Series robot arm self-collision prevention method, prevention method and device

Abstract

The application relates to the technical field of robot motion safety protection and real-time robot motion control, and discloses a self-collision prediction method, a prevention method and equipment for a serial mechanical arm. The method comprises the following steps: acquiring an observation data set; wherein the observation data set comprises joint motion observation data of the mechanical arm in each control cycle; based on the observation data and joint instruction values of motion planning, a future multi-cycle joint angle sequence is predicted through multi-step recursion; based on the future multi-cycle joint angle sequence, the minimum link spacing is calculated from a pre-stored effective collision pairing set; according to the minimum link spacing, it is predicted whether the motion trajectory of the mechanical arm in a preset time period has a self-collision risk; a hierarchical prevention strategy is executed to avoid link self-collision in advance. The application is a self-collision prediction and prevention method for a mechanical arm, can improve the accuracy of collision prediction, and can meet the real-time safety protection scene.

Description

Technical Field

[0001] This application relates to the fields of robot motion safety protection and robot real-time motion control technology. Specifically, it relates to a self-collision prediction method, prevention method and device for a serial robotic arm. Background Technology

[0002] When the robotic arm is in actual operation, the links are prone to self-collision. At the same time, it is easily affected by external environmental interference and external force disturbance, resulting in trajectory deviation. If the collision risk cannot be predicted and avoided, it is very easy to cause wear of the robotic arm links and damage to the transmission structure.

[0003] In related technologies, the standard operating trajectory of the robotic arm is stored in advance, and the fixed movement path and speed of each joint are limited. During operation, the robotic arm is forced to follow the preset fixed trajectory. For fixed points and fixed movement ranges on the preset trajectory, fixed safety distance thresholds are set in advance. When the robotic arm is running, it is checked whether the real-time position of the robotic arm deviates from the preset fixed trajectory and fixed safety thresholds to determine whether there is a collision risk at the current point.

[0004] However, static verification methods for fixed points on a preset trajectory can only determine deviations from the preset trajectory. If the robotic arm is in certain working conditions (such as servo lag), the preset trajectory may deviate, leading to a decrease in the accuracy of collision risk prediction and thus making it impossible to accurately prevent collisions. Summary of the Invention

[0005] In view of this, this application provides a self-collision prediction method, prevention method and device for a serial robotic arm, to solve the problem that if the robotic arm is in certain working conditions (such as robotic arm servo lag), the preset trajectory deviates, resulting in a decrease in the accuracy of collision risk prediction, and thus the inability to accurately prevent collisions.

[0006] In a first aspect, this application provides a self-collision prediction method for a serial robotic arm. The method includes: acquiring an observation data set; wherein the observation data set includes joint motion observation data of the robotic arm in each control cycle; predicting the joint angle sequence of future multi-cycles in multiple steps based on the observation data set and the joint command values ​​of the motion plan; calculating the minimum link spacing from a pre-stored set of valid collision pairs based on the future multi-cycle joint angle sequence; and predicting whether there is a risk of self-collision in the motion trajectory of the robotic arm during a preset time period based on the minimum link spacing.

[0007] In one possible implementation, based on the observation dataset and the joint command values ​​from motion planning, a multi-step recursive prediction of the future multi-period joint angle sequence is performed, including: determining the joint angular velocity and joint angular acceleration corresponding to each joint based on continuous multi-frame joint motion observation data within the observation dataset; determining the temporal evolution law and state transition relationship between joint angle, angular velocity, and angular acceleration based on the joint motion observation data, joint angular velocity, and joint angular acceleration; performing state iterative updates based on the temporal evolution law, state transition relationship, and joint command values ​​to obtain the optimal joint motion state estimate; and recursively predicting the future multi-period joint angle sequence based on the optimal joint motion state estimate.

