Anti-slip gripping method and system for handling robot with dynamic adjustment of gripping force

Abstract

The application discloses a carrying robot anti-slip grabbing method and system with dynamic adjustment of clamping force, and belongs to the field of carrying robot grabbing control. The method comprises the following steps: synchronously collecting clamping force, jaw displacement and driving current, and constructing a grabbing timing signal; analyzing a mechanical arm motion instruction, extracting an action node and generating an event window; calculating displacement accumulation, clamping force change rate and current fluctuation, inputting a slip discrimination network, and obtaining a risk level and a risk trend associated with an event type; making a hierarchical determination according to the comparison result of the risk level and a risk threshold and the risk trend, superimposing a corresponding clamping force increase to obtain a target clamping force, and applying a maximum clamping force constraint; tracking the target value of the clamping force through closed-loop control; and when the target clamping force has reached the maximum and the risk is still high, generating an abnormal treatment instruction and triggering release. The application realizes hierarchical clamping force adjustment for an action stage, balances between anti-slip and anti-crushing, and effectively improves carrying stability and safety.

Description

Technical Field

[0001] This invention relates to the field of gripping control for handling robots, and in particular to a method and system for preventing slippage in handling robots by dynamically adjusting the gripping force. Background Technology

[0002] In scenarios such as building construction, prefabricated component stacking, and handling, handling robots often use end effectors to grip and place bricks or blocks. Gripping stability is easily affected by factors such as dust, moisture, and wear on the contact surface; fluctuations in the coefficient of friction and effective contact area can easily cause workpiece slippage. Simultaneously, the inertial loads and posture changes generated during the robotic arm's closing, lifting, acceleration, deceleration, turning, and lowering actions alter the gripping force distribution, increasing the risk of slippage or even falls; while excessive gripping force can cause chipping and damage to the block's edges and corners, requiring a precise balance between anti-slip and anti-damage measures.

[0003] Existing technologies often employ preset fixed clamping forces or empirical segmented compensation, failing to differentiate force characteristics for different stages of the operation. This results in lag in clamping force response and mismatched adjustment ranges. Slip detection relies heavily on single thresholds or a limited number of sensing features, making it sensitive to changes in operating conditions. Signal drift can easily lead to misjudgments or missed detections, triggering unnecessary additional clamping or safety protection failures. Furthermore, when the clamping force approaches the material's maximum tolerance, most systems lack a safety mechanism linked to the robotic arm's movement, failing to provide reliable protection under extreme conditions and making it difficult to meet the stable and safe handling requirements under complex working conditions.

[0004] Therefore, there is an urgent need for a method to prevent slippage and gripping of handling robots that can overcome the shortcomings of the existing technologies. This is a problem that needs to be solved by those skilled in the art. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for anti-slip gripping of a handling robot with dynamic adjustment of clamping force. During the entire process from the closing to the placement of bricks or blocks, in the face of changes in slippage risk caused by uncertainties in different motion stages and friction conditions, the system achieves slippage risk identification and dynamic adjustment of clamping force for different motion stages, under the constraint of not exceeding the maximum allowable clamping force of the material. When the clamping force margin is exhausted and the risk is still high, the system will coordinate with the robotic arm to complete the safe handling. On the one hand, a method for preventing slippage in a handling robot with dynamically adjustable clamping force is provided, the method comprising: At each sampling moment, the gripping force feedback, gripper displacement and drive current during the gripper closing and transport process are acquired synchronously to obtain the gripping timing signal; Analyze the robotic arm's motion commands, extract action nodes such as closing, lifting, acceleration / deceleration, turning, and lowering, segment the grasping timing signal according to the action nodes and label the event type to obtain the event window; The cumulative displacement, clamping force change rate, and current fluctuation of the event window are obtained and input into the slip discrimination network to obtain slip risk parameters associated with the event type. The slip risk parameters are compared with the preset risk threshold to determine the risk level and obtain the matching target clamping force, and the maximum clamping force constraint is applied to the target clamping force. The clamping force error is calculated based on the target clamping force and the clamping force feedback, and a clamping drive command is generated based on the clamping force error to make the clamping force feedback track the target clamping force. When the target clamping force reaches the maximum clamping force and the risk level of the slippage risk parameter meets the second risk threshold, an abnormal handling instruction is generated; otherwise, a transport instruction is generated. The abnormal handling instruction or transport instruction is integrated with the clamping drive instruction to form an anti-slip gripping control instruction and applied to the robotic arm.

[0006] On the other hand, a handling robot anti-slip gripping system with dynamically adjustable clamping force is provided, the system comprising: The signal acquisition module synchronously acquires the gripping force feedback, gripper displacement and drive current during the gripper closing and transport process at each sampling moment to obtain the gripping timing signal. The event window generation module parses the robotic arm's motion commands, extracts action nodes such as closing, lifting, acceleration and deceleration, turning, and lowering, segments the grasping timing signal according to the action nodes and labels the event type to obtain the event window; The slip risk discrimination module obtains the cumulative displacement, clamping force change rate, and current fluctuation of the event window, inputs them into the slip discrimination network, and obtains slip risk parameters associated with the event type. The hierarchical decision-making module compares the slip risk parameters with the preset risk threshold to determine the risk level and obtain the matching target clamping force, and applies the maximum clamping force constraint to the target clamping force. The closed-loop control module calculates the clamping force error based on the target clamping force and the clamping force feedback, and generates clamping drive commands based on the clamping force error, so that the clamping force feedback tracks the target clamping force. The instruction fusion module generates an abnormal handling instruction when the target clamping force reaches the maximum clamping force and the risk level of the slippage risk parameter meets the second risk threshold; otherwise, it generates a transport instruction. The abnormal handling instruction or transport instruction is fused with the clamping drive instruction to form an anti-slip gripping control instruction and applied to the robotic arm.

[0007] The above technical solution has the following advantages or beneficial effects: The gripping force dynamically adjusted anti-slip gripping method for handling robots provided in this application constructs gripping timing signals and analyzes the robotic arm's motion commands to achieve event window segmentation for different action stages such as closing, lifting, acceleration / deceleration, turning, and lowering. Then, combined with multimodal features, it utilizes the risk level and risk trend associated with the output of the ModernTCN slip discrimination network and the event type. Based on this, the target gripping force is obtained by superimposing the corresponding first or second clamping increment based on the graded judgment results and applying the maximum gripping force constraint. Then, closed-loop control makes the gripping force track the target value in real time. When the target gripping force has reached the upper limit and the risk is still high, an abnormal handling command is generated and fused with the gripping drive command for output. Thus, adaptive gripping force adjustment and accurate slip risk identification are achieved for each action stage throughout the handling process, reducing misjudgments and omissions caused by signal fluctuations under different working conditions. While effectively preventing bricks or blocks from being crushed by excessive gripping force, the graded clamping increment and abnormal handling linkage mechanism significantly improves gripping stability and handling safety. Attached Figure Description

[0008] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0009] Figure 1 The flowchart shows the anti-slip gripping method for a handling robot with dynamically adjustable clamping force proposed in this invention. Figure 2 This is a flowchart illustrating the construction of reference parameters and gripping timing signals for the dynamic adjustment of gripping force anti-slip gripping method for handling robots proposed in this invention. Figure 3 The flowchart for motion node extraction and event window generation of the dynamic clamping force adjustment anti-slip gripping method for handling robots proposed in this invention is shown below. Figure 4 The ModernTCN slip discrimination network inference flowchart is shown below for the anti-slip gripping method for handling robots with dynamic adjustment of clamping force proposed in this invention. Figure 5 This is a flowchart illustrating the graded determination and target clamping force generation process of the dynamic adjustment clamping force anti-slip gripping method for handling robots proposed in this invention. Figure 6 This is a flowchart of the clamping force error closed loop and clamping drive command generation process of the dynamic clamping force adjustment anti-slip gripping method for handling robots proposed in this invention. Figure 7 This is a flowchart illustrating the fusion of anomaly handling and anti-slip gripping control commands in the dynamic adjustment of clamping force anti-slip gripping robot method proposed in this invention. Figure 8This is a schematic diagram of the gripper-brick contact and force constraint of the dynamic adjustment of gripping force anti-slip grasping method for a handling robot proposed in this invention. Figure 9 This is a schematic diagram comparing the traditional fixed clamping force of the dynamic clamping force adjustment method for anti-slip gripping of a handling robot proposed in this invention with the event-triggered clamping and closed-loop tracking anti-slip gripping method of this invention. Figure 10 This is a schematic diagram of the abnormal handling and safe placement process of the dynamic adjustment of clamping force anti-slip gripping robot proposed in this invention. Figure 11 The ModernTCN slip discrimination network diagram is shown for the anti-slip gripping method of the handling robot with dynamic adjustment of clamping force proposed in this invention. Detailed Implementation

[0010] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0011] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the invention. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0012] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0013] All data acquisition in this embodiment is carried out in accordance with laws and regulations and with user consent, and the data is used legally.

[0014] Example 1 like Figure 1 As shown, this embodiment provides a method for preventing slippage during handling by a transport robot based on a dynamic adjustment strategy for gripping force. The method includes: S1. Simultaneously acquire the gripping force feedback, gripper displacement, and drive current during the gripper closing and transport process at each sampling time to obtain the gripping timing signal, such as... Figure 2 As shown.

[0015] Specifically, in this embodiment, a gripper is installed at the end of the robotic arm, the gripper is connected to a gripping actuator, and the gripper is equipped with a gripping force sensor, a gripper displacement sensor, and a drive current acquisition module.

[0016] The moment when the grippers begin to close is taken as the data acquisition trigger moment, and the trigger moment is denoted as . Set a fixed sampling period. At each sampling time Synchronous acquisition of clamping force feedback gripper displacement and drive current The data collection process is as follows: The process begins by continuously covering the closing and transporting of the grippers, and continues until the end of the placement action, so that the gripping timing signal includes the state change segments corresponding to closing, lifting, acceleration / deceleration, turning, and placement.

[0017] Using a robotic arm to control a clock as a unified time reference, for , and Perform sampling time alignment; when any sensor output has a sampling phase deviation, map the timestamp reading of that sensor to a unified sampling time sequence and use linear interpolation to obtain the corresponding... The readings form the same The corresponding triplet data. Then, the triplet data from each sampling time point are assembled into a three-channel capture timing signal in chronological order. The displacement of the gripper Used for subsequent displacement accumulation calculation and clamping force feedback. Used for subsequent calculation of the rate of change of clamping force, driving current This is used for subsequent current fluctuation calculations, thereby providing a consistent three-channel timing basis for the ModernTCN slip discriminant network input to the event window.

[0018] The three-channel capture timing signal can be further specified as follows:

[0019] in, For the first Each sampling time, This represents the gripper displacement at that sampling moment. This is the clamping force feedback at that sampling moment. This represents the driving current at that sampling moment. This represents the total number of sampling points.

