Collision detection method for cooperative robot

Abstract

The invention provides a collision detection method for a collaborative robot, and relates to the technical field of robot collision detection, and the method comprises the steps: determining the influence of a transmission mode and a transmission ratio between joints of the collaborative robot on joint friction through a test; the load, rigidity, precision, stability and speed requirements of the collaborative robot are comprehensively considered, the transmission mode and the transmission ratio between joints of the collaborative robot are adjusted, namely, the structural design for reducing joint friction of the collaborative robot is adopted, the collision torque during collision detection is effectively reduced, and the collision detection sensitivity is greatly improved; comprising the steps that a collaborative robot joint friction model based on a DLinear model is established, modeling is conducted through the DLinear model, that is, a friction modeling problem is classified into a time sequence prediction problem, the model structure is simple, and calculation is convenient; according to the method, force which is equal to calculated joint friction force in magnitude and opposite to the calculated joint friction force in direction is added into feed-forward torque, friction torque in the joint of the collaborative robot is effectively compensated, and the sensitivity of collision detection is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of robot collision detection, and in particular to a collision detection method for a collaborative robot. Background Art

[0002] Collaborative robots are robots that can work together with humans on the production line, giving full play to the efficiency of robots and the intelligence of humans. Such robots are not only cost-effective, but also safe and convenient, and can greatly promote the development of manufacturing companies.

[0003] Collaborative robots are widely used in industrial and service applications. When collaborative robots work with other equipment, they can easily cause damage to people, equipment, and the robot itself. Therefore, in order to avoid collisions with the outside, collaborative robots need to have collision detection capabilities.

[0004] In order to realize robot collision detection, patent CN111504328A discloses a robot path planning method. Collision detection based on path planning generally requires the use of visual sensors, laser radars, etc. to establish a spatial model of the real scene where the object is located, and perform collision detection on the object and the collision object in the spatial model. This method has very high real-time requirements for dynamic modeling of the real scene, and may fail when the sensor is blocked; patent CN107848122A discloses a joint torque sensor that can be used for robot collision detection. The force sensor equipped with the robot can quickly identify external forces and then stop movement when a collision occurs, thereby preventing damage caused by the collision. The detection using the force sensor has higher sensitivity and accuracy. However, the application of force sensors makes the robot structure design complex, and the control algorithm involved is also complex. On the other hand, the high cost of these sensors will greatly increase the overall cost of using collaborative robots; patent CN109940622A compares the error value between the output current of the mechanical arm joint motor and the estimated current output by the neural network with the collision detection threshold for collision judgment. The collision detection method based on the motor feedback current generally requires dynamic modeling. The accuracy of the dynamic model will directly affect the force control accuracy, thereby affecting the sensitivity of collision detection. Summary of the invention

[0005] The present invention discloses a collision detection method for a collaborative robot, which solves the problems that the existing collaborative robots generally need to use visual sensors, laser radars, etc. for collision detection based on path planning, have very high real-time requirements for dynamic modeling of real scenes, and may fail when the sensors are blocked; use of force sensors for collision detection makes the robot structure design complicated, the control algorithms involved are also complicated, and the overall cost of using the collaborative robot is greatly increased; and use of sensorless robot collision detection based on motor feedback current has low force control accuracy due to the existence of nonlinear friction in the joints, which seriously affects the collision detection sensitivity. A structure for reducing joint friction is designed, and the DLinear model is used for friction modeling of the collaborative robot. The friction force is further compensated by feedforward torque, thereby effectively improving the collision detection sensitivity of the collaborative robot.

[0006] To achieve the above object, the technical solution of the present invention is specifically implemented as follows:

[0007] The present invention discloses a collision detection method for a collaborative robot, comprising the following steps:

[0008] S1: Under sensorless conditions, the effects of the transmission mode and transmission ratio between the joints of the collaborative robot on joint friction are determined through experiments;

[0009] S2: Comprehensively consider the load, stiffness, accuracy, stability and speed requirements of the collaborative robot, and adjust the transmission mode and transmission ratio between the joints of the collaborative robot;

[0010] S3: Establish a collaborative robot joint friction model based on the DLinear model;

[0011] S4: Calculate the joint friction of the collaborative robot;

[0012] S5: adding a force equal in magnitude and opposite in direction to the joint friction force calculated in step S4 to the feedforward torque;

[0013] S6: Calculate the error between the feedforward torque and the actual torque. When the error between the feedforward torque and the actual torque exceeds a predetermined torque threshold, a collision is detected.