[0008] In one possible implementation, based on temporal evolution patterns, state transition relationships, and joint command values, iterative state updates are performed to obtain the optimal joint motion state estimate. This includes: deriving the predicted joint motion state estimate for the current period based on temporal evolution patterns, state transition relationships, joint command values, and the optimal joint motion state estimate from the previous period; correcting the predicted joint motion state estimate based on the joint motion observation data for the current period within the observation dataset to obtain an intermediate joint motion state estimate; repeating the steps from deriving the predicted joint motion state estimate for the current period based on temporal evolution patterns, state transition relationships, joint command values, and the optimal joint motion state estimate from the previous period to correcting the predicted joint motion state estimate based on the joint motion observation data for the current period within the observation dataset to obtain an intermediate joint motion state estimate, until the fluctuation amplitude of the intermediate joint motion state estimate is less than a preset error threshold, thus obtaining the optimal joint motion state estimate.

[0009] In one possible implementation, the observation dataset also includes external force observation data of the robotic arm under force control conditions; based on the observation dataset and the joint command values ​​of the motion planning, a multi-step recursive prediction of the future multi-cycle joint angle sequence is performed, including: determining the joint angle sequence to be processed based on continuous multi-frame joint motion observation data and joint command values ​​of the motion planning within the observation dataset; performing perturbation compensation correction on the joint angle to be processed corresponding to the current prediction cycle based on the external force observation data of the robotic arm under force control conditions to obtain the corrected joint angle; using the corrected joint angle as the input for the next prediction cycle, repeatedly performing perturbation compensation correction on the joint angle to be processed corresponding to the current prediction cycle based on the external force observation data of the robotic arm under force control conditions to obtain the corrected joint angle, until all joint angles to be processed in the joint angle sequence to be processed have completed perturbation compensation correction, thus obtaining the future multi-cycle joint angle sequence.

[0010] In one possible implementation, based on the external force observation data of the robotic arm under force control conditions, the angle of the joint to be processed corresponding to the current prediction cycle is perturbed and compensated to obtain the corrected joint angle. This includes: determining the degree of external disturbance of the robotic arm based on the external force observation data of the robotic arm under force control conditions; adjusting the uncertainty parameter based on the degree of external disturbance of the robotic arm to obtain the correction iteration parameter; wherein, the uncertainty parameter indicates the correction coefficient used to correct the amplitude of time series evolution and the amplitude of state transition; and performing perturbed compensation correction on the angle of the joint to be processed corresponding to the current prediction cycle based on the correction iteration parameter to obtain the corrected joint angle.

[0011] In one possible implementation, the minimum link spacing is calculated from a pre-stored set of valid collision pairs based on a future multi-cycle joint angle sequence. This includes: determining the world coordinate system pose of the robotic arm link at the target prediction time based on the future joint angle sequence; wherein the target prediction time is any prediction time corresponding to the future joint angle sequence; determining the minimum Euclidean spacing of the risk link pair at the target prediction time based on the link world coordinate system pose at the target prediction time and the pre-stored set of valid collision pairs; and determining the smallest of the multiple minimum Euclidean spacings as the minimum link spacing.

[0012] In one possible implementation, predicting whether the robot arm's motion trajectory has a self-collision risk during a preset time period based on the minimum link spacing includes: detecting whether the minimum link spacing is less than a first preset safety threshold; if the minimum link spacing is less than the first preset safety threshold, predicting that the robot arm's motion trajectory has a self-collision risk during the preset time period; if the minimum link spacing is not less than the first preset safety threshold, predicting that the robot arm's motion trajectory does not have a self-collision risk during the preset time period.

[0013] In one possible implementation, the method further includes: obtaining link pairing combinations; based on the parent-child relationship of the links, removing link pairs with direct parent-child connections from the link pairing combinations, and removing link pairs that will not interfere with the robot arm in any pose, to obtain a pre-stored set of valid collision pairs.

