward is 100. The assembly trajectory planning ends when the maximum allowable duration or when rope interference occurs. （12） For at any time End effector position coordinates , (14) Tooling fixtures lift parts from the material area to the assembly area through adsorption, picking, clamping, handling and placement actions; (15) The camera identifies the assembled parts and guides the assembled parts to the assembly position. The sensor detects data such as position, pressure and force, and realizes hole alignment and mounting surface fit through impedance control.

3. The assembly unit based on a wire-controlled robot according to claim 1, characterized in that, The mobile deployment and assembly system in step (2) is implemented as follows: (21) The guided assembly system includes a control system and a laser projection device; (22) The laser projection equipment has corresponding customized software to import and construct CAD three-dimensional models, including typical components such as cabins and bottoms of complex products; (23) The projection information includes reference points, reference holes and assembly guidance information on typical component features. The projection information is used to construct a physical reference. The physical reference is compared with the CAD model to determine the reference error between the two. If the error exceeds the limit, the model is refitted. (24) The control system projects the feature information of the part onto the surface of the assembled parts, positions the part so that it coincides with the laser beam, or guides the assembly of the part according to the assembly auxiliary information, and verifies the accuracy of the projected position by changing the coordinate system or by using traditional measurement methods.

Images