[0014] Furthermore, the specific implementation of step S3 includes:

[0015] S31: Establish the dynamic model of collaborative robots;

[0016]

[0017] Where q∈R n is the position variable of the robot in the joint space; n represents the nth joint of the collaborative robot; is the joint speed of the collaborative robot; is the joint acceleration of the collaborative robot; D(q) is the inertia matrix of the collaborative robot; represents the Coriolis force and centrifugal force of the collaborative robot, G(q) is the gravity of the collaborative robot, represents the friction force of the collaborative robot, and τ represents the driving torque of each joint of the collaborative robot;

[0018] S32: Modeling the nonlinear friction of collaborative robot joints using the DLinear model;

[0019] For a time series containing C variables, given the historical review window data X;

[0020]

[0021] Where L is the length of historical data; is the variable value of the i-th variable at the t-th time step;

[0022] The time series prediction task is to predict the value Y of the next T time steps;

[0023]

[0024] in, represents the predicted value of the i-th variable at the t-th time step; R T×C Where R represents a real number set, T and C represent the dimension of the real number set R;

[0025] S33: DLinear model decomposes historical time series data into trend data series and residual data series;

[0026] X t ∈R L×C (4)

[0027] X s =XX t |X s ∈R L×C (5)

[0028] Among them, X t is the trend data series; X s is the remaining data sequence; R L×C Where R represents a real number set, L and C represent the dimension of the real number set R;

[0029] S34: Apply linear layer network W to the decomposed trend data sequence and the remaining data sequence respectively t ∈R T×L and W s ∈R T×L ,Right now

[0030] H t =W t X t ∈R T×C (6)

[0031] H s =W s X s ∈R T×C (7)

[0032] Among them, W t represents a linear layer network; W s represents a linear layer network; R T×L Where R represents the real number set, T and L represent the dimension of the real number set R; H t Represents X t Through W t The result obtained after s Represents X s Through W s The result obtained later;

[0033] S35: Output the sum of the output results of the two linear layers Y∈R T×C ;

[0034] Y=H t +H s (8)

[0035] S36: The 8-dimensional time series data of timestamp, reference position, actual position, position error, reference speed, actual speed, theoretical torque and actual torque of each joint of the collaborative robot are input into the DLinear model, and the actual torque is used as the target sequence for prediction.

[0036] Furthermore, the specific implementation of step S4 is as follows:

[0037] The joint friction force of the collaborative robot is the difference between the actual torque and the theoretical torque.

[0038] Furthermore, the specific implementation of step S6 is as follows:

[0039] The predetermined torque threshold adopts a dynamic threshold, setting a larger torque threshold when the collaborative robot is accelerating and decelerating; setting a larger torque threshold when the collaborative robot joint speed is higher; and setting a lower torque threshold when the collaborative robot joint speed is lower.

[0040] Beneficial technical effects:

[0041] 1. The present invention discloses a collision detection method for a collaborative robot, which includes, in the absence of a sensor, determining the influence of the transmission mode and transmission ratio between the joints of the collaborative robot on the joint friction through experiments; and then comprehensively considering the load, stiffness, accuracy, stability and speed requirements of the collaborative robot, adjusting the transmission mode and transmission ratio between the joints of the collaborative robot, that is, finding a structural design scheme to reduce the joint friction of the collaborative robot, effectively reducing the collision torque during collision detection, and thus greatly improving the collision detection sensitivity;

[0042] 2. A collaborative robot collision detection method disclosed in the present invention includes establishing a collaborative robot joint friction model based on a DLinear model. The DLinear model is used to classify the friction modeling problem as a time series prediction problem. The model structure is simple and the calculation is convenient.

[0043] 3. The present invention discloses a collision detection method for a collaborative robot, which includes adding a force equal to the calculated joint friction force and opposite in direction to the feedforward torque, thereby effectively compensating for the friction torque in the joints of the collaborative robot and greatly improving the sensitivity of collision detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solution of the present invention, the drawings required for describing the embodiments are briefly introduced below.