[0014] Secondly, this application provides a self-collision prevention method for a serial robotic arm. The method includes: using the aforementioned self-collision prediction method for a serial robotic arm to determine the self-collision risk assessment result of the robotic arm's motion trajectory during a preset time period; if the self-collision risk assessment result indicates that the robotic arm's motion trajectory during the preset time period has a self-collision risk, detecting whether the minimum link spacing is less than a second preset safety threshold; if the minimum link spacing is not less than the second preset safety threshold, determining a first operation of the robotic arm from a preset association relationship; if the minimum link spacing is less than the second preset safety threshold, determining a second operation of the robotic arm from a preset association relationship.

[0015] Thirdly, this application provides a self-collision prediction device for a serial robotic arm. The device includes: an acquisition module for acquiring an observation data set, wherein the observation data set includes joint motion observation data of the robotic arm in each control cycle; a first prediction module for recursively predicting a future multi-cycle joint angle sequence based on the observation data set and the joint command values ​​of the motion plan; a calculation module for calculating the minimum link spacing from a pre-stored set of valid collision pairs based on the future multi-cycle joint angle sequence; and a second prediction module for predicting whether there is a risk of self-collision in the motion trajectory of the robotic arm during a preset time period based on the minimum link spacing.

[0016] Fourthly, this application provides a computer device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the serial robotic arm self-collision prediction method of the first aspect or any corresponding embodiment and the serial robotic arm self-collision prevention method of the second aspect or any corresponding embodiment.

[0017] Fifthly, this application provides a computer-readable storage medium storing computer instructions for causing a computer to execute the serial robotic arm self-collision prediction method of the first aspect or any corresponding embodiment thereof, and the serial robotic arm self-collision prevention method of the second aspect or any corresponding embodiment thereof.

[0018] Sixthly, this application provides a computer program product, including computer instructions for causing a computer to execute the serial robotic arm self-collision prediction method of the first aspect or any corresponding embodiment thereof, and the serial robotic arm self-collision prevention method of the second aspect or any corresponding embodiment thereof.

[0019] The self-collision prediction method for serial robotic arms provided in this application acquires an observation data set and then combines the observation data set with the joint command values ​​of motion planning to predict the future multi-cycle joint angle sequence in multiple steps. It does not rely on the pre-stored standard operating trajectory of the robotic arm and can dynamically deduce the real motion trend based on the joint motion observation data collected in real time by the robotic arm. It can adapt to various working conditions such as servo lag of the robotic arm that cause trajectory deviation. Furthermore, it calculates the minimum link spacing from the pre-stored effective collision pairing set based on the future multi-cycle joint angle sequence, and then predicts whether there is a risk of self-collision in the motion trajectory of the robotic arm in a preset time period based on the minimum link spacing. Compared with the static verification logic for fixed points of the preset trajectory, it can dynamically restore the real motion deviation of the robotic arm and avoid the prediction distortion problem caused by only making deviation judgments on the preset trajectory, thereby improving the accuracy of collision risk prediction. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating the self-collision prediction method for a serial robotic arm according to an embodiment of this application; Figure 2 This is a flowchart illustrating another serial robotic arm self-collision prediction method according to an embodiment of this application; Figure 3 This is a flowchart illustrating another self-collision prediction method for a serial robotic arm according to an embodiment of this application; Figure 4 This is a flowchart illustrating another serial robotic arm self-collision prediction method according to an embodiment of this application; Figure 5 This is a flowchart illustrating another serial robotic arm self-collision prediction method according to an embodiment of this application. Figure 6 This is a flowchart illustrating a self-collision prevention method for a serial robotic arm according to an embodiment of this application. Figure 7 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] According to an embodiment of this application, a method for predicting self-collision of a serial robotic arm is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0024] This embodiment provides a self-collision prediction method for serial robotic arms, which can be used in the aforementioned computer devices, such as computers and servers. Figure 1This is a flowchart of a self-collision prediction method for a serial robotic arm according to an embodiment of this application, such as... Figure 1 As shown, the process includes the following steps: Step S101: Obtain the observation data set; wherein, the observation data set includes joint motion observation data of the robotic arm in each control cycle.

[0025] The observation data can indicate the data continuously collected according to the control cycle of the robotic arm, specifically including joint motion observation data for each cycle.

[0026] As an example, under force-controlled conditions, the observation dataset can also include external force observation data.

[0027] Specifically, the observation data set can be obtained by reading the joint motion observation data fed back by the joint photoelectric encoders configured in each joint of the robotic arm. Alternatively, the observation data set can be obtained by collecting joint motion observation data from other acquisition devices. The specific method is not limited here and can be implemented by those skilled in the art.

[0028] For example, using the UR5e six-axis serial collaborative robotic arm, its controller is set to a 2ms control cycle. In each cycle, it synchronously reads the pulse signals of 6 joint encoders and converts them into joint motion observation data; it continuously caches the observation values ​​of multiple frames in a continuous time series to generate an observation data set.

[0029] Step S102: Based on the observation data set and the joint command values ​​of motion planning, predict the future multi-cycle joint angle sequence in multiple steps.

[0030] The joint command values ​​of motion planning can indicate the target rotation angle that each joint of the robotic arm should theoretically reach when performing the current task, without errors and without external force disturbances.

[0031] In the future, multi-cycle joint angle sequences can characterize the motion trajectory of the robotic arm within the predicted cycle.

[0032] Specifically, after determining the set of observation data, the joint command values ​​from motion planning can be further combined to predict future multi-cycle joint angle sequences.

[0033] As an example, a third-order uniformly accelerated motion model and a temporal state transition matrix are fitted based on the joint motion observation data in the observation dataset. The joint command values ​​of the motion planning are used as the control input of the Kalman filter. The optimal joint motion state estimate for the current cycle is obtained by iterative correction and convergence. The number of predicted recursive steps N is dynamically adjusted according to the motion speed of the robotic arm end effector, where N=6 for low speed, N=8 for medium speed, and N=10 for high speed. The joint prediction angles for all future control cycles are calculated recursively for each cycle and integrated to obtain the joint angle sequence for future multi-cycles.

[0034] As an example, for the special working condition of a heavy-duty serial robotic arm with nonlinear flexible deformation of the robotic arm links, the discrete Kalman filter used for time-series state prediction is replaced with an extended Kalman filter to complete the calculation. Specifically, based on the algorithmic characteristics of the extended Kalman filter, the nonlinear motion state of the robotic arm link is approximated by a first-order Taylor expansion to eliminate the nonlinear interference of the motion model itself. Then, using the observation dataset as the measured correction input and the joint command values ​​of the motion planning as the ideal control input, iterative calculations of the prediction and correction stages are executed repeatedly to continuously converge and solve for the optimal joint motion state estimate that adapts to the flexible deformation of the link. After completing the solution of the optimal joint motion state estimate, multiple (e.g., 10) robotic arm control cycles are forwarded, and the corresponding estimated joint angle is calculated iteratively cycle by cycle. All joint angle estimates are integrated in time sequence to generate a future multi-cycle joint angle sequence.

[0035] Step S103: Calculate the minimum link spacing from the pre-stored set of valid collision pairs based on the future multi-cycle joint angle sequence.

[0036] The pre-stored set of valid collision pairs indicates the link pair dataset retained after the offline preprocessing stage filters all links of the robotic arm for pairwise combinations. Specifically, the pre-stored set of valid collision pairs may include link combinations that have been removed due to parent-child connections and no possibility of interference throughout the process.

[0037] The minimum link spacing indicates the minimum value among all the Euclidean distances of the links calculated after traversing the future prediction time and effective collision pairing. It is used to characterize the distance between the closest robotic arm links and the position with the highest collision risk in the predicted trajectory.

[0038] Specifically, after determining the future multi-cycle joint angle sequence, the minimum link spacing can be further calculated from the pre-stored set of valid collision pairs.

