Method and system for flexibly disassembling double shaft holes of industrial robot

Abstract

The invention discloses a method and a system for flexibly disassembling double shaft holes of an industrial robot. The method comprises the following steps: constructing three mechanism models for disassembling the double shaft holes of the industrial robot based on mathematical description; constructing a near-end strategy optimization-based industrial robot double-shaft-hole flexible disassembly strategy model and an online training environment thereof; according to dynamic data in an online training environment, setting a state, an action, a reward item and a penalty item suitable for a double-shaft-hole flexible disassembly problem of the industrial robot, and training a double-shaft-hole flexible disassembly strategy model of the industrial robot, one penalty item being calculated based on values of three mechanism models in a disassembly process; and according to the industrial robot double-shaft-hole smooth disassembling strategy model, smooth disassembling of the double shaft holes is completed. According to the method, the disassembling process of the double shaft holes can be optimized, the applicability to the batch disassembling process of the double shaft holes is improved, and then the generalization ability of the double shaft hole disassembling strategy of the industrial robot is effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the cross-research field of intelligent manufacturing combined with artificial intelligence, and in particular to a double-shaft-hole compliant disassembly method and system for industrial robots based on near-end policy optimization under dynamic parameters. BACKGROUND

[0002] In recent years, with the continuous acceleration of China's industrialization process, the product update iteration speed is continuously improving, and a large number of waste products will be generated in China every year. If the large number of waste products are not properly handled, it will not only cause great waste of resources, but also cause serious environmental pollution. Recycling waste products and making full use of the residual value of waste products can significantly reduce environmental pollution. At the same time, if the value contained in the waste products is fully utilized and re-injected into the production and manufacturing process of new products, the resource utilization rate can be greatly improved, the manufacturing cost of new products can be effectively reduced, and it has great significance for the sustainable development of China's manufacturing industry.

[0003] Disassembly is the primary step to realize the recycling of waste products, and industrial robot disassembly is an important means to promote the intelligent development of waste products. According to the different types of parts, the waste product disassembly process includes gear shaft disassembly, single-shaft-hole disassembly, double-shaft-hole disassembly, etc. Among them, double-shaft-hole disassembly is a common link in the waste product disassembly process. When disassembling the double-shaft-hole with the industrial robot, if the motion control parameters of the industrial robot disassembly process are not reasonably set, the disassembly force / torque of the industrial robot may be too large, which may cause damage to the end effector of the industrial robot or the double-shaft-hole, and thus cause damage to the industrial robot or the double-shaft-hole cannot be recovered.

[0004] Therefore, effectively reducing the disassembly force / torque of the industrial robot disassembly double-shaft-hole process and realizing the compliant disassembly of the industrial robot double-shaft-hole play an important role in promoting the intelligence of the waste product disassembly process and improving the economic benefits.

[0005] Due to the different factors such as user habits, part maintenance strategies, etc. during the product service process, the various physical parameters (double-shaft-hole hole distance, double-shaft-hole radius, inclination angle, etc.) of the batch waste products in the disassembly process have dynamic characteristics. At the same time, in the waste product disassembly process, such physical parameters have a very obvious influence on the disassembly force / torque. If the dynamic nature of these parameters is not considered, the compliant disassembly strategy under the expected static parameters is used to realize the compliant disassembly of the double-shaft-hole under the dynamic parameters, which may cause the problem that the expected scheme is difficult to apply. Therefore, under the condition of realizing the batch double-shaft-hole robot disassembly, proposing a compliant disassembly strategy for the industrial robot double-shaft-hole under dynamic parameters can improve the applicability of the batch double-shaft-hole disassembly process, and thus effectively improve the generalization ability of the industrial robot double-shaft-hole disassembly strategy. SUMMARY

[0006] The present application mainly aims to provide a dynamic parameter-based near-end strategy optimization industrial robot double shaft hole flexible disassembly method and system which can effectively improve the generalization ability of the double shaft hole disassembly strategy of the industrial robot.

[0007] The technical scheme adopted by the present application is:

[0008] A dynamic parameter-based near-end strategy optimization industrial robot double shaft hole flexible disassembly method is provided, comprising the following steps:

[0009] Three mechanism models of the industrial robot double shaft hole disassembly based on mathematical description are constructed, wherein the first mechanism model describes the distance moved by the single-sided shaft during the entire two-point contact process in the single-sided hole; the second mechanism model describes the distance moved by the double shaft during the entire two-point contact process on the outside of the double hole; and the third mechanism model describes the distance moved by the double shaft during the entire two-point contact process on the inside of the double hole.

[0010] A near-end strategy optimization industrial robot double shaft hole flexible disassembly strategy model and its online training environment are constructed, wherein the online training environment construction process is: constructing a double shaft and double hole three-dimensional model in a simulation environment, deploying an industrial robot model to the simulation environment, fixing the double shaft model on the force / torque sensor at the end of the industrial robot; by specifying the target position and attitude, controlling the movement of the target point in the process of disassembling the double shaft from the double hole by the industrial robot, and obtaining the dynamic parameters in the disassembly process, calculating the values of the three mechanism models in the disassembly process according to the dynamic parameters;

[0011] According to the dynamic data in the online training environment, the state, action, reward item and penalty item suitable for the industrial robot double shaft hole flexible disassembly problem are set, and the industrial robot double shaft hole flexible disassembly strategy model is trained, wherein one of the penalty items is calculated based on the values of the three mechanism models in the disassembly process.

[0012] According to the industrial robot double shaft hole flexible disassembly strategy model, the flexible disassembly of the double shaft hole is completed.

[0013] According to the above technical scheme, the construction process of the three mechanism models is: according to the force balance relationship between the double shaft and the double hole in the process of disassembling the double shaft hole by the industrial robot under the quasi-static condition, the mathematical relationship between various structural elements is described, and then the three mechanism models are constructed according to the mathematical relationship.

[0014] According to the above technical scheme, the construction process of the double shaft and double hole three-dimensional model is: creating a cube model and a cylinder model in the simulation environment, and constructing the double shaft and double hole three-dimensional model by combining the cylinder and the cube.

[0015] According to the technical scheme, in the online training environment construction process, the inverse kinematics model of the UR5 robot is constructed, and the movement of the industrial robot in the disassembly process is controlled by specifying the target position and attitude.

[0016] According to the technical scheme, the penalty term in the online training process of the industrial robot double-shaft hole compliant disassembly strategy model further includes a force and torque penalty term, a relative position amount penalty term of the double-shaft holes in the x and y directions, and a relative deflection angle amount penalty term of the double-shaft holes in the x, y and z directions.

[0017] According to the technical scheme, the industrial robot double-shaft hole compliant disassembly strategy model is continuously updated, and the updated parameters are used for continuous training.

[0018] According to the technical scheme, in the training process, the effect of the current disassembly strategy model is evaluated by estimating the advantage function, and the disassembly strategy model is guided to optimize and improve, and in the optimization and improvement process, a clipping mechanism is introduced to limit the amplitude of strategy update, so as to ensure the smoothness in the strategy change process.

[0019] According to the technical scheme, while fixing the double-shaft model on the force / torque sensor at the end of the industrial robot, a virtual target point is generated and fixed on the double-shaft center line.

[0020] The application also provides an industrial robot double-shaft hole compliant disassembly system based on a proximal strategy optimization under dynamic parameters, comprising:

[0021] A mechanism model construction module is configured to construct three mechanism models of industrial robot double-shaft hole disassembly based on mathematical description, wherein the first mechanism model describes the distance moved by a single-sided shaft in the whole process of always two-point contact in a single-sided hole; the second mechanism model describes the distance moved by a double-shaft in the whole process of always two-point contact on the outside of a double hole; and the third mechanism model describes the distance moved by a double-shaft in the whole process of always two-point contact on the inside of a double hole.

[0022] A disassembly strategy model construction and training module is configured to construct an industrial robot double-shaft hole compliant disassembly strategy model based on a proximal strategy optimization and an online training environment thereof; wherein the online training environment construction process comprises: constructing a double-shaft and double-hole three-dimensional model in a simulation environment, deploying an industrial robot model to the simulation environment, and fixing a double-shaft model on a force / torque sensor at the end of the industrial robot; controlling the movement of a target point in the process of disassembling the double-shaft from the double hole by the industrial robot by specifying the target position and attitude, and obtaining dynamic parameters in the disassembly process; and calculating the values of the three mechanism models in the disassembly process according to the dynamic parameters.

[0023] A compliant disassembly module is configured to complete the compliant disassembly of the double-shaft hole according to the industrial robot double-shaft hole compliant disassembly strategy model.

[0024] The application also provides a computer storage medium, which stores a computer program executable by a processor, and the computer program executes the industrial robot double-axis hole compliant disassembly method based on a proximal policy optimization under dynamic parameters.

[0025] The application has the following beneficial effects: the industrial robot double-axis hole compliant disassembly strategy model constructed by the application is based on a proximal policy optimization algorithm (PPO), and the algorithm is used to solve the continuous action space control problem; the application pre-constructs mechanism models of three situations in which the disassembly process of the double-axis hole is prone to cause excessive disassembly force of the industrial robot, and calculates the value of the mechanism model in the online training environment to calculate one of the penalty terms of the training disassembly strategy model, so as to avoid the three situations in which the disassembly force is excessive and optimize the disassembly process of the double-axis hole.

[0026] Further, in view of the low compliance of the industrial robot double-axis hole disassembly process, the data interaction method between the double-axis hole disassembly simulation environment and the deep reinforcement learning algorithm is combined, and in the online training process, the disassembly force / torque of the double-axis hole disassembly process is reduced through the force and torque penalty terms, so as to effectively improve the compliance of the industrial robot double-axis hole disassembly process.

[0027] Further, in the online training environment construction process, the inverse kinematics model of the UR5 robot is constructed, the motion of the industrial robot disassembly process is controlled by specifying the target position and attitude, and the double-axis hole disassembly data used for training is generated.

[0028] Further, in the training process, the effect of the current disassembly strategy model is evaluated through the estimation of the advantage function, and the disassembly strategy model is guided to optimize and improve, and in the optimization and improvement process, the clipping mechanism is introduced to effectively limit the amplitude of the strategy update, and the smoothness in the strategy change process is ensured.

[0029] Of course, implementing any product of the application does not necessarily need to achieve all the advantages described above. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0031] Figure 1Ais a flow chart of a dual-axis hole compliant disassembly method of an industrial robot based on a proximal policy optimization under dynamic parameters according to an embodiment of the present application;

[0032] Figure 1B is a flow chart of a dual-axis hole compliant disassembly method of an industrial robot based on a proximal policy optimization under dynamic parameters according to another embodiment of the present application;

[0033] Figure 2 is a schematic diagram of a dual-axis hole compliant disassembly model according to an embodiment of the present application;

[0034] Figure 3 is a schematic diagram of four contact states of a dual-axis hole disassembly according to the present application;

[0035] Figure 4 is a schematic diagram of an initial state of a dual-axis hole disassembly according to the present application;

[0036] Figure 5 is a schematic diagram of a two-point contact state of a dual-axis hole disassembly according to the present application;

[0037] Figure 6 is a schematic diagram of a single-axis hole two-point contact according to the present application;

[0038] Figure 7 is a schematic diagram of an outside two-point contact according to the present application;

[0039] Figure 8 is a schematic diagram of an inside two-point contact according to the present application;

[0040] Figure 9 is a schematic diagram of a structure framework of a dual-axis hole compliant disassembly strategy of an industrial robot based on a proximal policy optimization according to the present application;

[0041] Figure 10 is a schematic diagram of a relationship between a dual-axis hole disassembly force / torque and a depth under dynamic parameters of test group 1 according to the present application. DETAILED DESCRIPTION

[0042] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0043] It should be noted that the diagrams provided in the embodiments of the present application only schematically illustrate the basic concepts of the present application, and thus only the components related to the present application are shown in the diagrams, rather than the number, shape and size of the components when actually implemented. The shapes, number and proportions of the components when actually implemented can be arbitrarily changed, and the layout pattern of the components can also be more complex.

[0044] In the present application, it also needs to be explained that, as the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" and the like appear, the indicated orientation or positional relationship is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the indicated device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, as the terms "first", "second" appear, they are only for description and differentiation purposes, and cannot be understood as indicating or implying relative importance.

[0045] In addition, it also needs to be explained that the features of various embodiments of the present application can be combined or integrated partially or entirely, and can interact and operate in different ways as understood by those skilled in the art. Each embodiment can be implemented independently of each other, or in an associated relationship.

[0046] As shown in Figure 1A The industrial robot double-shaft hole compliant disassembly method based on near-end policy optimization under dynamic parameters according to the embodiment of the present application comprises the following steps:

[0047] S1, three mechanism models of industrial robot double-shaft hole disassembly based on mathematical description are constructed, wherein the first mechanism model describes the distance moved in the whole process of always two-point contact of single-sided shaft in single-sided hole; the second mechanism model describes the distance moved in the whole process of always two-point contact of outer side of double shaft in double hole; and the third mechanism model describes the distance moved in the whole process of always two-point contact of inner side of double shaft in double hole;

[0048] S2, an industrial robot double-shaft hole compliant disassembly strategy model based on near-end policy optimization and its online training environment are constructed; wherein the online training environment construction process is: constructing a double-shaft and double-hole three-dimensional model in a simulation environment, deploying an industrial robot model to the simulation environment, fixing a double-shaft model on a force / torque sensor at the end of the industrial robot; by specifying a target position and attitude, controlling the movement of the target point in the process of disassembling the double shaft from the double hole by the industrial robot, and obtaining dynamic parameters in the disassembly process, calculating the values of the three mechanism models in the disassembly process according to the dynamic parameters;

[0049] S3, according to the dynamic data in the online training environment, the state, action, reward item and penalty item suitable for the industrial robot double-shaft hole compliant disassembly problem are set, and the industrial robot double-shaft hole compliant disassembly strategy model is trained, wherein one of the penalty items is calculated based on the values of the three mechanism models in the disassembly process;

[0050] S4, the compliant disassembly of the double-shaft hole is completed according to the industrial robot double-shaft hole compliant disassembly strategy model.

[0051] The present application is directed to the dynamic nature of the surface friction coefficient, hole distance, shaft radius and hole radius and other parameters of the dual shaft hole in the disassembly process of a large number of different batches, which causes the problem that industrial robots are difficult to disassemble a large number of different batches of dual shaft holes flexibly. The dual shaft hole flexible disassembly strategy of the industrial robot based on the proximal strategy optimization under the dynamic parameters can effectively improve the dual shaft hole disassembly efficiency and generalization, and promote the intelligent development of the waste product disassembly process. Considering the dynamic nature of the surface friction coefficient, hole distance, shaft radius and hole radius and other parameters of the dual shaft hole, the present application constructs a mechanism model of the industrial robot disassembling the dual shaft hole based on mathematical description, analyzes the two-point contact range of the dual shaft hole disassembly process under different parameters, and provides a theoretical basis for the flexible disassembly strategy of the industrial robot dual shaft hole. Based on this, combined with the dual shaft hole disassembly simulation environment, a deep reinforcement learning algorithm based on proximal strategy optimization is used to construct a flexible disassembly strategy model of the industrial robot dual shaft hole. The present application can provide theoretical and technical support for promoting the intelligent degree of the waste product disassembly process.

[0052] The proximal strategy optimization algorithm (Proximal Policy Optimization, PPO) of the above embodiment is a typical deep reinforcement learning algorithm mainly used to solve continuous action space control problems. The present application applies it to the problem of the flexible disassembly strategy of the industrial robot dual shaft hole under dynamic parameters. The flexible disassembly strategy of the industrial robot dual shaft hole can dynamically adjust the motion control parameters of the robot according to the disassembly force / torque and the robot end pose perceived by the force sensor in the dual shaft hole disassembly process, so that the disassembly force / torque in the dual shaft hole disassembly process is as small as possible. Combined with the state data fed back in the industrial robot dual shaft hole disassembly process, the agent slowly learns how to flexibly disassemble the dual shaft hole, and then constructs the flexible disassembly strategy of the industrial robot dual shaft hole under dynamic parameters.

[0053] Further, the present application pre-constructs mechanism models of the dual shaft hole under three situations that are prone to cause the disassembly force of the industrial robot to be too large in the disassembly process, and calculates the mechanism model value in the online training environment to affect one of the penalty items of the training disassembly strategy model, so as to avoid the three situations that the disassembly force is too large as much as possible, and optimize the disassembly process of the dual shaft hole.

[0054] Further, the construction process of the three mechanism models is: according to the force balance relationship between the dual shaft and the dual hole in the process of the industrial robot disassembling the dual shaft hole under the quasi-static condition, the mathematical relationship between various structural elements is described, and then three mechanism models are constructed according to the mathematical relationship.

[0055] Further, the construction process of the dual shaft and dual hole three-dimensional model is: creating a cube model and a cylinder model in the simulation environment, and constructing a dual shaft and dual hole three-dimensional model by combining the cylinder and the cube.

[0056] Further, in the online training environment construction process, the inverse kinematics model of the UR5 robot is constructed, and the industrial robot is controlled to move in the disassembly process by specifying the target position and attitude.

[0057] Further, the penalty term in the online training process of the industrial robot double-axis hole compliant disassembly strategy model also includes a force and torque penalty term, a relative position quantity penalty term in the x and y directions between the double-axis holes, and a relative deflection angle quantity penalty term in the x, y and z directions between the double-axis holes.

[0058] Further, the industrial robot double-axis hole compliant disassembly strategy model is continuously updated, and the updated parameters are used for continuous training.

[0059] Further, in the training process, the effect of the current disassembly strategy model is evaluated by estimating the advantage function, and the disassembly strategy model is guided to optimize and improve, and in the optimization and improvement process, a clipping mechanism is introduced to limit the amplitude of strategy update, so as to ensure the smoothness in the strategy change process.

[0060] Further, while fixing the double-axis model on the force / torque sensor at the end of the industrial robot, a virtual target point is generated and fixed on the double-axis center line.

[0061] In another embodiment of the present application, based on the mathematical relationship between the structures of various elements in the double-axis hole, a first mechanism model Lh1 in the two-point contact situation of the single-axis hole, a second mechanism model Lh2 in the two-point contact situation of the outer side, and a third mechanism model Lh3 in the two-point contact situation of the inner side are constructed; in view of the low compliance of the industrial robot double-axis hole disassembly process, a deep reinforcement learning algorithm based on the proximal policy optimization is adopted to train and generate a double-axis hole compliant disassembly strategy model by combining the data interaction method between the double-axis hole disassembly simulation environment and the deep reinforcement learning algorithm, thereby reducing the disassembly force / torque in the double-axis hole disassembly process and effectively improving the compliance of the industrial robot double-axis hole disassembly process.

[0062] As shown in Figure 1B The dynamic parameter-based industrial robot double-axis hole compliant disassembly method based on proximal policy optimization of the embodiment mainly includes the following steps:

[0063] (1) Construct an industrial robot double-axis hole disassembly mechanism model based on mathematical description. Under the quasi-static condition, the force balance relationship between the double-axis and the double-hole in the industrial robot disassembly double-axis hole process is combined to describe the distance L C, the friction coefficient μ, the hole distance Ld, and according to the mathematical relationship between multiple factors, the industrial robot double-hole disassembly mechanism model based on mathematical description is constructed, including the first mechanism model Lh1 under the single-hole two-point contact condition, the second mechanism model Lh2 under the outer two-point contact condition, and the third mechanism model Lh3 under the inner two-point contact condition.

[0064] (2) An industrial robot double-hole compliant disassembly strategy model based on a deep reinforcement learning algorithm of a proximal policy optimization is proposed. Considering the dynamics of the industrial robot double-hole disassembly process parameters, a deep reinforcement learning algorithm environment for the industrial robot double-hole compliant disassembly strategy based on the proximal policy optimization is constructed by using the CoppeliaSim simulation environment, the state, action, and reward functions suitable for the industrial robot double-hole compliant disassembly problem are set, and the industrial robot double-hole compliant disassembly strategy model is trained. Thereafter, the double-hole parameters are dynamically adjusted, and the running effect of the industrial robot double-hole compliant disassembly strategy model under the dynamic parameters is analyzed.

[0065] In step (1), the distance L C from the compliant center of the industrial robot to the center of the double-hole bottom end is constructed C , the hole radius R / r, the friction coefficient μ, the hole distance Ld, and the mathematical relationship between the single-hole two-point contact distance Lh1, the outer two-point contact distance Lh2, and the inner two-point contact distance Lh3, and the industrial robot double-hole disassembly mechanism model is constructed, and the specific process is as follows:

[0066] The industrial robot double-hole compliant disassembly mechanism model is shown in Figure 2 , L C is the distance from the compliant center of the industrial robot to the center of the double-hole bottom end, R / r is the hole radius and the shaft radius respectively, Ld is the hole distance, O C represents the compliant center, K x and K θ represent the lateral stiffness and angular stiffness respectively. In the industrial robot double-hole disassembly process, there can be several contact states between the double shaft and the double hole, such as no contact, line contact, one-point contact, and two-point contact, as shown in Figure 3 . Among them, when the double-hole is in a two-point contact state, it is easy to cause the disassembly force of the industrial robot to be too large. The initial state of the industrial robot disassembling the double-hole is shown in Figure 4 , the initial inclination angle of the double shaft is θ0, the initial horizontal distance from the compliant center and the shaft bottom center to the hole axis is U0 and ε0 respectively, the initial disassembly depth of the double-hole is h0, the lateral error is δ0, and the angle error is β0.

[0067] When the industrial robot double-hole disassembly process is in a two-point contact state, it can be subdivided into the following three types: single-hole two-point contact, outer two-point contact, and inner two-point contact, as shown in Figure 5 .

[0068] ① Single-axis hole two-point contact

[0069] Single-axis hole two-point contact state of industrial robot disassembly biaxial hole process, as shown in Figure 6 During the biaxial hole disassembly process, the physical quantities such as the horizontal distance U of the compliant center to the hole axis, the horizontal distance ε of the shaft bottom center to the hole axis, and the shaft inclination angle θ satisfy the following mathematical relationship:

[0070] U+ε=L C sinθ (1)

[0071] U-U0+(ε-ε0)=L C (θ-θ0) (2)

[0072]

[0073] U-U0=L C (θ-θ0) (4)

[0074] When in single-axis hole two-point contact state, at small angle θ, the geometric constraint mathematical model of biaxial hole can be described by formula (5):

[0075]

[0076] When turning from two-point contact to one-point contact, combined with the force and torque generated by the compliant center, and the force and torque conditions of the biaxial hole in the disassembly process, the description of F x and M parameters can be obtained:

[0077] F x =-F N1 =-Kx(δ0+U0-U) (6)

[0078] M=(h-μr)F N1 =KxL C (δ0+U0-U)+K θ (β0+θ-θ0)(7)

[0079] Solving the above formulas (1)-(7) gives

[0080] αh 2 +βh+γ=0(8)

[0081] α=Kxδ0+KxL C θ0 (9)

[0082] β=(θ0-β0)(K θ -KxL C 2 -μrKxL C) - (L C + μr)Kx(L C β0+ δ0) - 2(R - r)KxL C (10)

[0083] γ = -2cR(K θ - KxL C 2 - μrKxL C ) (11)

[0084] The critical depth h of the single-axis hole two-point contact is two solutions of formula (8), and the two-point contact distance Lh1 can be described as:

[0085]

[0086] ② Outer two-point contact

[0087] The outer two-point contact of the double-axis hole can be equivalent to the two-point contact of the single-axis hole, as shown in Figure 7 , which satisfies the following relationship:

[0088]

[0089] During the disassembly process of the double-axis hole, the compliance center motion coupling relationship can be described by the following formula:

[0090] U + ε = L C sinθ (14)

[0091]

[0092] When in the outer two-point contact state, under a small angle θ, the geometric constraint mathematical model of the double-axis hole can be described by formula (17):

[0093]

[0094] When turning from two-point contact to one-point contact, combined with the force and torque generated by the compliance center, and the force and torque conditions of the double-axis hole during disassembly, the description of F x and M parameters can be obtained:

[0095] F x = -F N1 = -Kx(δ0+U0-U) (18)

[0096] M = (h - μr') F N1 = Kx L C (δ0+U0-U) + K θ (β0+θ-θ0) (19)

[0097] The above formulas (13)-(19) are obtained by combining the formulas

[0098] α′h 2 +β′h+γ′=0(20)

[0099] α′=Kx(δ0+c′R′-ε0)+KxL C θ0 (21)

[0100] β′=(θ0-β0)(K θ -KxL C 2 -μr′KxL C )-(L C +μr′)Kx(δ0+2c′R′-ε0) (22)

[0101] γ′=-2c′R′(K θ -KxL C 2 -μr′KxL C ) (23)

[0102] The critical depth h of the outer two-point contact is the two solutions of equation (20), and the two-point contact distance Lh2 in this state is:

[0103]

[0104] ③ Inner two-point contact

[0105] The inner two-point contact of the industrial robot biax hole disassembly process can be considered as an inverted single-axis hole. As shown in Figure 8 The shadow part is the equivalent shaft, which satisfies the following relationship:

[0106]

[0107] The motion coupling relationship of the compliance center of the industrial robot in the biax hole disassembly process is described by the following formula:

[0108] U+ε=L C sinθ (26)

[0109] ε=r″cosθ-R″ (27)

[0110] U-U0=L C (θ-θ0) (28)

[0111] When in the inner two-point contact state, at a small angle θ, the geometric constraint mathematical model of the biax hole can be described by equation (31):

[0112]

[0113] When the two-point contact is converted to one-point contact, the force and moment generated by the combined compliance center, as well as the force and moment of the biax during disassembly, can be obtained as F x and the description of M parameters:

[0114] F x = -F N2 = -Kx(δ0+U0-U) (30)

[0115] M = μR''F N2 = Kx(L C -h)(δ0+U0-U)+K θ (β0+θ-θ0)(31)

[0116] The above formulas (25)-(31) are obtained by combining the formulas:

[0117] α''h 2 +β''h+γ''=0 (32)

[0118] α''=Kx(δ0+L C θ0) (33)

[0119] β''=(θ0-β0)(K θ -KxL C 2 -μR''KxL C )-(L C +μR'')Kx(δ0+L C β0)+2c''R''KxL C (34)

[0120] γ''=-2c'R'(K θ -KxL C 2 -μr'KxL C ) (35)

[0121] The critical depth of the two-point contact inside the biax hole during the disassembly process of the industrial robot biax hole is the two solutions of formula (32), and the two-point contact distance Lh3 is:

[0122]

[0123] In step (2), in order to realize the data interaction of the industrial robot biax hole compliant disassembly simulation environment and the depth reinforcement learning algorithm, an online training environment for the depth reinforcement learning algorithm of the industrial robot multi-class biax hole compliant disassembly process is constructed, the running effect of the industrial robot biax hole compliant disassembly strategy model under dynamic parameters is analyzed, and the specific operation is as follows:

[0124] ① Implement data interaction between PyCharm software and the CoppeliaSim simulation environment. First, create a cube model and a cylinder model in the CoppeliaSim simulation environment. Construct a dual-axis, dual-hole 3D model by combining the cylinder and cube. Then, deploy the UR5 robot model to the CoppeliaSim simulation environment. Fix the dual-axis model to the force / torque sensor at the UR5 robot's end effector. Simultaneously, generate a dummy point UR5_target and fix it on the dual-axis central axis, making it a child of the dual-axis. Construct the inverse kinematics model of the UR5 robot, enabling control of the industrial robot's disassembly process by specifying the target position and orientation of UR5_target. In the CoppeliaSim simulation environment, call the sim.simxStart and sim.simxStartSimulation functions to initiate data interaction between the CoppeliaSim server and the PyCharm client. The sim.simxSetObjectPosition and sim.simxSetObjectQuaternion functions can specify the target's spatial coordinates and quaternions to control the target's position and orientation, respectively. In PyCharm, calling the `sim.simxSetObjectPosition` and `sim.simxSetObjectQuaternion` functions enables data transfer from the PyCharm environment to the CoppeliaSim simulation environment for dual-axis hole disassembly operations, controlling the movement of target points during the industrial robot disassembly process. The `sim.simxGetObjectPosition`, `sim.simxGetObjectOrientation`, and `sim.simxReadForceSensor` functions can read the target's spatial coordinates, angle, and force sensor data, respectively, thereby enabling the data transfer of dual-axis hole disassembly data (such as disassembly force / torque, relative position of the dual-axis holes, and relative attitude of the dual-axis holes) from the CoppeliaSim simulation environment to the PyCharm environment.

[0125] ② Establish a compliant disassembly strategy model for dual-axis holes in industrial robots based on proximal strategy optimization. The PPO algorithm's elements include state (s), action (a), and reward (r); the intelligent system determines the appropriate action based on the current state s. t Choose the optimal disassembly action a t Among them, state s t It is a one-dimensional vector of length 12, as shown in formula (37). Where rp x ,rp y and rp zra x, ra y, and ra z represent the relative position of the dual-axis hole in the x, y, and z directions, respectively, ra α , ra β , and ra γ represent the relative deflection angle of the dual-axis hole in the x, y, and z directions, respectively.f x , f y , f z , m x , m y , and m z are the forces and moments in the x, y, and z directions read by the force / moment sensor, P max , O max , F max , and M max are set to 3 mm, 0.03 rad, 100 N, and 3 N·m, respectively. The action is a one-dimensional vector of length 6, as shown in equation (38), po x , po y , po z , ao x , ao y , and ao z are the position and angle offsets of the robot end (UR5_target). The range of po x , po y is -0.1 mm ~ 0.1 mm, the range of po z is -1.5 mm ~ 1.5 mm, the range of ao x , ao y , and ao z is -0.005 rad ~ 0.005 rad. The reward is defined by equations (39)-(46). Where, P F and P M represent the penalty terms of force and moment, P p represents the penalty term of the relative position amount between the dual-axis holes in the x and y directions, P a represents the penalty term of the relative deflection angle amount between the dual-axis holes in the x, y, and z directions, P Lh represents the penalty term of the two-point contact distance of the dual-axis hole, R H represents the reward term of the dual-axis disassembly distance, R S represents the reward term of the successful dual-axis disassembly, Lh max is set to 0.5 m, h represents the disassembly depth, H represents the dual-hole depth, a n represents the sum of the number of actions within the current round. Lh1, Lh2, Lh3 are obtained from equations (9)-(12), (21)-(24), (33)-(36), respectively, where θ0=0 rad, δ0=2 mm, β0=0 rad, K x =4 N / mm, K θ= 30 Nmm / rad, L C , R, r, μ, Ld are determined by the biaxial hole parameters. If a n > 100, P F ≤ -1, P M ≤ -1, P p ≤ -1 or P a ≤ -1.2, it means that the disassembly task of the current round fails, and if h ≥ H, it means that the disassembly task of the current round succeeds, both of which need to make the end condition judgment information done = 1, and the environment of the PPO algorithm is reset, and a new disassembly process starts again.

[0126]

[0127] action(a) = [po x , po y , po z , ao x , ao y , ao z ] (38)

[0128] reward(r) = P F + P M + P p + P a + P Lh + R H + R S (39)

[0129]

[0130] The PPO algorithm can collect experience samples through the interaction between the agent and the environment, and optimize the policy based on the experience samples. The structural framework of the biaxial hole compliant disassembly strategy of the industrial robot based on the proximal policy optimization is shown in Figure 9 After the PPO algorithm completes the policy update, the used experience samples are discarded, and new sample data is collected using the updated policy to ensure that the data used in the training process is consistent with the current policy. In the PPO algorithm, the policy network is responsible for generating the action of the biaxial hole disassembly process of the industrial robot, and the evaluation network estimates the effect of the current policy network through the estimation of the advantage function, and guides the improvement of the policy network. The loss function of the PPO algorithm is shown in formula (47). Wherein, ∈ is the clipping coefficient, π old represents the policy with parameters θ old , and A is the advantage function. By introducing the clipping mechanism, the amplitude of the policy update can be effectively limited to ensure the smoothness of the policy change process.

[0131]

[0132] The hyperparameters used in the present patent are shown in Table 1.

[0133] Table 1 PPO hyperparameters

[0134]

[0135] First, the policy network generates an action policy a t through the current state s t . The industrial robot is controlled to perform a disassembly movement through the action policy a t , while the environment updates the state s t+1 and gives a reward calculation value r t , and stores the tuple <s t , a t , s t+1 , r t > in the experience pool. When the number of experiences in the experience pool reaches the batch sample number, the experiences are taken from the experience pool for neural network parameter update, and the updated policy network continues to generate an action policy.

[0136] ③ Realize the industrial robot double-axis hole compliant disassembly strategy under dynamic parameters.

[0137] Table 2 Double-axis hole parameters

[0138]

[0139] Based on the second group of algorithm parameters in Table 2, the industrial robot double-axis hole compliant disassembly strategy model is trained, and the converged deep reinforcement learning model is used to realize the compliant disassembly of the double-axis hole under different parameters (①, ④, ⑦, ⑩ in Table 2). The average value m_F and m_T of the absolute amount of force and torque in a single disassembly process, and the median mid_F and mid_T of the absolute amount of force and torque are measured. The converged strategy is used as test group 1, and the untrained strategy is used as test group 2. The results of the two test groups are shown in Table 3.

[0140] Table 3 Industrial robot double-axis hole disassembly data under dynamic parameters

[0141]

[0142] According to the results in Table 3, under the same parameters, the m_F, m_T, mid_F and mid_T of test group 1 are lower than those of test group 2, which shows that the converged deep reinforcement model helps to reduce the disassembly force and torque in the process of disassembling the double-axis hole of the industrial robot. The relationship between the double-axis hole disassembly force / torque and the depth of test group 1 under dynamic parameters is shown in Figure 10 , and the strategy model shows good compliance in the process of disassembling the double-axis hole under dynamic parameters.

[0143] The embodiment constructs an industrial robot double-shaft hole disassembly mechanism model, provides a theoretical basis for proposing an industrial robot double-shaft hole compliant disassembly strategy under dynamic parameters. A deep reinforcement learning algorithm environment of an industrial robot double-shaft hole compliant disassembly strategy based on a near-end strategy optimization is constructed by using a CoppeliaSim simulation environment, combined with state, action and reward elements suitable for the double-shaft hole compliant disassembly problem, a PPO algorithm-based training industrial robot double-shaft hole compliant disassembly strategy model is proposed, and the running effect of the industrial robot double-shaft hole compliant disassembly strategy model under dynamic parameters is analyzed, and the effectiveness of the proposed compliant disassembly strategy is proved.

[0144] The embodiment of the application is an industrial robot double-shaft hole compliant disassembly system based on near-end strategy optimization under dynamic parameters, comprising:

[0145] A mechanism model construction module is used to construct three mechanism models of industrial robot double-shaft hole disassembly based on mathematical description, wherein the first mechanism model describes the distance moved by a single-sided shaft in a single-sided hole during the whole process of always two-point contact; the second mechanism model describes the distance moved by a double-sided shaft in a double-sided hole during the whole process of always two-point contact on the outside; and the third mechanism model describes the distance moved by a double-sided shaft in a double-sided hole during the whole process of always two-point contact on the inside.

[0146] A disassembly strategy model construction and training module is used to construct an industrial robot double-shaft hole compliant disassembly strategy model based on near-end strategy optimization and an online training environment thereof; wherein the online training environment construction process is: constructing a double-shaft and double-hole three-dimensional model in a simulation environment, deploying an industrial robot model to the simulation environment, fixing a double-shaft model on a force / torque sensor at the end of the industrial robot; by specifying a target position and attitude, controlling the movement of the target point in the process of disassembling the double-shaft from the double-hole by the industrial robot, and obtaining dynamic parameters in the disassembly process, calculating the values of the three mechanism models in the disassembly process according to the dynamic parameters.

[0147] A compliant disassembly module is used to complete the compliant disassembly of the double-shaft hole according to the industrial robot double-shaft hole compliant disassembly strategy model.

[0148] Each module is mainly used to realize each step of the above method embodiment, which will not be described here.

[0149] The application also provides a computer readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card memory (for example, an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, a server, an App application store, and the like, which stores a computer program, and the program is executed by a processor to realize corresponding functions. The computer readable storage medium of the embodiment is executed by the processor to realize the method embodiment of the dynamic parameter based on the near-end strategy optimization of the industrial robot double-axis hole flexible disassembly method.

[0150] In summary, in view of the urgent needs of waste product recycling, and in view of the problem that the existing method is difficult to effectively construct the industrial robot double-axis hole flexible disassembly strategy under dynamic parameters, the application proposes an industrial robot double-axis hole flexible disassembly strategy based on near-end strategy optimization under dynamic parameters, constructs a mathematical description model of physical quantities such as double-axis hole friction coefficient and double-hole distance about the two-point contact distance of the double-axis hole in the process of industrial robot disassembly double-axis hole, and proposes an industrial robot double-axis hole flexible disassembly strategy based on near-end strategy optimization. The method provides theoretical support for promoting the intelligent development of the waste product recycling process, and can actively respond to the urgent needs of sustainable development and green development in China.

[0151] It should be noted that, according to the needs of implementation, each step / component described in the application can be split into more steps / components, or two or more steps / components or part of the operation of the steps / components can be combined into a new step / component, to realize the purpose of the application.

[0152] The size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiment of the application.

[0153] It should be understood that those skilled in the art can improve or change according to the above description, and all these improvements and changes should belong to the protection scope of the claims of the application.

Claims

1. A method for dual-axis hole compliant disassembly of industrial robots based on on-policy optimization under dynamic parameters, characterized in that, The method comprises the following steps: Three mechanism models of industrial robot double shaft hole disassembly based on mathematical description are constructed, wherein the first mechanism model describes the distance moved by a single shaft in a single hole during the whole process of two-point contact; the second mechanism model describes the distance moved by a double shaft in a double hole during the whole process of two-point contact on the outside; and the third mechanism model describes the distance moved by a double shaft in a double hole during the whole process of two-point contact on the inside; An industrial robot double shaft hole compliant disassembly strategy model based on a proximal policy optimization and an online training environment thereof are constructed; wherein the online training environment construction process comprises: constructing a double shaft and double hole three-dimensional model in a simulation environment, deploying an industrial robot model into the simulation environment, fixing a double shaft model on a force / torque sensor at the end of the industrial robot; controlling the movement of the target point in the process of disassembling the double shaft from the double hole by the industrial robot through specifying a target position and attitude, and obtaining dynamic parameters in the disassembly process, and calculating the values of the three mechanism models in the disassembly process according to the dynamic parameters; According to the dynamic data in the online training environment, the state, action, reward item and penalty item suitable for the industrial robot double shaft hole compliant disassembly problem are set, and the industrial robot double shaft hole compliant disassembly strategy model is trained, wherein one of the penalty items is calculated based on the values of the three mechanism models in the disassembly process; The compliant disassembly of the double shaft hole is completed according to the industrial robot double shaft hole compliant disassembly strategy model.

2. The method of claim 1, wherein the method is based on a near-end policy optimization of a dynamic parameter of a dual-axis hole compliant disassembly of an industrial robot, and wherein the method further comprises: The construction process of the three mechanism models comprises: describing the mathematical relationship between the structural elements according to the force balance relationship between the double shaft and the double hole in the process of disassembling the double shaft hole by the industrial robot under the quasi-static condition, and then constructing the three mechanism models according to the mathematical relationship.

3. The method of claim 1, wherein the method is based on a near-end policy optimization of a dynamic parameter of a dual-axis hole compliant disassembly of an industrial robot, and wherein the method further comprises: The construction process of the double shaft and double hole three-dimensional model comprises: creating a cube model and a cylinder model in the simulation environment, and constructing the double shaft and double hole three-dimensional model by combining the cylinder and the cube.

4. The method of claim 1, wherein, In the online training environment construction process, the inverse kinematics model of the UR5 robot is constructed, and the movement of the disassembly process is controlled by specifying the target position and attitude.

5. The method of claim 1, wherein, The penalty items in the online training process of the industrial robot double shaft hole compliant disassembly strategy model further include the penalty items of force and torque, the penalty items of the relative position amount between the double shaft hole in the x and y directions, and the penalty items of the relative deflection angle amount between the double shaft hole in the x, y and z directions.

6. The method of claim 1, wherein, In the training process, the industrial robot double shaft hole compliant disassembly strategy model continuously updates the parameters, and continues to train with the updated parameters.

7. The method of claim 1, wherein, In the training process, the effect of the current disassembly strategy model is evaluated by estimating the advantage function, and the disassembly strategy model is guided to optimize and improve, and in the optimization and improvement process, a clipping mechanism is introduced to limit the amplitude of strategy update, so as to ensure the smoothness in the strategy change process.

8. The method of claim 1-7, wherein, A virtual target point is generated and fixed on the double shaft center axis while the double shaft model is fixed on the force / torque sensor at the end of the industrial robot.

9. A dual-axis hole compliant disassembly system for industrial robots based on on-policy optimization of dynamic parameters, characterized in that, The method comprises the following steps: The mechanism model construction module is configured to construct three mechanism models of the industrial robot double-shaft hole disassembly based on mathematical description, wherein the first mechanism model describes the distance moved by a single-sided shaft in a single-sided hole during the whole process of always two-point contact; the second mechanism model describes the distance moved by a double-sided shaft in a double-sided hole during the whole process of always two-point contact on the outer side; and the third mechanism model describes the distance moved by a double-sided shaft in a double-sided hole during the whole process of always two-point contact on the inner side. The disassembly strategy model construction and training module is configured to construct an industrial robot double-shaft hole compliant disassembly strategy model based on proximal policy optimization and an online training environment thereof; wherein the online training environment construction process is as follows: constructing a three-dimensional model of the double shaft and the double hole in the simulation environment, deploying the industrial robot model into the simulation environment, fixing the double shaft model on the force / torque sensor at the end of the industrial robot; by specifying the target position and attitude, controlling the movement of the target point in the process of disassembling the double shaft from the double hole by the industrial robot, and obtaining the dynamic parameters in the disassembly process, calculating the values of the three mechanism models in the disassembly process according to the dynamic parameters; The compliant disassembly module is configured to complete the compliant disassembly of the double-shaft hole according to the industrial robot double-shaft hole compliant disassembly strategy model.

10. A computer storage medium, characterized in that The computer program stored therein can be executed by the processor, and the computer program executes the industrial robot double-shaft hole compliant disassembly method based on proximal policy optimization under dynamic parameters according to any one of claims 1-8.

Images