A self-adaptive control method for a multi-working-condition car coupler uncoupling robot

Abstract

The present application belongs to the technical field of car coupler uncoupling, and in particular to a car coupler uncoupling robot operation adaptive control method under multiple working conditions, which comprises obtaining a car coupler uncoupling original data set in time dimension and action stage dimension and performing standardization processing; performing multi-modal analysis on the standardized data set, constructing an execution parameter feature set of trajectory response parameters, force feedback parameters and visual recognition parameters; integrating and aligning the execution parameter feature set based on time and action stage dimension, generating an execution action feature set; comparing the execution action feature set with a preset ideal strategy parameter set to obtain a simulation deviation data set, and generating a compensation strategy based on deviation degree and deviation direction data; adjusting the execution action feature set through parameter compensation to form an adjusted action parameter set; generating a driving instruction using the adjusted action parameter set, driving the robot to perform the uncoupling action according to the updated trajectory, force control and visual following strategy, and realizing adaptive adjustment and control precision improvement under multiple working conditions.

Description

Technical Field

[0001] This invention relates to the field of coupler uncoupling technology, specifically an adaptive control method for coupler uncoupling robots operating under multiple working conditions. Background Technology

[0002] Couplers are key mechanical devices used in railway vehicles or rail transit equipment to connect trains. Their core function is to complete the mechanical coupling and uncoupling operations between vehicles through the precise cooperation of components such as the coupler head, coupler tongue, and coupler lock iron. When coupling, the coupler tongue rotates and embeds into the other coupler head to form a rigid connection, while the buffer device absorbs the impact force during train operation. When uncoupling, the coupler tongue is returned to its original position by operating the handle or external force source to separate the vehicles.

[0003] In existing technologies, by integrating visual recognition, force control feedback, and model reference adaptive algorithms, sensors are used to collect data on coupler model, position, and connection status in real time. Combined with dynamic environmental perception technology, the force, speed, and trajectory of the uncoupling action are automatically adjusted to ensure that the actuator parameters can be dynamically corrected through adaptive control laws under different vehicle models, motion states, and working conditions, thereby achieving precise uncoupling and ensuring operational stability.

[0004] The above-mentioned solutions still have some problems in practical applications. Although the existing technology can complete the coupling and uncoupling operation, due to the influence of various working conditions such as train speed, connection angle, track impact vibration, and ambient temperature and humidity, the robot is prone to response delay, force control imbalance, or trajectory deviation when performing the coupling and uncoupling action. These deviations will cause the coupler to fail to connect accurately or the action to be incomplete. When the robot works continuously, these action deviations will accumulate, reducing the success rate of coupling and uncoupling and reducing the consistency of actions, thereby affecting the smooth progress of subsequent train marshalling or demarcation operations, and causing operation delays or interruptions. Due to the instability of the coupling and uncoupling action, the subsequent train marshalling or demarcation operations will be directly affected, resulting in disordered operation sequence, action delays or interruptions, thereby reducing the efficiency and reliability of the entire marshalling operation.

[0005] Therefore, the present invention provides an adaptive control method for coupler / uncoupling robot operation under multiple working conditions. Summary of the Invention

[0006] This application provides an adaptive control method for coupler / uncoupling robot operations under multiple working conditions, enabling the device to maintain the stability, consistency, and accuracy of uncoupling actions under complex working conditions.

[0007] To achieve the above objectives, this application adopts the following technical solution:

[0008] This application provides an adaptive control method for a coupler / uncoupling robot operating under multiple working conditions. The method includes:

[0009] Obtain the original dataset of coupler uncoupling that includes time and action stage dimensions, and perform standardization processing on the original dataset to obtain a standardized dataset;

[0010] Multimodal analysis is performed on the standardized dataset to construct an execution parameter feature set, which includes trajectory response parameters, force feedback parameters, and visual recognition parameters.

[0011] Based on the time dimension and the action stage dimension, the trajectory response parameters, the force feedback parameters and the visual recognition parameters in the execution parameter feature set are integrated and aligned to generate an execution action feature set.

[0012] The execution action feature set is compared with the preset ideal strategy parameter set item by item. The simulation deviation dataset is obtained by calculating the quantitative difference between the trajectory response parameter, the force feedback parameter and the visual recognition parameter and the corresponding ideal value.

[0013] The simulation deviation dataset is evaluated by calculating the quantitative index of the overall deviation to obtain deviation degree data, and by identifying the feature dimensions that cause the main deviation to obtain deviation direction data.

[0014] Based on the deviation degree data and the deviation direction data, a compensation strategy for trajectory response parameters, force feedback parameters or visual recognition parameters is generated, and the adjusted action parameter set is obtained by calculating the action parameter set corresponding to the execution action feature set generated by the compensation strategy.

[0015] The adjusted motion parameter set is used to generate drive instructions, which are used to drive the robot to perform the unhooking action according to the updated trajectory, force control and vision following strategy.

[0016] In some possible implementations, the process of integrating and aligning the trajectory response parameters, force feedback parameters, and visual recognition parameters in the execution parameter feature set based on the time dimension and the action stage dimension to generate an execution action feature set includes:

[0017] Establish a unified timeline synchronized with the robot control system clock;

[0018] Based on the time series of the trajectory response parameters, the force feedback parameters and the visual recognition parameters are time-stamp matched and the sampling rate is unified to generate a spatiotemporally synchronized multimodal data sequence.

[0019] Based on predefined sequential action phases, phase identifiers are assigned to the spatiotemporally synchronized multimodal data sequence. The sequential action phases include a positioning and approach phase, a hook-and-tongue probing phase, a force-controlled unlocking phase, and a separation and withdrawal phase.

[0020] Based on the stage identifier, the spatiotemporally synchronized multimodal data sequence is divided into data subsets corresponding to each stage, and features are extracted and combined from the data subsets of each stage to generate an execution action feature set.

[0021] In some possible implementations, the step of extracting and combining features from subsets of data at each stage to generate an action feature set includes:

[0022] The displacement rate of change, velocity rate of change, and acceleration features of the trajectory response parameters are extracted from the data subsets of each stage to obtain the trajectory response parameter features;

[0023] Extract the torque variation characteristics and contact force variation characteristics of the force feedback parameters from the data subsets of each stage to obtain the force feedback parameter characteristics;

[0024] Extract the key point position change features and relative pose features of the visual recognition parameters from the data subsets of each stage to obtain the visual recognition parameter features;

[0025] The trajectory response parameter features, the force feedback parameter features, and the visual recognition parameter features are combined within a stage to form a stage feature vector that comprehensively represents the action execution characteristics of that stage.

[0026] The stage feature vectors of each stage are integrated sequentially to generate the execution action feature set.

[0027] In some possible implementations, the step of comparing the execution action feature set with a preset ideal strategy parameter set item by item, and obtaining a simulation deviation dataset by calculating the quantitative differences between the trajectory response parameters, the force feedback parameters, and the visual recognition parameters and their corresponding ideal values, includes:

[0028] The trajectory response parameters, force feedback parameters, and visual recognition parameters in the execution action feature set are indexed and sorted according to the time dimension and the action stage dimension.

[0029] For the trajectory response parameters, calculate the absolute difference and relative difference with the corresponding trajectory response parameters in the ideal strategy parameter set in each action stage to generate a subset of trajectory response parameter deviations;

[0030] For the force feedback parameters, calculate the absolute and relative differences between the force feedback parameters and the stress feedback parameters in the ideal strategy parameter set during each action stage to generate a force feedback parameter deviation subset;

[0031] For the visual recognition parameters, calculate the absolute position difference and pose difference with the corresponding visual recognition parameters in the ideal strategy parameter set during each action stage to generate a visual recognition parameter deviation subset;

[0032] The trajectory response parameter deviation subset, force feedback parameter deviation subset, and visual recognition parameter deviation subset are integrated according to the time dimension and the action stage dimension to generate a simulation deviation dataset.

[0033] In some possible implementations, evaluating the simulation deviation dataset by calculating a quantitative index of the overall deviation to obtain deviation data includes:

[0034] Based on the trajectory response parameter deviation subset, the force feedback parameter deviation subset, and the visual recognition parameter deviation subset, the deviation distribution characteristics of each parameter under the time dimension and the action stage dimension are calculated respectively to obtain the local deviation index of each parameter dimension.

[0035] The local deviation index is normalized and then weighted and aggregated according to the time dimension and the action stage dimension to generate an overall deviation evaluation matrix.

[0036] Based on the overall deviation evaluation matrix, a composite quantitative value of the deviation is calculated to characterize the overall execution deviation of the robot in the multimodal execution parameter space, thereby obtaining deviation data.

[0037] In some possible implementations, the method of obtaining deviation direction data by identifying the feature dimensions that cause the main deviation includes:

[0038] Based on the simulation deviation dataset and its corresponding local deviation index, the parameter dimensions with significant deviations in the time dimension and the action stage dimension are identified. The parameter dimensions include specific components or key features of trajectory response parameters, force feedback parameters and visual recognition parameters.

[0039] For the parameter dimensions marked as having significant deviations in each action phase, analyze their deviation directions. The deviation directions include the displacement, velocity, and acceleration deviations of the trajectory response parameters, the torque and contact force deviations of the force feedback parameters, and the key point position and pose deviations of the visual recognition parameters.

[0040] The parameters with significant deviations obtained from the analysis and their corresponding deviation directions are integrated to generate deviation direction data.

[0041] In some possible implementations, generating a compensation strategy for the trajectory response parameters, force feedback parameters, or visual recognition parameters based on the deviation degree data and the deviation direction data includes:

[0042] The compensation strategy is generated based on a predefined sequence of action phases, which includes a positioning and approach phase, a hook tongue probing phase, a force-controlled unlocking phase, and a separation and withdrawal phase.

[0043] Using the compensation function model, at least two of the trajectory response parameters, force feedback parameters, and visual recognition parameters are jointly adjusted based on the deviation direction data to generate the compensation parameter set.

[0044] In some possible implementations, the step of generating the action parameter set corresponding to the action feature set by the compensation strategy and performing calculations to obtain the adjusted action parameter set includes:

[0045] The compensation strategy is mapped step by step to the feature vectors of each stage of the execution action feature set according to the time dimension and the action stage dimension, so that each stage feature vector corresponds to the compensation parameters of trajectory response parameters, force feedback parameters and visual recognition parameters.

[0046] The trajectory response parameters, force feedback parameters, and visual recognition parameters in the feature vectors of each stage of the execution action feature set are numerically corrected based on the mapping compensation parameters.

[0047] The modified subsets of motion parameters for each stage are integrated according to the sequence of motion stages to generate a complete sequence of motion parameters with cross-stage continuity and temporal consistency, which serves as the adjusted set of motion parameters.

[0048] In some possible implementations, the generation of drive instructions using the adjusted motion parameter set, the drive instructions being used to drive the robot to perform an unhooking action according to the updated trajectory, force control, and vision following strategy, includes:

[0049] The adjusted set of motion parameters is used to generate drive commands, which drive the robot to perform the unhooking action; during this process, the robot's actual execution status data is acquired in real time.

[0050] The actual execution status data is compared with the expected state of the adjusted action parameter set to generate a real-time deviation;

[0051] The drive command sequence is verified and corrected in real time based on the real-time deviation to ensure that the robot accurately executes the unhooking action according to the updated trajectory, force control and vision following strategy.

[0052] The beneficial effects of this invention are as follows:

[0053] 1. The adaptive control method for a coupler / uncoupling robot under multiple working conditions described in this invention, based on an execution action feature set constructed in the time dimension and action stage dimension, and a joint compensation strategy for multimodal parameters, achieves coordinated adjustment of trajectory response parameters, force feedback parameters, and visual recognition parameters. During the robot's uncoupling action, this scheme can dynamically generate a cross-stage compensation parameter set, precisely controlling the values ​​and directions of multimodal execution parameters, enabling the robot to maintain continuity and stability in each action stage, such as positioning, contact, unlocking, and withdrawal, thereby significantly improving the control accuracy and action consistency of the robot's operation.

[0054] 2. The adaptive control method for coupler / uncoupling robot operation under multiple working conditions described in this invention achieves real-time feedback and compensation for the robot's actual execution state through a real-time adjustment mechanism driven by deviation degree data and deviation direction data. During the execution of the action, the system can generate correction drive commands based on real-time deviations and jointly optimize the trajectory, force control, and vision following strategies, enabling the robot to adaptively cope with changes in multiple working conditions, ensuring a high success rate and continuous stability of the uncoupling action, and improving the overall intelligence level and operational reliability of the operation. Attached Figure Description

[0055] The invention will now be further described with reference to the accompanying drawings.

[0056] Figure 1 This is a flowchart of an adaptive control method for a coupler / uncoupler robot operating under multiple working conditions according to the present invention. Detailed Implementation

[0057] The terms "first," "second," and "third," etc., used in this application specification, claims, and drawings are used to distinguish different objects, not to limit a specific order.

[0058] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0059] Research has revealed that while existing technologies can complete the uncoupling and coupling operations, the robot is prone to response delays, force control imbalances, or trajectory deviations when performing uncoupling actions due to variations in train speed, connection angle, track impact vibration, and ambient temperature and humidity. These deviations can lead to inaccurate coupling or incomplete action execution. When the robot operates continuously, these deviations accumulate, reducing the success rate of uncoupling and decreasing action consistency. This affects the smooth progress of subsequent train marshalling or demarcation operations, causing delays or interruptions. The instability of the uncoupling action directly impacts subsequent train marshalling or demarcation operations, resulting in disordered operation sequences, action delays, or interruptions, thus reducing the efficiency and reliability of the entire marshalling operation.

[0060] Example 1

[0061] To solve the above problems, such as Figure 1 As shown, this application provides an adaptive control method for a coupler / uncoupling robot operating under multiple working conditions, the method comprising:

[0062] Step 1: Obtain the original dataset of coupler uncoupling that includes the time dimension and the action stage dimension, and perform standardization processing on the original dataset to obtain a standardized dataset;

[0063] In step one: Time dimension: used to identify the sequence of consecutive time points in the process of the robot performing the unhooking operation, in order to characterize the temporal order of the actions;

[0064] Action Phase Dimension: A dimension identifier that divides the hook removal operation process into several sequential phases to distinguish the operation of different functional phases, including the positioning and approach phase, hook tongue probing phase, force control unlocking phase, and separation and withdrawal phase.

[0065] Original dataset for coupler uncoupling: A multimodal dataset containing trajectory response data, force feedback data, and visual recognition data collected during the robot's uncoupling operation;

[0066] Standardized dataset: A dataset with a uniform numerical scale and time order, formed by normalizing and time-aligning the original dataset of coupler uncoupling.

[0067] For example: When acquiring the original dataset of coupler uncoupling through the multi-channel synchronous acquisition mechanism of the robot control system, the control system first uses the global clock signal as a unified time reference, and simultaneously triggers the trajectory tracking module, force sensing module and vision recognition module to acquire data. At each time point, the displacement, velocity and acceleration signals of the end effector, as well as the contact force and torque information of the clamping end are recorded synchronously. At the same time, the three-dimensional spatial position and attitude parameters of the coupler key points are extracted. All data are accompanied by a unified timestamp and numbered according to the action stage. After the acquisition is completed, the trajectory response data, force feedback data and vision recognition data are normalized respectively. Data with different sampling frequencies are resampled in time, and the action stage boundaries are corrected, thereby forming a standardized dataset that is unified and aligned in the time dimension and action stage dimension and can be directly used for subsequent multimodal analysis.

[0068] It should be noted that by obtaining the original coupler and uncoupler dataset containing the time dimension and action stage dimension in step one and performing standardization processing, the uniformity of data from different sensor sources in terms of time and stage can be achieved, which significantly improves the accuracy and stability of subsequent multimodal parameter feature analysis and execution deviation identification, and provides highly consistent data support for the generation of adaptive control strategies.

[0069] Step 2: Perform multimodal analysis on the standardized dataset to construct an execution parameter feature set, which includes trajectory response parameters, force feedback parameters, and visual recognition parameters;

[0070] In step two: multimodal analysis: joint processing and feature extraction of data from different sensor channels to characterize the robot's overall execution state at the same action stage and time point;

[0071] Execution parameter feature set: A set of features extracted by performing multimodal analysis on a standardized dataset to characterize the robot’s behavioral characteristics during different stages of action execution, including trajectory response parameters, force feedback parameters, and visual recognition parameters;

[0072] Trajectory response parameters: Characteristic information used to represent the spatial motion state of the end effector of the robot during the unhooking action, including changes in displacement, velocity, and acceleration;

[0073] Force feedback parameters: Characteristic information used to characterize the force situation of the robot during the unhooking process, including joint torque and end contact force changes;

[0074] Visual recognition parameters: Information used to characterize the robot vision system's recognition of key points and relative pose of the coupler, including key point position changes and posture deviations;

[0075] For example: By performing multimodal analysis on a standardized dataset, firstly, the displacement, velocity, and acceleration change rates of the trajectory response data are calculated according to the time dimension to obtain the motion characteristic curves of the end effector at each time point. At the same time, the average change rate and extreme value changes within each action stage are extracted to reflect the dynamic response at different stages. Next, the incremental calculation of joint torque and end contact force is performed on the force feedback data to analyze the force change law at each stage, and the force signal within a continuous time period is smoothed and normalized. Subsequently, the displacement and relative pose changes of the key points of the coupler in three-dimensional space are analyzed on the visual recognition data to extract key frame features and stage feature values. Finally, the trajectory response parameters, force feedback parameters, and visual recognition parameters are aligned and integrated in the time dimension and action stage dimension to form a multimodal execution parameter feature set that reflects the overall execution characteristics of the robot in each action stage.

[0076] It should be noted that by constructing the execution parameter feature set in step two, multimodal data from different sensor channels can be fused into a unified feature representation, accurately characterizing the robot's execution behavior at each stage of the unhooking action, providing basic data support for subsequent deviation analysis and adaptive control strategy generation, thereby improving control accuracy and response stability.

[0077] Step 3: Based on the time dimension and the action stage dimension, integrate and align the trajectory response parameters, force feedback parameters and visual recognition parameters in the execution parameter feature set to generate an execution action feature set;

[0078] In step three: the execution action feature set refers to the multimodal feature set formed after aligning and integrating trajectory response parameters, force feedback parameters, and visual recognition parameters in the time dimension and action stage dimension;

[0079] For example: By processing the execution parameter feature set, firstly, the trajectory response parameters, force feedback parameters, and visual recognition parameters are aligned according to a unified timestamp in the time dimension to ensure that each parameter corresponds to the same action state at the same time point. Then, based on the predefined sequential action stages, the time series is divided into the positioning and approach stage, the hook tongue probing stage, the force control unlocking stage, and the separation and withdrawal stage. Various parameters in each stage are aggregated and processed to extract the average value, rate of change, and extreme value features within the stage. At the same time, different parameter components are combined to form a stage feature vector. Finally, the stage feature vectors of each stage are integrated sequentially according to the action stage order to generate a complete execution action feature set.

[0080] It should be noted that by integrating and aligning the execution parameter feature set in terms of time and stage in step three, the correspondence between multimodal parameters at the same action stage and time point can be clearly ensured, thereby providing accurate and directly usable multimodal action feature data for subsequent deviation calculation, compensation strategy generation and adaptive control.

[0081] Step 4: Compare the execution action feature set with the preset ideal strategy parameter set item by item, and obtain the simulation deviation dataset by calculating the quantitative difference between the trajectory response parameters, the force feedback parameters and the visual recognition parameters and the corresponding ideal values;

[0082] In step four: Simulation deviation dataset: refers to the set of deviations obtained by quantitative calculation after comparing various parameters in the action feature set with the corresponding ideal policy parameters. It is used to characterize the degree and direction of deviation of each parameter from the ideal value during the multimodal execution of the robot.

[0083] Preset ideal strategy parameter set: refers to a predefined set of reference parameters used to guide the robot's hook removal action, including the expected values ​​of trajectory response, force feedback and visual recognition;

[0084] For example: By comparing the set of execution action features with the set of ideal strategy parameters item by item, the trajectory response parameters, force feedback parameters and visual recognition parameters are first indexed and sorted according to the time dimension and action stage dimension to ensure that the parameters at each time point and action stage can correspond to the ideal strategy parameters. Then, the absolute difference and relative difference of displacement, velocity and acceleration of trajectory response parameters in each action stage are calculated to generate a subset of trajectory response parameter deviations. The absolute deviation and trend deviation of joint torque and end contact force are calculated for force feedback parameters to generate a subset of force feedback parameter deviations. The spatial position deviation and pose deviation of key points are calculated for visual recognition parameters to generate a subset of visual recognition parameter deviations.

[0085] Finally, the trajectory response parameter deviation subset, force feedback parameter deviation subset, and visual recognition parameter deviation subset are integrated according to the time dimension and action stage dimension to form a complete simulation deviation dataset.

[0086] It should be noted that by generating the simulation deviation dataset in step four, the deviation of the robot in the multimodal execution parameter space can be fully quantified, providing an accurate data foundation for subsequent deviation assessment, identification of major deviation features, and generation of adaptive compensation strategies, thereby improving the accuracy and reliability of the unhooking action.

[0087] Step 5: Evaluate the simulation deviation dataset by calculating the quantitative index of the overall deviation to obtain the deviation degree data;

[0088] In step five: Deviation data: refers to the numerical value obtained by performing an overall quantitative analysis of the simulation deviation dataset, which is used to characterize the robot's overall execution deviation from the ideal strategy in the multimodal execution parameter space;

[0089] Overall deviation quantification index: refers to the numerical value formed by integrating deviation information from various parameter dimensions, used to quantify the degree of deviation of the robot's overall actions;

[0090] For example: By processing the simulation deviation dataset, firstly, the local deviation indices of the trajectory response parameter deviation subset, force feedback parameter deviation subset, and visual recognition parameter deviation subset are calculated in the time dimension and action stage dimension, including average deviation, maximum deviation, and rate of change, etc. Then, the local deviation indices of each parameter are normalized to eliminate differences in different dimensions and numerical ranges. Then, they are weighted and aggregated according to the time dimension and action stage dimension to form an overall deviation evaluation matrix, which reflects the comprehensive deviation of the robot at each time point and action stage. Finally, the deviation synthesis quantification value is calculated based on the overall deviation evaluation matrix, and the deviation degree of trajectory response, force feedback, and visual recognition parameters is integrated into a single index to generate complete deviation degree data.

[0091] It should be noted that by generating deviation data in step five, the overall deviation of the robot in the multimodal execution parameter space can be accurately characterized, providing a reliable data foundation for subsequent key deviation dimension identification, compensation strategy generation, and adaptive control, thereby improving the execution accuracy and control stability of the unhooking action.

[0092] Step Six: Obtain deviation direction data by identifying the feature dimensions that cause the main deviation;

[0093] In step six: Deviation direction data: refers to the data set formed by analyzing the simulation deviation dataset and the local deviation indices of each parameter to identify the parameter dimension that contributes the most to the overall deviation and its deviation direction, which is used to guide the generation of compensation strategies;

[0094] Feature dimension: refers to the components or key features with clear physical meaning in trajectory response parameters, force feedback parameters and visual recognition parameters, such as displacement, velocity, acceleration, torque, contact force or key point position and pose bias;

[0095] Deviation direction: refers to the deviation trend of the feature dimension at each action stage and time point, including positive or negative changes, used to characterize the directionality of the parameter deviating from the ideal value;

[0096] For example: By analyzing the simulation deviation dataset and corresponding local deviation indices, we first screen out the feature dimensions with significant deviations in the time and action phase dimensions, and mark the trajectory response parameters, force feedback parameters, and visual recognition parameters that contribute the most to the overall deviation. Then, we calculate the deviation direction for each significant feature dimension, including the displacement, velocity, and acceleration deviation of the trajectory response parameters, the torque and contact force deviation of the force feedback parameters, and the key point position and pose deviation of the visual recognition parameters. Finally, we integrate the feature dimensions with significant deviations and their corresponding deviation directions to form a deviation direction dataset, which provides a basis for the generation of subsequent compensation strategies.

[0097] It should be noted that by generating deviation direction data in step six, it is possible to identify which parameters and their components cause the main execution deviations in each action phase of the robot, and to understand their deviation directions. This provides data support for the accurate generation of compensation strategies, thereby effectively improving the control accuracy and response reliability of the unhooking action.

[0098] Step 7: Based on the deviation degree data and the deviation direction data, generate a compensation strategy for the trajectory response parameters, force feedback parameters, or visual recognition parameters;

[0099] In step seven: Compensation strategy: A parameter correction scheme determined based on deviation degree data and deviation direction data is used to jointly adjust at least two execution parameters among trajectory response parameters, force feedback parameters and visual recognition parameters based on a predefined sequence of action stages, so as to reduce deviations during robot execution and improve control accuracy;

[0100] Parameter compensation model: A joint adjustment model established based on deviation degree data and deviation direction data, which can reflect the coupling relationship between multiple parameters under different action stages, and is used to determine the adjustment amount and adjustment direction of each execution parameter;

[0101] Adjustment amount: The specific correction value calculated for each execution parameter, used to reflect the magnitude of the adjustment required;

[0102] Adjustment direction: The positive or negative trend of the parameter correction is used to guide the parameter to converge toward the ideal value;

[0103] For example: After the feature set of the executed action is evaluated for deviation, the system first determines the severity level of the overall deviation based on the deviation degree data, and identifies the key execution parameter dimensions that cause the main deviation in the current action stage based on the deviation direction data. Subsequently, for different action stages, the parameter compensation model calls the joint compensation logic that matches the current stage to coordinately adjust the trajectory response parameters, force feedback parameters, and visual recognition parameters. In the positioning and approach stage, when the deviation direction data indicates that the deviation is mainly caused by the lateral position deviation of the coupler key point in the visual recognition parameters and the end effector speed deviation in the trajectory response parameters, the compensation strategy adjusts the visual recognition parameters and trajectory response parameters simultaneously: calculating the lateral correction amount based on the visual position deviation, and adjusting the approach speed based on the speed deviation, thereby generating a compensation parameter set that synchronously corrects the spatial position and motion speed, enabling the robot to smoothly adjust the approach dynamics while correcting its pose.

[0104] During the hook-and-tongue probing phase, when the deviation direction data simultaneously identifies insufficient contact force in the force feedback parameters and deviation in the probing angle in the trajectory response parameters, the compensation strategy jointly adjusts the force feedback parameters and trajectory response parameters. It calculates the required increase in force control output based on the contact force deviation and fine-tunes the attitude angle based on the angle deviation. The adjustment amounts of both are coupled and output in the parameter compensation model, thereby ensuring that the robot probing at a better angle while increasing the contact force, avoiding jamming caused by incorrect posture.

[0105] During the force-controlled unlocking phase, when the deviation direction data shows a persistently high unlocking torque in the force feedback parameters and a slight offset in the relative pose of the hook tongue in the visual recognition parameters, the compensation strategy jointly adjusts the force feedback parameters and visual recognition parameters. It dynamically corrects the force control center reference position based on the visual pose deviation and finely adjusts the force control output stiffness based on the torque deviation. This achieves coordinated control through "visual correction assisting force control positioning and force perception feeding back to visual tracking," effectively preventing "hard prying" or "slippage" during unlocking. Through multi-parameter joint analysis at each stage, the parameter compensation model ultimately outputs a compensation parameter set containing the adjustment amounts of each execution parameter and their collaborative relationships, forming a complete compensation strategy that achieves phased and coordinated correction of trajectory response parameters, force feedback parameters, and visual recognition parameters.

[0106] It should be noted that by generating a compensation strategy in step seven, multimodal execution parameters can be collaboratively adjusted based on deviation degree data and deviation direction data, taking the action phase as the context. This method overcomes the limitations of traditional single-parameter independent compensation and can effectively handle the dynamic coupling relationship between trajectory response parameters, force feedback parameters, and visual recognition parameters under multiple working conditions, thereby significantly improving the robot's motion intelligence, adaptability, and overall control accuracy during coupler uncoupling operations.

[0107] Step 8: Calculate the action parameter set corresponding to the action feature set generated by the compensation strategy to obtain the adjusted action parameter set;

[0108] In step eight: Action parameter set: A set of parameters obtained by mapping the action feature set and that can be directly used to generate robot control commands. The parameter set includes trajectory control parameters, force control adjustment parameters, and vision correction parameters.

[0109] Adjusted motion parameter set: A new set of parameters generated after correcting the parameter values ​​in the original motion parameter set under the action of the compensation strategy, used to characterize the control command characteristics after compensation;

[0110] Parameter calculation module: A calculation unit used to combine the compensation strategy with the action parameter set, and outputs the adjusted action parameter set through calculation methods such as parameter superposition, weighting or mapping transformation;

[0111] For example: After generating the compensation strategy, the parameter calculation module first reads the action parameter set corresponding to the action feature set, including trajectory control parameters, force control adjustment parameters, and visual correction parameters. Then, the adjustment amounts of each parameter in the compensation strategy are matched with the corresponding action parameters, and superimposed and corrected according to preset calculation rules: For trajectory control parameters, the displacement, velocity, and acceleration setpoints of the end effector are corrected based on the adjustment amount of the trajectory response parameters in the compensation strategy; for force control adjustment parameters, the control thresholds of joint torque and end contact force are dynamically corrected based on the adjustment amount of the force feedback parameters in the compensation strategy; for visual correction parameters, the spatial position and attitude angle information of the coupler key points are corrected based on the adjustment amount of the visual recognition parameters in the compensation strategy.

[0112] After completing the calculation of each parameter, the parameter calculation module recombines the corrected parameters to form a new set of motion parameters, and verifies the time synchronization and correspondence between various parameters and the motion stage to ensure that the adjusted set of motion parameters can be directly applied in the control system, thereby realizing the adaptive control optimization of the robot in each motion stage.

[0113] It should be noted that by performing calculations on the compensation strategy and the action parameter set corresponding to the action feature set in step eight, the control parameters can be accurately corrected, thereby obtaining the adjusted action parameter set. This enables the robot to have self-optimization capabilities when performing the unhooking operation, effectively reducing action errors and improving the overall coordination and stability of the execution.

[0114] Step 9: Generate driving instructions using the adjusted motion parameter set. The driving instructions are used to drive the robot to perform the unhooking action according to the updated trajectory, force control and vision following strategy.

[0115] In step nine: Drive commands: A sequence of robot control signals generated based on the adjusted motion parameter set, used to instruct the motion path, force control output, and visual feedback response of each joint execution unit of the robot; the drive commands may include three parts: trajectory control commands, force control adjustment commands, and visual following commands;

[0116] Trajectory control commands: Motion path control commands generated based on trajectory response parameters, used to guide the robot's end effector to move in space along the corrected trajectory curve;

[0117] Force control adjustment command: The force control output control signal generated based on the force feedback parameters is used to realize the dynamic control of the clamping force and the applied torque, so as to ensure that the force output is stable and matches the target contact conditions during the unhooking action;

[0118] Visual follow instruction: Spatial pose correction instruction generated based on visual recognition parameters, used to dynamically fine-tune the target position according to real-time image feedback during the execution of the action, so as to achieve precise positioning under visual guidance;

[0119] Updated trajectory, force control, and vision following strategy: A reconstructed execution control scheme based on the combined effect of compensation strategy and adjusted motion parameter set, designed to make robot movements more closely follow the ideal trajectory and have real-time correction capabilities;

[0120] For example: After compensation strategy correction, an adjusted set of motion parameters is obtained. Trajectory response parameters are used to correct the spatial position of path points, force feedback parameters are used to correct the amplitude and direction of force control output, and visual recognition parameters are used to optimize the matching accuracy of visual perception. The control system first generates trajectory control commands based on the trajectory response parameters to ensure the robot's end effector moves along the ideal path. Then, it generates force control adjustment commands based on the force feedback parameters to correct the output torque of the drive motor, achieving compliant control during the coupler separation phase. Simultaneously, it generates visual following commands based on the visual recognition parameters, enabling the robot to dynamically adjust its spatial posture based on real-time image data during execution. Finally, the trajectory control commands, force control adjustment commands, and visual following commands together constitute the drive commands, driving the robot to perform the unhooking action.

[0121] It should be noted that by generating drive instructions in step eight, the dynamic fusion of trajectory, force control and vision control can be achieved, enabling the robot to adaptively perform unhooking operations based on the adjusted set of motion parameters, thereby further improving the accuracy, stability and coordination of the operation, and reducing execution deviations caused by environmental disturbances or mechanical errors.

[0122] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An adaptive control method for a coupler / uncoupling robot operating under multiple working conditions, characterized in that, The method includes: Obtain the original dataset of coupler uncoupling that includes time and action stage dimensions, and perform standardization processing on the original dataset to obtain a standardized dataset; Multimodal analysis is performed on the standardized dataset to construct an execution parameter feature set, which includes trajectory response parameters, force feedback parameters, and visual recognition parameters. Based on the time dimension and the action stage dimension, the trajectory response parameters, the force feedback parameters and the visual recognition parameters in the execution parameter feature set are integrated and aligned to generate an execution action feature set. The execution action feature set is compared with the preset ideal strategy parameter set item by item. The simulation deviation dataset is obtained by calculating the quantitative difference between the trajectory response parameter, the force feedback parameter and the visual recognition parameter and the corresponding ideal value. The simulation deviation dataset is evaluated by calculating the quantitative index of the overall deviation to obtain deviation degree data, and by identifying the feature dimensions that cause the main deviation to obtain deviation direction data. Based on the deviation degree data and the deviation direction data, a compensation strategy for trajectory response parameters, force feedback parameters or visual recognition parameters is generated, and the adjusted action parameter set is obtained by calculating the action parameter set corresponding to the execution action feature set generated by the compensation strategy. The adjusted motion parameter set is used to generate drive instructions, which are used to drive the robot to perform the unhooking action according to the updated trajectory, force control and vision following strategy.

2. The adaptive control method for a coupler / uncoupling robot under multiple working conditions according to claim 1, characterized in that, The process of integrating and aligning the trajectory response parameters, force feedback parameters, and visual recognition parameters in the execution parameter feature set based on the time dimension and the action stage dimension to generate an execution action feature set includes: Establish a unified timeline synchronized with the robot control system clock; Based on the time series of the trajectory response parameters, the force feedback parameters and the visual recognition parameters are time-stamp matched and the sampling rate is unified to generate a spatiotemporally synchronized multimodal data sequence. Based on predefined sequential action phases, phase identifiers are assigned to the spatiotemporally synchronized multimodal data sequence. The sequential action phases include a positioning and approach phase, a hook-and-tongue probing phase, a force-controlled unlocking phase, and a separation and withdrawal phase. Based on the stage identifier, the spatiotemporally synchronized multimodal data sequence is divided into data subsets corresponding to each stage, and features are extracted and combined from the data subsets of each stage to generate an execution action feature set.

3. The adaptive control method for a coupler / uncoupling robot under multiple working conditions according to claim 2, characterized in that, The process of extracting and combining features from data subsets at each stage to generate an action feature set includes: The displacement rate of change, velocity rate of change, and acceleration features of the trajectory response parameters are extracted from the data subsets of each stage to obtain the trajectory response parameter features; Extract the torque variation characteristics and contact force variation characteristics of the force feedback parameters from the data subsets of each stage to obtain the force feedback parameter characteristics; Extract the key point position change features and relative pose features of the visual recognition parameters from the data subsets of each stage to obtain the visual recognition parameter features; The trajectory response parameter features, the force feedback parameter features, and the visual recognition parameter features are combined within a stage to form a stage feature vector that comprehensively represents the action execution characteristics of that stage. The stage feature vectors of each stage are integrated sequentially to generate the execution action feature set.

4. The adaptive control method for a coupler / uncoupling robot under multiple working conditions according to claim 3, characterized in that, The step involves comparing the executed action feature set with a preset ideal strategy parameter set item by item, and obtaining a simulation deviation dataset by calculating the quantitative differences between the trajectory response parameters, the force feedback parameters, and the visual recognition parameters and their corresponding ideal values. This includes: The trajectory response parameters, force feedback parameters, and visual recognition parameters in the execution action feature set are indexed and sorted according to the time dimension and the action stage dimension. For the trajectory response parameters, calculate the absolute difference and relative difference with the corresponding trajectory response parameters in the ideal strategy parameter set in each action stage to generate a subset of trajectory response parameter deviations; For the force feedback parameters, calculate the absolute and relative differences between the force feedback parameters and the stress feedback parameters in the ideal strategy parameter set during each action stage to generate a force feedback parameter deviation subset; For the visual recognition parameters, calculate the absolute position difference and pose difference with the corresponding visual recognition parameters in the ideal strategy parameter set during each action stage to generate a visual recognition parameter deviation subset; The trajectory response parameter deviation subset, force feedback parameter deviation subset, and visual recognition parameter deviation subset are integrated according to the time dimension and the action stage dimension to generate a simulation deviation dataset.

5. The adaptive control method for a coupler / uncoupling robot under multiple working conditions according to claim 4, characterized in that, The evaluation of the simulation deviation dataset, which involves calculating a quantitative index of the overall deviation to obtain deviation degree data, includes: Based on the trajectory response parameter deviation subset, the force feedback parameter deviation subset, and the visual recognition parameter deviation subset, the deviation distribution characteristics of each parameter under the time dimension and the action stage dimension are calculated respectively to obtain the local deviation index of each parameter dimension. The local deviation index is normalized and then weighted and aggregated according to the time dimension and the action stage dimension to generate an overall deviation evaluation matrix. Based on the overall deviation evaluation matrix, a composite quantitative value of the deviation is calculated to characterize the overall execution deviation of the robot in the multimodal execution parameter space, thereby obtaining deviation data.

6. The adaptive control method for a coupler / uncoupling robot under multiple working conditions according to claim 5, characterized in that, The process of obtaining deviation direction data by identifying the feature dimensions that cause the main deviation includes: Based on the simulation deviation dataset and its corresponding local deviation index, the parameter dimensions with significant deviations in the time dimension and the action stage dimension are identified. The parameter dimensions include specific components or key features of trajectory response parameters, force feedback parameters and visual recognition parameters. For the parameter dimensions marked as having significant deviations in each action phase, analyze their deviation directions. The deviation directions include the displacement, velocity, and acceleration deviations of the trajectory response parameters, the torque and contact force deviations of the force feedback parameters, and the key point position and pose deviations of the visual recognition parameters. The parameters with significant deviations obtained from the analysis and their corresponding deviation directions are integrated to generate deviation direction data.

7. The adaptive control method for a coupler / uncoupling robot under multiple working conditions according to claim 6, characterized in that, The method for generating a compensation strategy for the trajectory response parameters, force feedback parameters, or visual recognition parameters based on the deviation degree data and the deviation direction data includes: The compensation strategy is generated based on a predefined sequence of action phases, which includes a positioning and approach phase, a hook tongue probing phase, a force-controlled unlocking phase, and a separation and withdrawal phase. Using the compensation function model, at least two of the trajectory response parameters, force feedback parameters, and visual recognition parameters are jointly adjusted based on the deviation direction data to generate the compensation parameter set.

8. The adaptive control method for a coupler / uncoupling robot under multiple working conditions according to claim 7, characterized in that, The step of generating the action parameter set corresponding to the action feature set by the compensation strategy and performing calculations to obtain the adjusted action parameter set includes: The compensation strategy is mapped step by step to the feature vectors of each stage of the execution action feature set according to the time dimension and the action stage dimension, so that each stage feature vector corresponds to the compensation parameters of trajectory response parameters, force feedback parameters and visual recognition parameters. The trajectory response parameters, force feedback parameters, and visual recognition parameters in the feature vectors of each stage of the execution action feature set are numerically corrected based on the mapping compensation parameters. The modified subsets of motion parameters for each stage are integrated according to the sequence of motion stages to generate a complete sequence of motion parameters with cross-stage continuity and temporal consistency, which serves as the adjusted set of motion parameters.

9. The adaptive control method for a coupler / uncoupling robot under multiple working conditions according to claim 8, characterized in that, The process of generating drive instructions using the adjusted motion parameter set, wherein the drive instructions are used to drive the robot to perform a hook-unhooking action according to the updated trajectory, force control, and vision following strategy, includes: The adjusted set of motion parameters is used to generate drive commands, which drive the robot to perform the unhooking action; during this process, the robot's actual execution status data is acquired in real time. The actual execution status data is compared with the expected state of the adjusted action parameter set to generate a real-time deviation; The drive command sequence is verified and corrected in real time based on the real-time deviation to ensure that the robot accurately executes the unhooking action according to the updated trajectory, force control and vision following strategy.

Images