[0039] As an example, for each predicted moment within a future multi-cycle joint angle sequence, the pose of the link in the world coordinate system is calculated in the forward direction. Using an offline pre-generated and locally stored set of effective collision pairs, a BVH layered bounding volume space pruning algorithm is used for coarse judgment. If the spacing between higher-level bounding volumes is greater than the warning threshold, fine ranging of the lower-level triangular facets is skipped, and distance calculation is only performed for close-range pairs. The spacing values ​​of all moments and all pairs are summarized, and the minimum global value is selected as the minimum link spacing.

[0040] As an example, the minimum Euclidean distance is iteratively calculated for each link pair within the effective collision pairing set. Specifically, for the 3D convex hull geometry model corresponding to each link pair, the search interval for the distance between the two convex hulls is continuously reduced iteratively until convergence, and the minimum Euclidean distance of that link pair is output. After traversing all link pairs within the effective collision pairing set, the minimum Euclidean distance of all link pairs at the current prediction time is summarized, and the minimum Euclidean distance at that time is selected. Then, all prediction times included in the future multi-cycle joint angle sequence are traversed, and the minimum Euclidean distances corresponding to all times are compared to determine the minimum link distance of the predicted trajectory.

[0041] Step S104: Based on the minimum link spacing, predict whether there is a risk of self-collision in the movement trajectory of the robotic arm during a preset time period.

[0042] Specifically, after determining the minimum link spacing, it is possible to further predict whether there is a risk of self-collision in the movement trajectory of the robotic arm during a preset time period.

[0043] As an example, the calculated minimum link spacing is compared with the first preset safety threshold (a pre-set value). If the minimum link spacing is less than the first preset safety threshold, it is determined that the movement trajectory of the robotic arm has a risk of self-collision within a preset time period consisting of multiple cycles in the future. If the minimum link spacing is not less than the first preset safety threshold, it is determined that there is no risk of self-collision within the preset time period.

[0044] The self-collision prediction method for serial robotic arms provided in this application acquires an observation data set and then combines the observation data set with the joint command values ​​of motion planning to predict the future multi-cycle joint angle sequence in multiple steps. It does not rely on the pre-stored standard operating trajectory of the robotic arm and can dynamically deduce the real motion trend based on the joint motion observation data collected in real time by the robotic arm. It can adapt to various working conditions such as servo lag of the robotic arm that cause trajectory deviation. Furthermore, it calculates the minimum link spacing from the pre-stored effective collision pairing set based on the future multi-cycle joint angle sequence, and then predicts whether there is a risk of self-collision in the motion trajectory of the robotic arm in a preset time period based on the minimum link spacing. Compared with the static verification logic for fixed points of the preset trajectory, it can dynamically restore the real motion deviation of the robotic arm and avoid the prediction distortion problem caused by only making deviation judgments on the preset trajectory, thereby improving the accuracy of collision risk prediction.

[0045] In one possible implementation, such as Figure 2 As shown, the multi-step recursive prediction of future multi-cycle joint angle sequences based on the observation data set and motion planning joint command values ​​in step S102 above may include the following steps: Step S2021: Based on the joint motion observation data of multiple consecutive frames in the observation data set, determine the joint angular velocity and joint angular acceleration corresponding to each joint.

[0046] The joint observation angles of multiple consecutive frames can be understood as the actual rotation angles of multiple joints that are continuously sampled by the robotic arm's sensors according to the control cycle and arranged in a temporal order.

[0047] Joint angular velocity indicates the amount of angle change of a single joint of a robotic arm per unit time, and is used to characterize how fast the joint rotates.

[0048] Joint angular acceleration can indicate the change in angular velocity of a single joint of a robotic arm per unit time, and is used to characterize the trend of joint rotation speed.

[0049] Specifically, after determining the set of observation data, the joint observation angles of multiple consecutive frames can be further determined to determine the joint angular velocity and joint angular acceleration corresponding to the joint observation angles.

[0050] As an example, based on the robotic arm's 10ms control cycle, 20 frames of time-series joint observation angles are continuously and equally collected to construct continuous multi-frame joint motion observation data; a first-order difference algorithm is used to perform differential operations on the joint observation angles of adjacent frames to calculate the joint angular velocity corresponding to each frame; then, a second-order difference operation is performed on the obtained joint angular velocity to obtain the corresponding joint angular acceleration.

[0051] Step S2022: Based on joint motion observation data, joint angular velocity, and joint angular acceleration, determine the temporal evolution law and state transition relationship between joint angle, angular velocity, and angular acceleration.

[0052] The temporal evolution law can characterize the continuous change trend, temporal correlation characteristics, and dynamic evolution rules of the joint observation angle, joint angular velocity, and joint angular acceleration of the robotic arm over time, and is used to constrain the temporal continuity of joint motion.

[0053] The state transition relationship can characterize the mapping relationship of the iterative update of the joint motion angle of the robotic arm between adjacent sampling frames and adjacent prediction periods.

[0054] Specifically, after determining the joint motion observation data, joint angular velocity, and joint angular acceleration, the temporal evolution law and state transition relationship between joint angle, angular velocity, and angular acceleration can be further determined.

[0055] As an example, we summarize the joint observation angle, joint angular velocity, and joint angular acceleration corresponding to multiple consecutive frames, and statistically analyze the temporal change amplitude, change rate, and coupling correlation characteristics of the joint observation angle, joint angular velocity, and joint angular acceleration to obtain the temporal evolution law of joint motion. Based on the motion state difference and iterative change characteristics of adjacent frames, we construct the inter-frame state transition relationship of joint motion state.

[0056] As an example, the time sequence intervals are divided according to the movement stages of the robotic arm, into three types of movement intervals: the start-up acceleration segment, the constant speed operation segment, and the deceleration and reset segment. Based on the change characteristics of the three parameters in different movement intervals, the time sequence evolution law of the corresponding intervals is fitted piecewise, and differentiated time sequence constraint coefficients are matched. At the same time, the state transition relationship of the corresponding intervals is constructed piecewise.

[0057] Step S2023: Based on the temporal evolution law, state transition relationship and joint command value, perform state iterative update to obtain the optimal joint motion state estimate.

[0058] Optimal joint motion state estimation can be understood as the joint command value that integrates the observation data set and motion planning.

[0059] Specifically, after determining the temporal evolution pattern, state transition relationship, and joint command value, the optimal joint motion state estimate can be further determined.

[0060] As an example, the temporal evolution law is combined to constrain the trend of joint motion temporal change, and the predicted state of joint motion in the current frame is obtained by deducing frame by frame according to the state transition relationship. The predicted state is first theoretically corrected by the ideal joint angle in the current frame to eliminate theoretical trajectory deviation. Then, the joint observation angle, joint angular velocity, and joint angular acceleration measured in the current frame are used for the second measured correction to correct the real working condition error caused by mechanical servo lag and transmission backlash, and determine the optimal joint motion state estimate of the robotic arm.

[0061] Specifically, step S2023 above may include the following steps: Step a1: Based on the temporal evolution pattern, state transition relationship, joint command value, and the optimal joint motion state estimate of the previous cycle, the joint predicted running state estimate for the current cycle is derived.

[0062] Specifically, after determining the temporal evolution pattern, state transition relationship, joint command value, and the optimal joint motion state estimate of the previous cycle, the joint prediction operation state estimate of the current cycle can be further deduced.

[0063] As an example, a third-order uniformly accelerated motion model is used to fit the temporal evolution law and state transition matrix; the optimal joint motion state estimate of the previous cycle is loaded, the joint command value issued by the motion planning is fused as the control input, Kalman filter prediction operation is performed, and the joint prediction running state estimate of the current cycle is generated recursively.

[0064] Step a2: Based on the joint motion observation data of the current period in the observation dataset, the joint predicted running state estimate is corrected to obtain the intermediate joint motion state estimate.

[0065] