Self-adaptive variable scale mapping teleoperation control method

Abstract

The invention discloses a self-adaptive variable scale mapping teleoperation control method which comprises the following steps: acquiring Cartesian space position information of the tail end of a main end mechanical arm to obtain a main end rate factor Dm; target position coordinates, coordinates of a point, closest to the tail end of the slave end mechanical arm, of the surrounding environment and actual pose information of the tail end of the slave end mechanical arm are obtained, and a target distance factor Dg and an environment safety factor De are obtained; inputting Dm, De and Dg into a mapping coefficient network, and recording an output value of the mapping coefficient network, namely a mapping coefficient k; according to the mapping coefficient k and the Cartesian space pose information of the tail end of the master-end mechanical arm, expected pose information of the tail end of the slave-end mechanical arm is obtained; and joint control is conducted according to the expected pose information, and the actual pose information of the tail end of the slave end mechanical arm is obtained. Smooth variable-scale mapping operation control of the master-slave teleoperation robot in the heterogeneous space can be achieved, operation safety is improved, and the method has wide application prospects in the fields of medical operations, space deep-sea exploration, disaster rescue and the like.

Description

Technical Field

[0001] The invention relates to an adaptive variable scale mapping teleoperation control method, belonging to the technical field of teleoperation robot master-slave control. Background Art

[0002] Teleoperated robots combine human intelligence with robotics. At a time when current artificial intelligence (AI) technologies haven't yet enabled robots to achieve intelligent decision-making capabilities comparable to humans, they can replace humans in complex, unstructured environments and meet flexible mission requirements, completing tedious or dangerous tasks such as medical surgery, deep-sea exploration, and disaster relief. Currently, a variety of master-slave teleoperation architectures exist, with Cartesian space mapping in heterogeneous spaces being the most versatile. However, due to the conflict between rapid movement and precise operation associated with single position mapping, variable scale mapping and rate mapping are currently the primary implementation methods.

[0003] Existing methods divide the slave robot's workspace into an area far from the target and an area close to the target. When moving in the area far from the target, a constant-scale absolute workspace mapping method is used; when moving in the area close to the target, a variable-scale incremental workspace mapping method based on pose separation is used. However, this method has the following main problems:

[0004] (1) The scale coefficients of the constant-scale absolute and variable-scale incremental workspace mappings based on posture separation need to be determined according to the operator's needs and input into the central control computer through the keyboard. They cannot be adaptively adjusted according to the operating range and operating tasks, which increases the operator's operation complexity.

[0005] Existing methods also include adaptively setting the gain coefficient K based on the size of the master hand's displacement increment to determine the spatial mapping relationship between the master hand and the slave hand, facilitating fast and accurate control of the slave hand. However, this method has the following main problems:

[0006] (1) This method adaptively sets the gain coefficient K according to the size of the master hand displacement increment. It can adaptively adjust the coefficient, but it is adjusted according to the master hand displacement increment and divided into precise mode, normal mode, and fast mode, representing the speed attribute, without considering the position restriction of the operating environment. There may be a situation where the proportional coefficient is high even if the object is close to the operation target, which increases the hidden danger of operation safety.

[0007] (2) Posture control is not absolute and is a non-intuitive operation, which increases the control difficulty for the operator. In addition, the feedback force does not use the data of the force sensor, but is calculated through the master-slave position deviation, which has a certain hysteresis, increases the probability of collision, and affects the safety of the master-end operation. Summary of the Invention

[0008] The present invention aims to overcome these shortcomings by providing an adaptive variable-scaling mapping teleoperation control method, resolving the technical issues of poor safety and operational difficulties associated with existing mapping teleoperation. This method enables smooth variable-scaling mapping control of master-slave telerobots in heterogeneous spaces, improving operational safety. The method has broad application prospects in fields such as medical surgery, space and deep-sea exploration, and disaster relief.

[0009] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0010] An adaptive variable-scale mapping teleoperation control method, comprising:

[0011] The S1 master robot obtains the Cartesian spatial position information of the end of the master manipulator and obtains the master rate factor D based on the Cartesian spatial position information of the end of the master manipulator. m ;

[0012] The slave robot obtains the target position coordinates, the coordinates of the nearest point between the surrounding environment and the end of the slave manipulator, and the actual posture information of the end of the slave manipulator. The target distance factor D is obtained based on the target position coordinates, the coordinates of the nearest point between the surrounding environment and the end of the slave manipulator, and the actual posture information of the end of the slave manipulator. g and environmental safety factor D e ;

[0013] S2 will factor the master rate D m , target distance factor D e , Environmental safety factor D g Input the mapping coefficient network and record the output value of the mapping coefficient network, which is the mapping coefficient k at this moment;

[0014] According to the mapping coefficient k and the Cartesian space pose information of the end of the master end manipulator, the desired pose information of the end of the slave end manipulator is obtained;

[0015] The slave robot is controlled by joints according to the desired pose information of the end of the slave manipulator arm to obtain the actual pose information of the end of the slave manipulator arm;

[0016] When the S3 operation task is completed, the control process ends; otherwise, steps S1 and S2 are repeated.

[0017] Further,

[0018] Where, X ′ m Y is the difference between the x-axis position of the end of the main end robot arm at the current moment and the x-axis position of the end of the main end robot arm at the previous moment. ′ m is the position difference of the end of the master end robot in the y-axis direction at the current moment, Z′ m It is the position difference of the end of the master robot arm in the z-axis direction at the current moment.

[0019] Further,

[0020] Where, X s 、Y s , Z s They are the position coordinates of the slave end robot arm in the x-axis, y-axis, and z-axis directions at the current moment, respectively. g 、Y g , Z g are the coordinates of the target in the x-axis, y-axis, and z-axis directions.

[0021] Further,

[0022] Where, X s 、Y s , Z s They are the position coordinates of the slave end robot arm in the x-axis, y-axis, and z-axis directions at the current moment, respectively. e 、X e 、X e The x-axis, y-axis, and z-axis coordinates of the point closest to the end of the slave robot arm.

[0023] Furthermore, it also includes:

[0024] In step S1, the master robot executes force feedback control instructions based on the force compensation algorithm;

[0025] In step S2, the actual load information of the end of the slave manipulator is recorded, and a force feedback control instruction based on a force compensation algorithm is generated according to the actual load information of the end of the slave manipulator.

[0026] Furthermore, the force feedback control instructions based on the force compensation algorithm include:

[0027] The control loop of the master-end robotic arm adopts a feedforward-feedback hybrid model. The feedforward control establishes the interference force model of the master-end robotic arm in advance. The compensation torque that each joint should output is calculated in real time through the joint status of the master-end robotic arm during operation and the data of the six-dimensional force sensor at the end of the master-end robotic arm, eliminating the interference force of each connecting rod and joint of the master-end robotic arm, so that the feedback force of the end of the master-end robotic arm on the human hand is the slave-end load force that eliminates the influence of its own gravity and friction. The slave-end load force is the actual load information of the end of the slave-end robotic arm.

[0028] Furthermore, the mapping coefficient network is based on the target distance factor D e , Environmental safety factor D g , master-side rate factor D mis the input, the mapping coefficient k is the output, and the network model learning and training is performed in advance;

[0029] The training sample rules for network model learning and training are as follows:

[0030]

[0031] Among them, D safe is the safety margin factor, D e med is the target distance median factor, D g med is the median factor of environmental safety, D m med The median rate factor of the master end, k max is the maximum allowable mapping ratio coefficient related to the entire task. The above five parameters are pre-set according to the task attributes.

[0032] Furthermore, the sample set input array {D e , D g , D m} and the output array {k} are generated segment by segment according to the training sample rules:

[0033] k=0 segment:

[0034] Random array generation:

[0035] D e =rand(1:100)D e max

[0036] D g =rand(1:100)D g max

[0037] D m =rand(1:100)D m max

[0038] Among them, D e max is the maximum value of the target distance factor, D g max is the maximum value of the environmental safety factor, D m max The maximum value of the master-side rate factor;

[0039] D e ≤D safe or D g ≤D safe With D mArrange and combine to randomly select 100 groups of data as the input array, and the output array is always 0, and the array selection is 100;

[0040] 0 < k < 1 segment:

[0041] The input array {D e , D g , D m} is generated respectively according to the following formula:

[0042] D e (n) = n * (D e med e - D safe ) / 100 + D safe

[0043] D g (n) = n * (D g med - D safe ) / 100 + D safe

[0044] D m (n) = n * (D m med ) / 100

[0045] Among them, n is an array starting from 1 and incrementing by 1 step by step to 100;

[0046] The output array {k} is generated according to the formula k(n) = n / 100;

[0047] k = 1 segment:

[0048] The input array {D e , D g , D m} is {D e s med , D g med , D m med};

[0049] [[ID=#72]]The output array {k} is {1};

[0050] 1 < k < k max segment:

[0051] The input array {D e , D g , D m} is generated respectively according to the following formula:

[0052] D e (n) = n * (D e max - De med ) / 100+D e med

[0053] D g (n)=n*(D g max -D g med ) / 100+D g med

[0054] D m (n)=n*(D m max -D m med ) / 100+D m med

[0055] The output array {k} is based on the formula k(n)=n*(k max -1) / 100+1 generated.

[0056] Furthermore, the network model adopts the RBF neural network model, which includes an input layer, a hidden layer, and an output layer;

[0057] The output of the input layer is the input of the hidden layer; the output of the hidden layer is the input of the output layer;

[0058] The number of hidden layer nodes is 3-6.

[0059] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0060] (1) The variable-scale mapping coefficient of the present invention fully considers factors such as environmental boundary restrictions, distance from the target location, and master-end operation speed, and comprehensively calculates the size of the mapping coefficient, thereby improving work efficiency and operation accuracy while greatly reducing the probability of collision and improving the safety of master-end operation;

[0061] (2) The present invention calculates the variable-scale mapping coefficients by network training, has generalization capability, and the coefficient transformation is transitional, which reduces the inadaptability of the master-end operator;

[0062] (3) The present invention designs different control strategies for position and posture respectively, which improves the intuitiveness of posture mapping and reduces the difficulty of operation for the operator. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 This is a flow chart of the adaptive variable scale mapping remote control method of the present invention;

[0064] Figure 2This is the block diagram of the adaptive variable scale mapping remote control of the present invention;

[0065] Figure 3 Schematic diagram of the mapping coefficient network model of the present invention. DETAILED DESCRIPTION

[0066] The following detailed description of the present invention will make the features and advantages of the present invention more clear and explicit.

[0067] The word "exemplary" is used exclusively herein to mean "serving as an example, example, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.

[0068] This paper proposes an adaptive variable-scale mapping teleoperation control method. This method, a teleoperation mapping method for master-slave heterogeneous spaces, adaptively adjusts mapping scale parameters based on environmental safety boundaries, distance to the target, and operating speed. This method enables both rapid long-distance access and precise close-range operations. This method is suitable for teleoperated intelligent robots in medical surgery, deep-sea exploration, disaster relief, and other robotic systems requiring remote operator control of high-precision operations. It can be used to remotely control robots to perform precise and dexterous operations in complex environments.

[0069] like Figure 1 and Figure 2 As shown in FIG, an adaptive variable scale mapping teleoperation control method includes the following four steps:

[0070] Step 1: Start the master-slave operation task. Power on the master robot system and perform the calibration procedure to complete the calibration of each sensor. Power on the slave robot system and prepare for the task execution.

[0071] Step 2: Data acquisition and processing stage: The master robot system executes the force feedback control instructions based on the force compensation algorithm, collects the Cartesian space position and posture information of the master manipulator end, and calculates the master velocity factor D based on the displacement value at each moment. m The slave robot system collects image information based on its visual system and identifies the coordinates of the target position {X g Y g Z g}, the coordinates of the point closest to the surrounding environment and the end of the slave robot arm {X e Y e Z e}, and calculate the target distance factor D according to the actual posture information of the slave end g , Environmental safety factor De .

[0072] The gravity, friction and inertia of the main arm (collectively referred to as interference force) will be reflected at the end of the main arm in the form of the load at the end of the main arm and loaded onto the wearer's hand. Based on the force feedback control program of the force compensation algorithm, the control loop of the main end robot's mechanical arm (referred to as the main arm) adopts a feedforward feedback hybrid model. The feedforward control establishes the interference force model of the main arm's gravity, friction and inertia in advance. After adjusting the kf parameters through the joint status of the main arm during operation and the load data of the six-dimensional force sensor at the end of the main arm, the compensation torque that each joint should output is calculated in real time, and most of the interference forces of the main arm's links and joints are eliminated, so that the interference force of the main arm end on the human hand is reduced to a smaller range, so as to realize the feedback of the load force from the end to the end of the human hand, which eliminates the influence of interference forces such as the main arm's gravity and friction.

[0073] Calculate the master end rate factor D according to the displacement value at each moment m :

[0074]

[0075] Where, X ′ m is the position difference of the master end in the x-axis direction at this moment, Y ′ m is the position difference of the main end in the y-axis direction at this moment, Z ′ m is the position difference of the master end in the z-axis direction at this moment.

[0076] Calculate the target distance factor D based on the actual posture information of the slave terminal g :

[0077]

[0078] Where, X s 、Y s , Z s They are the position coordinates of the terminal at the current moment in the x-axis, y-axis, and z-axis directions, {X g Y g Z g} is the coordinate of the target position.

[0079] Calculate the environmental safety factor D based on the actual posture information of the slave terminal e :

[0080]

[0081] Where, X s 、Y s , Z sThey are the position coordinates of the terminal at the current moment in the x-axis, y-axis, and z-axis directions, {X e Y e Z e} is the coordinate of the point closest to the surrounding environment and the slave end.

[0082] Step 3: If Figure 3 , Mapping coefficient calculation and slave robot control: The master end rate factor D calculated in step 2 is m , target distance factor D e , Environmental safety factor D g Input the mapping coefficient network and record the output value of the mapping coefficient network, which is the mapping coefficient k at this moment; calculate the expected posture information of the slave end at this moment based on the mapping coefficient k and the Cartesian space posture information of the master end, and then perform inverse kinematics solution to realize joint control, and record the actual slave end load and actual posture information for the master end force compensation force feedback control and the slave end target distance factor D respectively. e , Environmental safety factor D g calculate.

[0083] Mapping coefficient network to target distance factor D e , Environmental safety factor D g , master-side rate factor D m For input {D e , D g , D m}, the mapping coefficient k is output {k}, and the network model learning, training and updating are performed in advance.

[0084] The training sample rules for network model learning and training are as follows:

[0085]

[0086] Among them, D safe is the safety margin factor, D e med is the target distance median factor, D g med is the median factor of environmental safety, D m med The median rate factor of the master end, k max is the maximum allowable mapping ratio coefficient related to the entire task. The above five parameters are pre-set according to the task attributes.

[0087] Sample set input array {D e , D g , D m} and the output array {k} are generated segment by segment according to the training sample rules:

[0088] (a) k=0 segment:

[0089] Array random generation:

[0090] D e = rand(1:100)D e max

[0091] D g = rand(1:100)D g max

[0092] D m = rand(1:100)D m max

[0093] Among them, D e max is the maximum value of the target distance factor, D g max is the maximum value of the environmental safety factor, D m max is the maximum value of the master - end rate factor.

[0094] D e ≤ D safe or D g ≤ D safe and D m Randomly select 100 groups of data as the input array through permutation and combination. The output array is always 0, and the array size is also selected as 100.

[0095] (b) 0 < k < 1 segment:

[0096] The input array {D e , D g , D m} is generated according to the following formula respectively: [[ID= June 2]]

[0097] D e (n) = n * (D e med - D safe ) / 100 + D safe

[0098] D g (n) = n * (D g med - D safe ) / 100 + D safe

[0099] D m (n) = n * (D m med ) / 100

[0100] Here, n is an array starting from 1 and gradually increasing by 1 to 100.

[0101] The output array {k} is generated according to the formula k(n)=n / 100.

[0102] (c) k = 1 segment:

[0103] Input array {D e , D g , D m} is {D e med , D g med , D m med}.

[0104] The output array {k} is {1}.

[0105] (d)1 <k<k max part:

[0106] Input array {D e , D g , D m} are generated according to the following formulas:

[0107] D e (n)=n*(D e max -D e med ) / 100+D e med

[0108] D g (n)=n*(D g max -D g med ) / 100+D g med

[0109] D m (n)=n*(D m max -D m med ) / 100+D m med

[0110] The output array {k} is based on the formula k(n)=n*(k max -1) / 100+1 generated.

[0111] Combine the input arrays and output arrays in sections (a) to (d) above to train the network model.

[0112] The network model adopts the RBF neural network model including input layer, hidden layer and output layer; the output of the input layer is the input of the hidden layer; the output of the hidden layer is the input of the output layer; the number of hidden layer nodes is 3-6 nodes.

[0113] Step 4: Use the visual system to observe in real time whether the slave end has completed the operation task. If the operation task has been completed, the entire control process ends. If the operation task has not been completed, the process of steps 2 and 3 is repeated.

[0114] The present invention has been described in detail above with reference to specific embodiments and exemplary examples. However, these descriptions should not be construed as limiting the present invention. Those skilled in the art will appreciate that various equivalent substitutions, modifications, or improvements may be made to the technical solutions and implementations of the present invention without departing from the spirit and scope of the present invention, all of which fall within the scope of the present invention. The scope of protection of the present invention shall be determined by the appended claims.

[0115] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. An adaptive variable scale mapping teleoperation control method, characterized in that: It includes: The S1 master robot obtains the Cartesian spatial position information of the end of the master manipulator and obtains the master rate factor D based on the Cartesian spatial position information of the end of the master manipulator. m ; The slave robot obtains the target position coordinates, the coordinates of the nearest point between the surrounding environment and the end of the slave manipulator, and the actual posture information of the end of the slave manipulator. The target distance factor D is obtained based on the target position coordinates, the coordinates of the nearest point between the surrounding environment and the end of the slave manipulator, and the actual posture information of the end of the slave manipulator. g and environmental safety factor D e ; S2 will factor the master rate D m , target distance factor D e , Environmental safety factor D g Input the mapping coefficient network and record the output value of the mapping coefficient network, which is the mapping coefficient k at this moment; Based on the mapping coefficient k and the Cartesian space pose information of the end of the master manipulator, obtain the desired pose information of the end of the slave manipulator; Perform joint control on the slave robot according to the desired pose information of the end of the slave manipulator to obtain the actual pose information of the end of the slave manipulator; When the operation task in S3 is completed, end the control process, otherwise repeat steps S1 and S2.

2. The adaptive variable-scaling mapping teleoperation control method according to claim 1, characterized in that: Where, X ′ m Y is the difference between the x-axis position of the end of the main end robot arm at the current moment and the x-axis position of the end of the main end robot arm at the previous moment. ′ m is the position difference of the end of the master end robot in the y-axis direction at the current moment, Z ′ m It is the position difference of the end of the master robot arm in the z-axis direction at the current moment.

3. The adaptive variable-scaling mapping teleoperation control method according to claim 1, characterized in that: Where, X s 、Y s 、Z s They are the position coordinates of the slave end robot arm in the x-axis, y-axis, and z-axis directions at the current moment, respectively. g 、Y g 、Z g are the coordinates of the target in the x-axis, y-axis, and z-axis directions.

4. The adaptive variable-scaling mapping teleoperation control method according to claim 1, characterized in that: Where, X s 、Y s 、Z s They are the position coordinates of the slave end robot arm in the x-axis, y-axis, and z-axis directions at the current moment, respectively. e 、X e 、X e The x-axis, y-axis, and z-axis coordinates of the point closest to the end of the slave robot arm.

5. The adaptive variable-scaling mapping teleoperation control method according to claim 1, characterized in that: It also includes: In step S1, the master robot executes a force feedback control instruction based on a force compensation algorithm; In step S2, record the actual load information of the end of the slave manipulator, and generate a force feedback control instruction based on the force compensation algorithm according to the actual load information of the end of the slave manipulator.

6. The adaptive variable-scaling mapping teleoperation control method according to claim 5, characterized in that: The force feedback control instruction based on the force compensation algorithm includes: The master manipulator control loop adopts a feedforward-feedback hybrid model. The feedforward control establishes a disturbance force model of the master manipulator in advance, and calculates the magnitude of the compensation torque that each joint should output in real time through the joint state during the operation of the master manipulator and the data of the six-dimensional force sensor at the end of the master manipulator, so as to eliminate the disturbance force of each link and joint of the master manipulator, and make the feedback force of the end of the master manipulator on the human hand be the slave load force after eliminating the influence of its own gravity and friction. The slave load force is the actual load information of the end of the slave manipulator.

7. The adaptive variable-scaling mapping teleoperation control method according to claim 1, characterized in that: The mapping coefficient network is based on the target distance factor D e , Environmental safety factor D g , master-side rate factor D m is input, the mapping coefficient k is output, and the network model learning and training is performed in advance.

8. The adaptive variable-scaling mapping teleoperation control method according to claim 7, characterized in that: The training sample rules for network model learning and training are as follows: Among them, D safe is the safety margin factor, D emed is the target distance median factor, D gmed is the median factor of environmental safety, D mmed The median rate factor of the master end, k max is the maximum allowable mapping ratio coefficient related to the entire task. The above five parameters are pre-set according to the task attributes.

9. The adaptive variable-scaling mapping teleoperation control method according to claim 8, characterized in that: Sample set input array {D e , D g , D m } and the output array {k} are generated segment by segment according to the training sample rules: For the segment of k = 0: Array is randomly generated: D e =rand(1:100)D emax D g =rand(1:100)D gmax D m =rand(1:100)D mmax Among them, D emax is the maximum value of the target distance factor, D gmax is the maximum value of the environmental safety factor, D mmax The maximum value of the master-end rate factor; D e ≤D safe or D g ≤D safe With D m The permutation and combination randomly selects 100 groups of data as the input array, the output array is always 0, and the array selects 100; For the segment of 0 < k < 1: Input array {D e , D g , D m } are generated according to the following formulas: D e (n)=n*(D emed -D safe ) / 100+D safe D g (n)=n*(D gmed -D safe ) / 100+D safe D m (n)=n*(D mmed ) / 100 Where, n is an array starting from 1 and incrementing by 1 step by step to 100; The output array {k} is generated according to the formula k(n) = n / 100; For the segment of k = 1: Input array {D e , D g , D m } is {D emed , D gmed , D mmed }; The output array {k} is {1}; 1 <k<k max part: Input array {D e , D g , D m } are generated according to the following formulas: D e (n)=n*(D emax -D emed ) / 100+D emed D g (n)=n*(D gmax -D gmed ) / 100+D gmed D m (n)=n*(D mmax -D mmed ) / 100+D mmed The output array {k} is based on the formula k(n)=n*(k max -1) / 100+1 generated.

10. The adaptive variable-scaling mapping teleoperation control method according to claim 8, characterized in that: The network model adopts an RBF neural network model, including an input layer, a hidden layer and an output layer; The output of the input layer is the input of the hidden layer; the output of the hidden layer is the input of the output layer; The number of nodes in the hidden layer is 3 - 6 nodes.

Images