[0045] Figure 1 A flowchart of specific steps of a collision detection method for a collaborative robot disclosed in the present invention;

[0046] Figure 2 This is a schematic diagram of the overall structure of a 4-joint collaborative robot;

[0047] Figure 3 This is a schematic diagram of the cross-sectional structure of a 4-joint collaborative robot;

[0048] Figure 4 for Figure 3 A magnified view of the structure of part B;

[0049] Figure 5 It is a schematic diagram of the relationship between the friction force and the transmission ratio of the harmonic reducer;

[0050] Figure 6 is the ball screw friction force of joint 1;

[0051] Figure 7 This is a structural diagram of joint 3 when all joints of a 4-joint collaborative robot are driven by synchronous wheels and synchronous belts;

[0052] Figure 8 The friction force variation curve of the synchronous belt under different transmission ratios;

[0053] Fig. 9 is the friction force of the synchronous belt drive of joint 1;

[0054] Fig.10 It is a schematic diagram of the structure of DLinear;

[0055] Fig.11 This is the principle block diagram of friction compensation;

[0056] Fig.12a The fitting diagram of the torque and actual torque of joint 1 after compensation by DLinear's joint friction model, ARIMA model, LSTM model and Informer model respectively;

[0057] Figure 12b The fitting diagram of the torque and actual torque of joint 2 after compensation by DLinear's joint friction model, ARIMA model, LSTM model and Informer model respectively;

[0058] Fig.12c The fitting diagram of the torque and actual torque of joint 3 after compensation by DLinear's joint friction model, ARIMA model, LSTM model and Informer model respectively;

[0059] Fig.12d The fitting diagram of the torque and actual torque of joint 4 after compensation by DLinear's joint friction model, ARIMA model, LSTM model and Informer model respectively;

[0060] Fig.13 A flowchart of the steps for collision detection;

[0061] Fig.14a Comparison curve between the feedforward torque and actual torque of joint 1 when the 4-joint collaborative robot adopts scheme 1 harmonic reducer + ball screw transmission;

[0062] Fig.14b When the 4-joint collaborative robot adopts scheme 1 harmonic reducer + ball screw transmission, the comparison curve of the collision torque of joint 1 and the predetermined torque threshold;

[0063] Fig.14c Comparison curve between the feedforward torque and actual torque of joint 2 when the 4-joint collaborative robot adopts scheme 1 harmonic reducer + ball screw transmission;

[0064] Fig.14d When the 4-joint collaborative robot adopts scheme 1 harmonic reducer + ball screw transmission, the comparison curve of the collision torque of joint 2 and the predetermined torque threshold;

[0065] Fig.14eComparison curve between the feedforward torque and actual torque of joint 3 when the 4-joint collaborative robot adopts scheme 1 harmonic reducer + ball screw transmission;

[0066] Fig.14f When the 4-joint collaborative robot adopts scheme 1 harmonic reducer + ball screw transmission, the comparison curve of the collision torque of joint 3 and the predetermined torque threshold;

[0067] Figure 14g Comparison curve between the feedforward torque and actual torque of joint 4 when the 4-joint collaborative robot adopts scheme 1 harmonic reducer + ball screw transmission;

[0068] Figure 14h When the 4-joint collaborative robot adopts scheme 1 harmonic reducer + ball screw transmission, the comparison curve of the collision torque of joint 4 and the predetermined torque threshold;

[0069] Fig.15a This is the comparison curve between the feedforward torque and actual torque of joint 1 when the 4-joint collaborative robot adopts the synchronous belt drive of scheme 2;

[0070] Fig.15b The comparison curve between the collision torque of joint 1 and the predetermined torque threshold when the 4-joint collaborative robot adopts the synchronous belt drive of scheme 2;

[0071] Fig.15c This is the comparison curve between the feedforward torque and actual torque of joint 2 when the 4-joint collaborative robot adopts the synchronous belt drive of scheme 2;

[0072] Fig.15d The comparison curve between the collision torque of joint 2 and the predetermined torque threshold when the 4-joint collaborative robot adopts the synchronous belt drive of scheme 2;

[0073] Fig.15e This is the comparison curve between the feedforward torque and actual torque of joint 3 when the 4-joint collaborative robot adopts the synchronous belt drive of scheme 2;

[0074] Fig.15f The comparison curve between the collision torque of joint 3 and the predetermined torque threshold when the 4-joint collaborative robot adopts the synchronous belt drive of scheme 2;

[0075] Figure 15g This is the comparison curve between the feedforward torque and actual torque of joint 4 when the 4-joint collaborative robot adopts the synchronous belt drive of scheme 2;

[0076] Figure 15h The comparison curve of the collision torque of joint 4 and the predetermined torque threshold when the 4-joint collaborative robot adopts the synchronous belt drive of scheme 2.

[0077] Among them, 1-joint 1; 2-joint 2; 3-joint 3; 4-joint 4, 5-first screw fixing seat, 6-guide rail, 7-ball screw, 8-second screw fixing seat, 9-motor, 10-encoder, 11-motor end cover, 12-bearing, 13-harmonic reducer, 14-synchronous wheel, 15-synchronous belt, 16-connecting rod. DETAILED DESCRIPTION

[0078] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0079] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0080] Unless otherwise specifically stated, the relative arrangement of the parts and steps described in these embodiments, numerical expressions and numerical values ​​do not limit the scope of the present invention. At the same time, it should be understood that, for ease of description, the sizes of the various parts shown in the accompanying drawings are not drawn according to the actual proportional relationship. The technology, method and equipment known to ordinary technicians in the relevant field may not be discussed in detail, but in appropriate cases, the technology, method and equipment should be regarded as a part of the authorization specification. In all examples shown and discussed here, any specific value should be interpreted as being merely exemplary, rather than as a limitation. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once a certain item is defined in an accompanying drawing, it does not need to be further discussed in subsequent drawings.

[0081] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0082] In order to solve various problems existing in the existing robot collision detection method, the present invention discloses a collision detection method for a collaborative robot, see Figure 1 The collaborative robot-human collision detection method disclosed in the present invention specifically comprises the following steps:

[0083] It should be noted in advance that this embodiment uses a 4-joint collaborative robot as the analysis object to describe this method in detail. Figure 2-Figure 4 The joint 1 of the collaborative robot serves as a mobile joint, that is, the Z-direction travel of the collaborative robot can be customized arbitrarily.

[0084] S1: Under sensorless conditions, the effects of the transmission mode and transmission ratio between the joints of the collaborative robot on joint friction are determined through experiments;

[0085] Specifically, the mechanical structural factors that affect joint friction mainly include transmission mode, transmission ratio, structure selection and related process parameters. The present invention studies the friction characteristics of three different transmission modes of the collaborative robot, namely, harmonic reducer, ball screw and synchronous belt, and different transmission ratios.

[0086] When joints 2 to 4 use harmonic reducers for transmission and deceleration, Figure 3 Taking the structure shown in the figure as an example, the friction of joints 2 to 4 is mainly composed of the composite friction of the harmonic reducer and the friction of the motor bearing. Joint 1 is driven by a non-self-locking ball screw, such as Figure 4 As shown, the friction force of joint 1 mainly includes the composite friction of the ball screw, the friction of the linear guide, the friction of the screw support bearing, the friction of the motor bearing, etc.

[0087] Compare the effects of different transmission ratios of harmonic reducers on joint friction: Taking joint 3 as an example, reducers with transmission ratios of 30, 50, 80, 100 and 120 are used to ensure that the maximum speed of the joint is consistent under different transmission ratios. When performing uniform motion of a single joint, the friction torque and motor speed at the motor end are collected and converted to the joint end respectively, such as Figure 5 As shown, it can be seen that the friction force at the joint end increases with the increase of the transmission ratio, which is approximately in direct proportion. This is due to the torque amplification effect of the harmonic reducer. The friction force at the motor end is also amplified accordingly. It can be said that the size of the transmission ratio directly determines the size of the joint friction. In addition, the Stribeck effect of the harmonic reducer is obvious at low speed.

[0088] For joint 1, ball screw transmission and deceleration are adopted. The lead of ball screw is 0.02m, and the equivalent transmission ratio is 2π / 0.02=314. When joint 1 moves at a constant speed, subtract the gravity term to obtain the friction torque and motor speed at the motor end, and convert them to the joint end, such as Figure 6 As shown in the figure, due to the large transmission ratio of the ball screw and the large number of friction components in the ball screw transmission, the friction composition is relatively complex, which ultimately leads to large friction at the joint end. The Stribeck effect of the ball screw is very significant, which indicates that the static friction of the ball screw is large, which will bring challenges to the robot's collision detection.

[0089] When joints 1 to 4 are driven and decelerated by synchronous wheels and synchronous belts, Figure 7 In the joint 3 structure shown, the friction of each joint mainly includes the friction between the synchronous wheel and the synchronous belt, the friction of the synchronous wheel support bearing, the friction of the drive motor bearing, etc.

[0090] Limited by the size of the connecting rod and the synchronous wheel, the transmission ratio of the synchronous belt and synchronous wheel is relatively small compared with the harmonic reducer drive. Taking joint 3 as an example, the first-stage transmission ratio is 3.75 and the second-stage transmission ratio is 9.765. The friction torque and motor speed at the motor end are collected and converted to the joint end, such as Figure 8 As shown in the figure, the friction force of synchronous belt drive will also increase with the torque amplification effect of the transmission ratio. In addition, due to the increase in friction components of multi-stage synchronous belts, the friction force increases more obviously. For joint 1 synchronous belt drive, the equivalent transmission ratio is 196.35. The friction torque and motor speed at the motor end are collected and converted to the joint end, as shown in the figure. Fig. 9 shown.

[0091] By comparing the harmonic reducer, ball screw and synchronous belt drive, the following conclusions are drawn: the choice of transmission mode and the size of the transmission ratio directly determine the type of joint friction and the order of magnitude of the friction, thereby determining the force control performance of the robot, and ultimately affecting the sensitivity of collision detection. Therefore, from the perspective of improving the force control performance of the robot, a transmission mode with a small transmission ratio should be selected as much as possible to reduce the parts involved in friction, and on this basis, the friction force can be accurately identified and modeled.

[0092] S2: Comprehensively consider the load, stiffness, accuracy, stability and speed requirements of the collaborative robot, and adjust the transmission mode and transmission ratio between the joints of the collaborative robot;

[0093] Specifically, in the absence of sensors, taking into account the requirements of load, stiffness, accuracy, stability, speed, etc., the present invention proposes a structural design scheme to reduce joint friction. In this embodiment, two transmission schemes of the 4-joint collaborative robot are provided to meet different application requirements, as shown in Table 1;

[0094] Table 1 Two transmission schemes of 4-joint collaborative robots

[0095]

[0096] S3: Establish a collaborative robot joint friction model based on the DLinear model;

[0097] Specifically, because the neural network model does not require prior knowledge, but before establishing the friction model based on the neural network, a dynamic model is established first, which can provide a reference for the availability of data and improve the generalization ability of the neural network model. Therefore, the Newton-Euler recursion method is used to carry out dynamic modeling of the collaborative robot.

[0098] S31: The robot dynamics model in the joint space is described as follows, with the friction force explicitly represented:

[0099] Establish the dynamic model of collaborative robots;

[0100]

[0101] Where q∈R n is the position variable of the robot in the joint space. In this embodiment, the independent variable q of joint 1 is the connecting rod offset d, and the independent variables q of joints 2 to 4 are the joint angles θ; n represents the nth joint of the collaborative robot; is the joint speed of the collaborative robot; is the joint acceleration of the collaborative robot; D(q) is the inertia matrix of the collaborative robot; represents the Coriolis force and centrifugal force of the collaborative robot, G(q) is the gravity of the collaborative robot, represents the friction force of the collaborative robot, and τ represents the driving torque of each joint of the collaborative robot;

[0102] In the dynamic model (1), the driving torque τ of each joint motor is motor The motor feedback current I and the motor torque constant K t Multiply them together to get;

[0103] τ motor =IK t

[0104] The driving torque τ of each joint end is determined by the torque τ of the motor end. motor Multiply by the gear ratio Gear, that is:

[0105] τ=Gear τ motor

[0106] The speed R of each joint motor divided by the transmission ratio Gear is the speed of the joint end.

[0107]

[0108] The first three terms in formula (1) The sum of G(q) is the theoretical torque of the robot, which can be obtained through theoretical calculation. The dynamic parameters required to calculate the theoretical torque can be given by the 3D model or obtained through parameter identification.

[0109] S32: Modeling the nonlinear friction of collaborative robot joints using the DLinear model;

[0110] In the process of friction modeling, the robot data used, such as joint position, joint speed, joint torque and current, are arranged in chronological order. Therefore, the present invention classifies the friction modeling problem as a time series forecasting (TSF) problem.

[0111] Specifically, for a time series containing C variables, given the historical review window data X;

[0112]

[0113] Where L is the length of historical data; is the variable value of the i-th variable at the t-th time step;

[0114] The time series prediction task is to predict the value Y of the next T time steps;

[0115]

[0116] in, represents the predicted value of the i-th variable at the t-th time step; R T×C Where R represents a real number set, T and C represent the dimension of the real number set R;

[0117] S33: DLinear model decomposes historical time series data into trend data series and residual data series;

[0118] X t ∈R L×C (4)

[0119] X s =XX t |X s ∈R L×C (5)

[0120] Among them, X t is the trend data series; X s is the remaining data sequence; R L×C Where R represents a real number set, L and C represent the dimension of the real number set R;

[0121] S34: Apply linear layer network W to the decomposed trend data sequence and the remaining data sequence respectively t ∈R T×L and W s ∈R T×L , see Fig.10 ,Right now

[0122] H t =W t X t ∈R T×C (6)

[0123] H s =W s X s ∈R T×C (7)

[0124] Among them, W t represents a linear layer network; W s represents a linear layer network; R T×L Where R represents the real number set, T and L represent the dimension of the real number set R; H t Represents X t Through W t The result obtained after s Represents X s Through W s The result obtained later;

[0125] S35: Output the sum of the output results of the two linear layers Y∈R T×C ;

[0126] Y=H t +H s (8)

[0127] S36: The 8-dimensional time series data of timestamp, reference position, actual position, position error, reference speed, actual speed, theoretical torque and actual torque of each joint of the collaborative robot are input into the DLinear model, and the actual torque is used as the target sequence for prediction.

[0128] S4: Calculate the joint friction of the collaborative robot;

[0129] Specifically, joint friction force is the difference between actual torque and theoretical torque.

[0130] S5: adding a force equal in magnitude and opposite in direction to the joint friction force calculated in step S4 to the feedforward torque;

[0131] Specifically, see Fig.11 , Taking the actual running trajectory of a 4-joint collaborative robot as an example, the beneficial effect of friction compensation is evaluated. Due to the large variation of the motor feedback current, the actual torque is filtered. The filtered torque data can prevent the friction model from overfitting.

[0132] In order to verify the accuracy of the joint friction model based on DLinear, the model is compared with the ARIMA model, LSTM model, and Informer model. The torque after compensation of each model is different from the actual torque. Figures 12a-12d As shown in the figure, it can be seen that the torque after compensation based on the DLinear friction model is closest to the actual torque after filtering, which plays an important role in improving the sensitivity of collision detection.

[0133] S6: Calculate the error between the feedforward torque and the actual torque. When the error between the feedforward torque and the actual torque exceeds a predetermined torque threshold, a collision is detected.

[0134] Under normal operating conditions, the controller calculates the feedforward torque τ feedforward and the actual torque τ actual The error Δτ between them is very small;

[0135] Δτ=τ actual -τ feedforwar

[0136] When the robot is hit by an external force, the joint current changes suddenly and sharply, and the error between the feedforward torque and the actual torque increases. When the torque threshold τ is exceeded, threshold When Δτ≥τ threshold , that is, a collision is detected, see Fig.13 At this time, the motors of the collaborative robot joints should be powered off in time, and corresponding safety measures should be taken to reduce the harm caused by the robot to personnel, surrounding equipment and other objects.

[0137] It should be noted that lowering the threshold is conducive to collision detection, thereby reducing the collision torque, but lowering the threshold may also cause misjudgment, so the feedforward torque after friction compensation should be as close to the actual torque as possible. This embodiment adopts a dynamic threshold to improve the sensitivity of collision detection and reduce misjudgment according to the operating conditions of the collaborative robot. The specific implementation is as follows: when the torque error is large during acceleration and deceleration, a larger torque threshold is set; when the joint speed is large, a larger torque threshold is also used, and when the joint speed is low, a lower torque threshold is implemented.

[0138] 14a-15h respectively show the collision torque of each joint when using scheme 1 harmonic reducer + ball screw transmission and scheme 2 synchronous belt transmission. The maximum value of Δτ is the joint collision torque.

[0139] The actual collision force was measured using a push-pull dynamometer. Table 2 shows the actual collision force of the 4-joint collaborative robot under different transmission modes and different speeds.

[0140] Table 2 Collision force of different transmission modes

[0141]

[0142] As can be seen from Table 2, a smaller transmission ratio is used in the robot joints, which can reduce joint friction and ultimately reduce collision force. The collision force at each joint of the robot using the synchronous belt solution is significantly smaller than that of the robot using the harmonic reducer + ball screw solution.

[0143] Table 3 shows the actual collision force of the 4-joint collaborative robot using different friction models.

[0144] Table 3 Collision force of different friction models

[0145]

[0146] From Table 3, it can be seen that using a more accurate friction model can improve the sensitivity of collision detection and reduce the collision force.

[0147] The present invention studies the influence of friction elements and mechanical structure design on the friction of collaborative robots. Specifically, the friction characteristics of joints under different transmission modes and transmission ratios are studied. In order to improve the collision detection performance of the robot, the present invention proposes a structural design scheme to reduce joint friction. The present invention classifies the friction modeling problem of the robot into a time series data regression prediction problem, adopts the DLinear model for friction modeling, and further compensates the friction force by feedforward torque. The collision detection method disclosed in the present invention can effectively detect abnormal collisions with high sensitivity.

[0148] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0149] The above embodiments are only descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary engineering and technical personnel in the field should fall within the protection scope determined by the claims of the present invention.

Claims

1. A collision detection method for a collaborative robot, characterized in that: The following steps are involved: S1: Under sensorless conditions, the effects of the transmission mode and transmission ratio between the joints of the collaborative robot on joint friction are determined through experiments; S2: Comprehensively consider the load, stiffness, accuracy, stability and speed requirements of the collaborative robot, and adjust the transmission mode and transmission ratio between the joints of the collaborative robot; S3: Establish a collaborative robot joint friction model based on the DLinear model; S4: Calculate the joint friction of the collaborative robot; S5: adding a force equal in magnitude and opposite in direction to the joint friction force calculated in step S4 to the feedforward torque; S6: Calculate the error between the feedforward torque and the actual torque. When the error between the feedforward torque and the actual torque exceeds a predetermined torque threshold, a collision is detected.

2. A collaborative robot collision detection method according to claim 1, characterized in that: The specific implementation of step S3 includes: S31: Establish the dynamic model of collaborative robots; Where q∈R n is the position variable of the robot in the joint space; n represents the nth joint of the collaborative robot; is the joint speed of the collaborative robot; is the joint acceleration of the collaborative robot; D(q) is the inertia matrix of the collaborative robot; represents the Coriolis force and centrifugal force of the collaborative robot, G(q) is the gravity of the collaborative robot, represents the friction force of the collaborative robot, and τ represents the driving torque of each joint of the collaborative robot; S32: Modeling the nonlinear friction of collaborative robot joints using the DLinear model; For a time series containing C variables, given the historical review window data X; Where L is the length of historical data; x i t is the variable value of the i-th variable at the t-th time step; The time series prediction task is to predict the value Y of the next T time steps; in, represents the predicted value of the i-th variable at the t-th time step; R T×C Where R represents a real number set, T and C represent the dimension of the real number set R; S33: DLinear model decomposes historical time series data into trend data series and residual data series; X t ∈R L×C (4) X s =X-X t |X s ∈R L×C (5) Among them, X t is the trend data series; X s is the remaining data sequence; R L×C Where R represents a real number set, L and C represent the dimension of the real number set R; S34: Apply linear layer network W to the decomposed trend data sequence and the remaining data sequence respectively t ∈R T×L and W s ∈R T×L ,Right now H t =W t X t ∈R T×C (6)H s =W s X s ∈R T×C (7) Among them, W t represents a linear layer network; W s represents a linear layer network; R T×L Where R represents the real number set, T and L represent the dimension of the real number set R; H t Represents X t Through W t The result obtained after s Represents X s Through W s The result obtained later; S35: Output the sum of the output results of the two linear layers Y∈R T×C ; Y=H t +H s (8) S36: The 8-dimensional time series data of timestamp, reference position, actual position, position error, reference speed, actual speed, theoretical torque and actual torque of each joint of the collaborative robot are input into the DLinear model, and the actual torque is used as the target sequence for prediction.

3. The method for detecting a collision of a collaborative robot according to claim 1, characterized in that: The specific implementation of step S4 is: The joint friction force of the collaborative robot is the difference between the actual torque and the theoretical torque.

4. The method for detecting collision of a collaborative robot according to claim 1, characterized in that: The specific implementation of step S6 is: The predetermined torque threshold adopts a dynamic threshold, setting a larger torque threshold when the collaborative robot is accelerating and decelerating; setting a larger torque threshold when the collaborative robot joint speed is high; and setting a lower torque threshold when the collaborative robot joint speed is low.

Images