[0020] S2. Receive the robotic arm's motion commands and extract the action nodes for closing, lifting, acceleration / deceleration, turning, and lowering; segment the grasping timing signal according to the action nodes and label the event type to obtain the event window. The specific process is as follows: Figure 3 As shown.

[0021] Specifically, in this implementation method, the robotic arm controller receives robotic arm motion commands and forms a sequence of commands updated according to the control cycle, recording the robotic arm motion commands as follows: The controller has... The field is parsed to obtain the gripper opening and closing settings arranged over time. End-effector pose setting and speed setting .in This indicates the setting value for the distance between the two fingers of the gripper. This includes end position setting and end direction setting. The vertical position component of the end position setting is denoted as... The end direction setting is denoted as The control cycle is denoted as To ensure that the determination of action termination has an objective standard, a preset number of stable determination control cycles is established. And preset the gripper opening and closing stability threshold respectively. Vertical position stability threshold Speed ​​stability threshold and angular velocity stability threshold .

[0022] The controller first The input is used to extract the closing action node. The gripper open state setting value is recorded as... The clamping state setting value is recorded as The clamping threshold is denoted as .when Depend on Towards Change and first crossing When this control moment is determined, it is set as the start moment of the closing action node; when achieve And in continuous The variation within each control cycle does not exceed When the corresponding control moment is determined, it will be the termination moment of the closing action node.

[0023] Subsequently To extract the lifting action node from the input, the vertical position setting value at the termination time of the closing action node is recorded as... The lifting displacement threshold is denoted as ,when relatively The increase exceeded for the first time At that time, determine the start time of the lifting action node; when In continuous The variation within each control cycle does not exceed At that time, determine the termination time of the lifting action node.

[0024] Again Extract acceleration / deceleration action nodes from the input and calculate the acceleration settings. The acceleration threshold is denoted as ,when First time exceeding When, determine the start time of the acceleration / deceleration action node; when In continuous No more than [amount] within each control cycle and The range of change within the corresponding interval does not exceed At that time, determine the termination time of the acceleration / deceleration action node.

[0025] Again To extract steering action nodes as input, for adjacent control cycles Calculate the direction change angle corresponding to the relative attitude change, and divide the direction change angle by Obtain angular velocity setting The angular velocity threshold is denoted as ,when First time exceeding The start time of the steering action node is determined at time; when In continuous No more than [amount] within each control cycle And the range of change within the corresponding interval does not exceed The timing of the steering action termination point is determined.

[0026] Finally To extract the dropping action node from the input, the gripper opening threshold is denoted as... ,when Depend on Towards Change and first crossing At that time, determine the start time of the drop-off action node; when achieve And in continuous The variation within each control cycle does not exceed At that time, determine the termination time of the drop action node.

[0027] The controller reads and captures timing signals. Sampling time sequence With sampling period The start time of each action node is mapped to the index of the nearest sampled time not greater than the start time, and the end time of each action node is mapped to the index of the nearest sampled time not less than the end time, resulting in segment boundary indices arranged in chronological order. If the segment boundary indices of adjacent action nodes overlap, the segment end index of the preceding action node is adjusted to the index of the previous sampled time before the segment start index of the following action node, ensuring that any sampled time falls within only one segment interval. The controller then adjusts the segment boundary indices accordingly. By extracting segments in chronological order, segmented signals corresponding to each action node are obtained.

[0028] The controller labels each segment signal as a unique event type, and the event type is denoted as... , Select from the closing event type, lifting event type, acceleration / deceleration event type, steering event type, and lowering event type. The controller will then match the segmented signals with the corresponding... Combine to get the event window ,in It contains three-channel captured time-series signal segments arranged in order of sampling time, with a sampling period of The sampling time is recorded as ,in The sampling point number is denoted as , and the total number of sampling points in the event window is denoted as . The controller will handle event types. Mapped to the event type output branch index of the ModernTCN sliding discriminant network And in the following When inputting the ModernTCN sliding discriminant network, the following is used: Choice and The matching event type output branch allows the same event window to output the slip risk parameters associated with the event type under different event types.

[0029] S3. Extract the cumulative displacement, clamping force change rate, and current fluctuation of the event window, and input the calculation results into the slip discrimination network to obtain slip risk parameters associated with the event type. These slip risk parameters include risk level and risk trend. The specific process is as follows: Figure 4 As shown.

[0030] Compared to general classification models that process signals with a uniform threshold or do not distinguish between action stages, this design matches the slip risk assessment with the force mechanism of specific handling actions and characterizes the direction of risk change within the window through risk trend information. This reduces misjudgments and omissions caused by signal fluctuations under different working conditions, making the risk judgment results more suitable for driving subsequent clamping control decisions.

[0031] Specifically, in this implementation method, S3 includes: S301: Event Receiver Window Event type Extract the gripper displacement subsequence according to the sampling time. Clamping force feedback subsequence and driving current subsequence .

[0032] With Event Window As a processing unit of the ModernTCN slip discrimination network, the controller receives event windows. Event type And according to the sampling time from Extract the gripper displacement subsequence sequentially Clamping force feedback subsequence and driving current subsequence ,in correspond Capture the results within the event window. correspond Capture the results within the event window. correspond Capture the results within the event window.

[0033] S302: Based on the gripper displacement subsequence Clamping force feedback subsequence and driving current subsequence Calculate the time-series feature sequence, which includes the displacement cumulative sequence. Clamping force change rate sequence Current fluctuation sequence .

[0034] The displacement cumulative sequence for k=2,3,…,N in, This represents the k-th sampling time within the event window; Indicates the sampling time within the event window. The collected gripper displacement values, where N is the total number of sampling points within the event window; the cumulative displacement at the first sampling moment is initialized to zero.

[0035] The clamping force change rate sequence for k=2,3,…,N in, This represents the clamping force feedback value at the k-th sampling moment within the event window; The sampling period is N; the total number of sampling points within the event window is N; the rate of change of clamping force at the first sampling moment is initialized to zero.

[0036] The current fluctuation sequence for k=2,3,…,N in, The current value is the driving current value at the kth sampling moment within the event window; N is the total number of sampling points within the event window; the current fluctuation at the first sampling moment is initialized to zero.

[0037] S303: Construct a three-channel input sequence based on the temporal feature sequence.

[0038] The controller will , , Align them at the same sampling time and construct a three-dimensional feature vector at each sampling time. ,

[0039] All sampling times The three-channel input sequence is formed by stacking them in chronological order. ,

[0040] in, For the first The total number of sampling points within each event window.

[0041] S304: Input the three-channel input sequence into the ModernTCN slip discrimination network to obtain the slip risk associated with the event type. .

[0042] The ModernTCN glide discriminant network consists of dilated causal convolutional residual blocks and sets output branches according to event type.

[0043] like Figure 11As shown, the ModernTCN sliding discriminant network consists of an ingress causal convolutional layer, multi-layer dilated causal convolutional residual blocks, a temporal convergence part, a trend extraction part, and output branches set according to event type. The ingress causal convolutional layer is used to perform causal convolutional encoding on the three-channel input sequence and outputs intermediate feature maps while keeping the sequence length unchanged. The multi-layer dilated causal convolutional residual blocks are used to extract features at different temporal scales layer by layer. Each dilated causal convolutional residual block includes two dilated causal convolution processes, a nonlinear activation process between the two convolutions, and a residual connection. When the number of channels is inconsistent, alignment is completed through a channel mapping convolutional layer. The temporal convergence part performs channel averaging on the end feature map in the temporal dimension to obtain a window representation vector. The trend extraction part divides the event window into a front segment and a back segment in chronological order, obtaining a front segment convergence vector and a back segment convergence vector, respectively, and obtains a trend representation vector through normalized difference. Finally, the window representation vector and the trend representation vector are concatenated, and the corresponding event type output branch is selected according to the event type. The two-dimensional sliding risk parameters, namely the risk level and risk trend, are output through a fully connected layer.

[0044] S3041: A causal convolutional layer that takes a three-channel input sequence as input to obtain an intermediate feature map of constant length.

[0045] Will The input to the ModernTCN glide discriminant network is the ingress causal convolutional layer, and the kernel length of the ingress causal convolutional layer is denoted as . The number of output channels is denoted as The method employs causal-constrained convolution to utilize only the input features from the current and historical sampling times, and maintains the output sequence length by padding the input sequence with zeros, thereby obtaining intermediate feature maps. ,in Corresponding to each sampling time 3D channel characteristics.

[0046] S3042: Transfer intermediate features Figure 1 The input is a multi-layer dilated causal convolutional residual block, which is then used to obtain the block output feature map.

[0047] The controller will use the intermediate feature map Input multi-layer dilated causal convolution residual blocks sequentially, the total number of residual blocks is denoted as . Each dilated causal convolution residual block is configured to perform dilated causal convolution on the input feature map twice, introduce a non-linear activation function between the two convolutions, and finally fuse the processed feature map with the input feature map through residual connections to obtain the block output feature map.

[0048] Specifically, taking the first Taking an extended causal convolutional residual block as an example, it receives the output feature map from the previous layer. The input feature map first passes through a first dilated causal convolutional layer to obtain the first convolutional output, then through a nonlinear activation layer to perform a pointwise nonlinear transformation, then through a second dilated causal convolutional layer to obtain the second convolutional output, and finally through a residual summing layer to combine the second convolutional output with the input feature map. The block output feature map is obtained by adding the channel values ​​at the same sampling time. . No. The first expansion rate of each residual block is denoted as... The second expansion rate is denoted as The values ​​are set to different values ​​to enhance the network's ability to extract features from different time series. If the number of channels in the two feature maps of the residual addition layer is inconsistent, a channel mapping convolutional layer is first used to adjust the number of channels in the input feature map to be consistent with the output of the second convolution, and then the residual addition is performed.

[0049] S3043: Calculate the channel average of the block output feature map in the time dimension to obtain the window representation vector.

[0050] Window representation vector Feature map output via end block Obtained by channel averaging over the time dimension:

[0051] in, Let be the channel feature vector at the k-th sampling time; This represents the total number of sampling moments within event window i.

[0052] S3044: Divide the event window into a front segment and a back segment in chronological order, obtain the front segment convergence vector and the back segment convergence vector, and calculate the trend representation vector.

[0053] The front-end convergence vector and backend convergence vector The calculation formula is as follows:

[0054]

[0055] in, Let be the channel feature vector at the k-th sampling time; This represents the total number of sampling moments within event window i. This is the number of sampling points in the front end, usually taken as... The first half of the sampling points; This represents the number of sampling points in the later stage.

[0056] The trend representation vector is calculated using the normalized difference method. as follows: ; in, The trend is represented by a vector; This is the convergence vector for the later segment; For the front-end convergence vector; symbol Represents the first norm and is calculated as the sum of the absolute values ​​of each dimension of the vector; The numerator is a preset positive constant; the numerator is a vector subtracted dimension by dimension, and the denominator is a scalar. The normalized difference method divides each dimension of the numerator vector by the denominator scalar to obtain the normalized vector output.

[0057] S3045: Concatenate the window representation vector with the trend representation vector, select the corresponding output branch according to the event type, and obtain the slip risk parameter associated with the event type.

[0058] Represent the vector in a window Trend representation vector The branch input vector is formed by concatenating the vectors along the dimensions. And according to the event type Select the event type output branch of the ModernTCN slip discrimination network, and Input the selected event type output branch. The event type output branch uses a fully connected layer to output two-dimensional slip risk parameters, with the risk level denoted as [risk level]. Risk trends are denoted as ; regarding risk level Execution scope limitation processing, when When less than zero Set to zero, when When it is greater than one Set as one; regarding risk trends Execution scope limitation processing, when When less than negative one, Set to negative one, when When it is greater than one Set to one, thus outputting As a slip risk parameter associated with the event type, it is input into the subsequent classification and clamping force adjustment process.

[0059] S4. Obtain the baseline clamping force, maximum clamping force, first risk threshold, and second risk threshold. Classify the risk level based on the slip risk parameters and the first and second risk thresholds. When the risk level meets the first risk threshold, determine the first clamping increment based on the event type and add it to the baseline clamping force to obtain the target clamping force. When the risk level meets the second risk threshold or the risk trend is rising, determine the second clamping increment based on the event type and add it to the baseline clamping force to obtain the target clamping force. Apply the maximum clamping force constraint to the target clamping force. The specific process is as follows: Figure 5 As shown.

[0060] This strategy differs from empirical methods that use fixed clamping force or uniform compensation amount. It allows the clamping range to vary with the stage of action and the level of risk, increasing the anti-slip margin when needed and avoiding ineffective clamping when not needed. It also completes risk-driven clamping force adjustment within the material strength boundary, taking into account both anti-slip and anti-damage.

[0061] Specifically, in this implementation method, S4 includes: The object to be grabbed is a brick or block, and its specifications include its external dimensions and weight.

[0062] The reference clamping force

[0063]

[0064] Where G is the gravity load during the transportation process; The lower limit of the preset friction coefficient is obtained from offline friction tests under dusty or humid conditions and written into the control parameters.

[0065] Will The reference clamping force is allocated to each clamping contact surface according to the number of clamping contact surfaces of the jaws, and is used to establish the initial mechanical boundary during the jaw closing stage.

[0066] The maximum clamping force

[0067]

[0068] in, The compressive strength is provided by material testing data or batch quality data of the brick or block; Effective contact area; This is the preset compressive safety factor.

[0069] like Figure 8 The geometric relationship between the clamps, bricks, and force constraints shown indicates that, to ensure the bricks are not crushed, the maximum clamping force must be constrained by the compressive strength of the material; at the same time, by adjusting the clamping force, the frictional force is made greater than the sliding driving force to achieve the purpose of anti-slip.

[0070] when hour, = This ensures that the reference clamping force is less than the maximum clamping force and reserves constraint space for subsequent event-driven clamping additions, where This is a preset safety margin.

[0071] In this embodiment To normalize the scalar output, the value range is limited to... to The first risk threshold is denoted as... The second risk threshold is denoted as ,Will and Settings to Within the range and satisfy Greater than When selecting the threshold, multiple sets of event window samples labeled as having or not having slipped are collected, and the percentages of samples that did not slip are statistically analyzed. Distribution and slippage of samples Distribution, will Set the upper bound threshold for samples that have not slipped, and then... Set as the lower bound threshold for samples that slip.

[0072] The controller is equipped with a graded judgment unit, an increase in clamping amount determination unit, and a clamping constraint unit, which are used to execute the following steps.

[0073] The hierarchical determination unit for event-driven clamping force closed-loop adjustment uses an event window. event types Slippage risk parameters, first risk threshold With the second risk threshold As input, output increment level ,in This is the event window number.

[0074] The controller will classify the risk level Compared with the first risk threshold The comparison generates the first judgment result. Used to characterize risk level Does it meet the first risk threshold? , Using binary identifiers, if ,but ,otherwise Risk level With the second risk threshold A second judgment result is generated through comparison. Used to characterize risk level Does it meet the second risk threshold? ,like ,but ,otherwise .

[0075] Risk level Risk Trends First Judgment Result Compared with the second judgment result Input the classification determination unit.

[0076] The classification and judgment unit assesses risk trends. Perform trend direction analysis and generate an upward trend indicator. Or a non-upward trend indicator. The grading judgment unit presets a trend judgment threshold. Trend determination threshold Risk Trends Using the same unit of measurement, the trend determination threshold is used in this embodiment. Set to zero; when hour, And the non-upward trend indicator is set to 0; otherwise, Furthermore, the non-upward trend indicator is set to 0.

[0077] The grading determination unit is based on the second determination result. Upward trend indicator Compared with the first judgment result Generate clipping levels Increase the clamping level The value set includes zero-address level, first-address level, and second-address level, denoted as follows: , , In this embodiment, the zero-addition level is coded as The first level of clipping is coded as The second additional level is coded as The grading determination unit uses the following formula to calculate the level of clipping. : ; in, To increase the level of clamping; The zero-incremental-level encoding value; This is the code value for the first level of clipping. This is the second-level code value; The second determination result is set to either zero or one. An upward trend is indicated and its value is either zero or one. The first determination result is set to either zero or one; This indicates the relationship between the numerical value and the numerical value. The operation of taking the smaller value between the two is used to determine the second judgment result. With upward trend indicator The triggering relationship is mapped to zero or one; the numerical value 1 in the formula is a constant.

[0078] The clamping amount determination unit pre-stores a first clamping amount set and a second clamping amount set. The first clamping amount, representing the closing event type, is denoted as... The first increase in the lifting event type is recorded as... The first increment of the acceleration / deceleration event type is denoted as The first increment of the turning event type is denoted as The first increment of the drop event type is recorded as . The second increment of the closing event type is denoted as... The second increase in the lifting event type is recorded as... The second increase in the acceleration / deceleration event type is denoted as... The second increment of the turning event type is denoted as The second increment of the drop event type is recorded as... Size validation is performed for each event type during parameter configuration to ensure... , , , , The verified clamping amount value is then written into the parameter table of the clamping amount determination unit.

[0079] The unit for determining the amount of clamping depends on the event type. With increasing the level Select the clamping increment, denoted as . When the level of clamping is increased For the first level of clipping At that time, Set as event type The first increment of the matching clamp; when the clamp level For the second level of additional clamping At that time, Set as event type Matching the second clipping amount; when the clipping level Zero-addition level At that time, Set it to zero.

[0080] Obtain the target clamping force ,in For the established reference clamping force value, This is the target clamping force value used for subsequent error closed-loop calculations.

[0081] The clamping constraint unit uses maximum clamping force Clamping force on target Apply maximum clamping force constraint, when Greater than When Limited to ,when Not greater than Keep constant.

[0082] like Figure 9As shown, a) represents the traditional fixed clamping force strategy (dashed line), where the clamping force does not change with the action stage or slippage risk, easily leading to under-clamping or over-clamping; b) represents the event-triggered clamping increase and closed-loop tracking curve of this invention. Within event windows such as closing, lifting, acceleration / deceleration, turning, and lowering, the clamping force is increased in stages according to the slippage risk level and risk trend, so that the clamping force increases rapidly when the risk increases and decreases reasonably when the risk decreases, thus suppressing relative slippage.

[0083] S5. Calculate the clamping force error based on the target clamping force and the clamping force feedback, and generate a clamping drive command based on the clamping force error so that the clamping force feedback tracks the target clamping force.

[0084] Specifically, such as Figure 6 As shown, in this embodiment, step S5 includes: S501: At each sampling moment of capturing the timing signal Perform closed-loop adjustment of clamping force, obtain target clamping force and clamping force feedback, and obtain clamping force error.

[0085] The clamping force error

[0086] in, The target clamping force after being constrained by the maximum clamping force; This is for clamping force feedback.

[0087] S502: Adjusting clamping force error Input error closed-loop unit to obtain clamping adjustment amount , Proportional adjustment component

[0088] Integral cumulative amount and integral adjustment component

[0089] in

[0090]

[0091] Differential adjustment component

[0092] Clamping adjustment amount

[0093] in, For proportional adjustment components; For proportional gain; This is the clamping force error; This refers to the accumulated points. Adjust the component for integral; This is the integral gain; Update the identifier for the target clamping force when When the target clamping force is greater than or equal to the target clamping force update threshold, Otherwise, it is 0; This is the differential adjustment component; This is the differential gain.

[0094] S503: Adjust the clamping amount Converted into drive current command for clamping actuator ,right Applying amplitude and rate of change constraints to obtain clamping drive commands .

[0095] The original current command

[0096] in, The current conversion coefficient is obtained by applying multiple sets of calibrated currents to the clamping actuator and measuring the corresponding steady-state clamping force to establish a conversion relationship. The compensation current is set under zero clamping force. The amplitude constraint uses an upper current limit. With current lower limit right Limit:

[0097] The rate of change constraint uses the maximum change amount. Limit current variation between adjacent sampling times:

[0098] in, This represents the maximum allowable change in current between adjacent sampling times.

[0099] S504: Clamping drive command Send to the clamping actuator, the clamping actuator according to Adjust the gripper force and output gripper force feedback at subsequent sampling times. The controller continues reading at the next sampling time. and Repeat steps S501 to S503 to provide clamping force feedback. Tracking target clamping force under closed-loop regulation .

[0100] S6. When the target clamping force reaches the maximum clamping force and the risk level of the slippage risk parameter meets the second risk threshold, an abnormal handling command is generated; otherwise, a transport command is generated. The abnormal handling command or transport command is integrated with the clamping drive command to form an anti-slip gripping control command and applied to the robotic arm.

[0101] This linkage mechanism provides a consistent safety handling path for scenarios where "the clamping force has reached its limit but the risk of slippage remains high," avoiding the control uncertainty caused by simply increasing clamping force or decelerating alone, thereby improving the stability and safety of the handling process.

[0102] Specifically, such as Figure 7 As shown, in this embodiment, step S6 includes: S601: Abnormal trigger condition judgment. When the target clamping force reaches the maximum clamping force and the risk level meets the second risk threshold, an abnormal handling instruction is generated.

[0103] The controller first and Perform consistency determination when Write as after being constrained by the clamping constraint unit When the target clamping force reaches its maximum, the controller then performs a threshold judgment on the risk level. Not less than The risk level is determined to meet the second risk threshold.

[0104] When the same event window corresponds to multiple sampling times, it remains within the sampling interval of that event window. Take the output value of the event window so that the risk level participates in consistent anomaly determination within the same event window.

[0105] S602: Perform branch processing When both of the above conditions are met, an exception trigger flag will be displayed. Set to 1 and generate exception handling instructions. Abnormal handling instructions The speed setting of the robotic arm's motion commands is reset to zero. Speed ​​settings Write it to zero and set the drop action node trigger flag. Write it as one; the controller will The input condition for the action node jump is used to input the robotic arm motion state machine, so that the robotic arm motion state machine exits the current handling trajectory segment and enters the execution process of the placement action node in the subsequent control cycle.

[0106] If the target clamping force does not reach the maximum clamping force or the risk level does not meet the second risk threshold, the received robotic arm motion command will be treated as a handling command. The controller will then display an abnormal trigger flag. Set to zero and record the transport instruction as ,in Take directly At sampling time The original field value enables the robotic arm to continue executing the transport trajectory according to the robotic arm motion command.

[0107] like Figure 10 As shown, when the target clamping force has reached the maximum clamping force and the slippage risk level still meets the second risk threshold, the system triggers the abnormal handling procedure, sets the speed of the robotic arm to zero and safely lowers it, and avoids the brick from falling or being crushed when it is impossible to continue to safely transport it.

[0108] S603: Integrates abnormal handling instructions or handling instructions with gripping drive instructions into anti-slip gripping control instructions and outputs them to the robotic arm.

[0109] The anti-slip gripping control command is denoted as The anti-slip gripping control command includes two parts: robotic arm movement sub-commands and gripping sub-commands. The controller uses a command selector based on anomaly trigger flags. Select the robotic arm motion sub-command when For a momentary choice and cover ,when Choose when zero The controller will permanently write the clamping sub-command as... The robotic arm motion sub-commands and gripping sub-commands are combined to form ,Will The output is sent to the robotic arm controller, so that abnormal handling and clamping closed-loop adjustment take effect simultaneously in the same control command.

[0110] Example 2 This embodiment provides an anti-slip gripping system for a handling robot based on a dynamic clamping force adjustment strategy, including: The signal acquisition module synchronously acquires the gripping force feedback, gripper displacement and drive current during the gripper closing and transport process at each sampling moment to obtain the gripping timing signal. The event window generation module parses the robotic arm's motion commands, extracts action nodes such as closing, lifting, acceleration and deceleration, turning, and lowering, segments the grasping timing signal according to the action nodes and labels the event type to obtain the event window; The slip risk discrimination module obtains the cumulative displacement, clamping force change rate, and current fluctuation of the event window, inputs them into the slip discrimination network, and obtains slip risk parameters associated with the event type. The hierarchical decision-making module compares the slip risk parameters with the preset risk threshold to determine the risk level and obtain the matching target clamping force, and applies the maximum clamping force constraint to the target clamping force. The closed-loop control module calculates the clamping force error based on the target clamping force and the clamping force feedback, and generates clamping drive commands based on the clamping force error, so that the clamping force feedback tracks the target clamping force. The instruction fusion module generates an abnormal handling instruction when the target clamping force reaches the maximum clamping force and the risk level of the slippage risk parameter meets the second risk threshold; otherwise, it generates a transport instruction. The abnormal handling instruction or transport instruction is fused with the clamping drive instruction to form an anti-slip gripping control instruction and applied to the robotic arm.

[0111] The proposed system can be implemented in other ways. For example, the system embodiments described above are merely illustrative, and the division of modules described above is only a logical functional division. In actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed.

Claims

1. A method for preventing slippage in a handling robot with dynamically adjustable clamping force, characterized in that, include: At each sampling moment, the gripping force feedback, gripper displacement and drive current during the gripper closing and transport process are acquired synchronously to obtain the gripping timing signal; Analyze the robotic arm's motion commands, extract action nodes such as closing, lifting, acceleration / deceleration, turning, and lowering, segment the grasping timing signal according to the action nodes and label the event type to obtain the event window; The cumulative displacement, clamping force change rate, and current fluctuation of the event window are obtained and input into the slip discrimination network to obtain slip risk parameters associated with the event type. The slip risk parameters are compared with the preset risk threshold to determine the risk level and obtain the matching target clamping force, and the maximum clamping force constraint is applied to the target clamping force. The clamping force error is calculated based on the target clamping force and the clamping force feedback, and a clamping drive command is generated based on the clamping force error to make the clamping force feedback track the target clamping force. When the target clamping force reaches the maximum clamping force and the risk level of the slippage risk parameter meets the second risk threshold, an abnormal handling instruction is generated; otherwise, a transport instruction is generated. The abnormal handling instruction or transport instruction is integrated with the clamping drive instruction to form an anti-slip gripping control instruction and applied to the robotic arm.

2. The anti-slip gripping method for a handling robot with dynamically adjustable clamping force as described in claim 1, characterized in that, The grasping timing signal includes: gripper displacement and gripping force reverse drive current at the same sampling time.

3. The anti-slip gripping method for a handling robot with dynamically adjustable clamping force as described in claim 1, characterized in that, The analysis of the robotic arm's motion commands extracts action nodes such as closing, lifting, acceleration / deceleration, turning, and lowering, including: The robotic arm motion commands are analyzed to obtain the gripper opening and closing settings, end effector pose settings, and speed settings arranged over time. The start and end times of the closing action node are extracted based on the change from opening to closing of the gripper opening and closing settings. The start and end times of the lifting action node are extracted based on the rise of the vertical position component of the end effector. The start and end times of the acceleration and deceleration action nodes are extracted based on the change of acceleration settings. The start and end times of the turning action node are extracted based on the change of end effector direction. The start and end times of the lowering action node are extracted based on the change from closing to opening of the gripper opening and closing settings.

4. The anti-slip gripping method for a handling robot with dynamically adjustable clamping force as described in claim 1, characterized in that, The obtained slippage risk parameters associated with the event type include: Receive the event window and event type, and extract the gripper displacement subsequence, gripping force feedback subsequence, and drive current subsequence according to the sampling time; Calculate the cumulative displacement sequence, the clamping force change rate sequence, and the current fluctuation sequence based on the subsequences; The cumulative displacement, the rate of change of clamping force, and the current fluctuation are aligned at the same sampling time to construct a three-dimensional feature vector, and then stacked in chronological order to form a three-channel input sequence. The three-channel input sequence is fed into the ModernTCN slip discrimination network to obtain slip risk parameters associated with the event type.

5. The anti-slip gripping method for a handling robot with dynamically adjustable clamping force as described in claim 4, characterized in that, The ModernTCN glide discriminant network consists of an ingress causal convolutional layer, a multi-layer dilated causal convolutional residual block, a temporal convergence part, a trend extraction part, and output branches set according to event type.

6. The anti-slip gripping method for a handling robot with dynamically adjustable clamping force as described in claim 4, characterized in that, The process of inputting the three-channel input sequence into the ModernTCN slip discrimination network to obtain slip risk parameters associated with the event type includes: The three-channel input sequence is fed into the ingress causal convolutional layer, and an intermediate feature map of invariant length is obtained through causal constraints and zero-padding at the front end. The intermediate feature maps are sequentially input into multi-layer dilated causal convolutional residual blocks, and the input feature maps are fused through residual connections to obtain the terminal feature map; Channel averaging is performed on the terminal feature map over the time dimension to obtain the window representation vector; The event window is divided into a front segment and a back segment according to the time sequence. Channel averaging is performed on the feature maps of the front segment and the back segment respectively to obtain the front segment convergence vector and the back segment convergence vector. The trend representation vector is calculated by normalized difference. The window representation vector and the trend representation vector are concatenated, and the corresponding output branch is selected according to the event type. The two-dimensional slip risk parameters are output through a fully connected layer and the range is limited to obtain the risk level and risk trend.

7. The anti-slip gripping method for a handling robot with dynamically adjustable clamping force as described in claim 1, characterized in that, The step of comparing the slip risk parameter with a preset risk threshold to determine the risk level and obtain a matching target clamping force includes: Obtain the baseline clamping force, maximum clamping force, first risk threshold, and second risk threshold. Classify and determine the risk level based on the slip risk parameter and the first and second risk thresholds. When the risk level meets the first risk threshold, determine the first clamping increment based on the event type and add it to the baseline clamping force to obtain the target clamping force. When the risk level meets the second risk threshold or the risk trend is rising, determine the second clamping increment based on the event type and add it to the baseline clamping force to obtain the target clamping force.

8. The anti-slip gripping method for a handling robot with dynamically adjustable clamping force as described in claim 7, characterized in that, The first clamping amount includes a clamping amount of closing event type, a clamping amount of lifting event type, a clamping amount of acceleration / deceleration event type, a clamping amount of turning event type, and a clamping amount of lowering event type; The second clamping amount includes the second clamping amount for closing event type, the second clamping amount for lifting event type, the second clamping amount for acceleration / deceleration event type, the second clamping amount for turning event type, and the second clamping amount for lowering event type, and the second clamping amount for each event type is greater than the first clamping amount for the corresponding event type.

9. The anti-slip gripping method for a handling robot with dynamically adjustable clamping force as described in claim 1, characterized in that, The step of calculating the clamping force error based on the target clamping force and the clamping force feedback, and generating a clamping drive command based on the clamping force error to make the clamping force feedback track the target clamping force, includes: At each sampling moment, the target clamping force and clamping force feedback are read, and the clamping force error is calculated. The clamping force error is input into the error closed-loop unit, and the clamping adjustment amount is obtained through proportional, integral, and differential operations. The clamping adjustment amount is converted into a drive current command, and the amplitude and rate of change constraints are applied to the drive current command to obtain the clamping drive command. The clamping drive command is sent to the clamping actuator to adjust the clamping force of the gripper. New clamping force feedback is obtained at subsequent sampling times. The above steps are repeated to make the clamping force feedback track the target clamping force.

10. A dynamic-adjustable clamping force anti-slip gripping system for a handling robot, characterized in that, include: The signal acquisition module synchronously acquires the gripping force feedback, gripper displacement and drive current during the gripper closing and transport process at each sampling moment to obtain the gripping timing signal. The event window generation module parses the robotic arm's motion commands, extracts action nodes such as closing, lifting, acceleration and deceleration, turning, and lowering, segments the grasping timing signal according to the action nodes and labels the event type to obtain the event window; The slip risk discrimination module obtains the cumulative displacement, clamping force change rate, and current fluctuation of the event window, inputs them into the slip discrimination network, and obtains slip risk parameters associated with the event type. The hierarchical decision-making module compares the slip risk parameters with the preset risk threshold to determine the risk level and obtain the matching target clamping force, and applies the maximum clamping force constraint to the target clamping force. The closed-loop control module calculates the clamping force error based on the target clamping force and the clamping force feedback, and generates clamping drive commands based on the clamping force error, so that the clamping force feedback tracks the target clamping force. The instruction fusion module generates an abnormal handling instruction when the target clamping force reaches the maximum clamping force and the risk level of the slippage risk parameter meets the second risk threshold; otherwise, it generates a transport instruction. The abnormal handling instruction or transport instruction is fused with the clamping drive instruction to form an anti-slip gripping control instruction and applied to the robotic arm.

Images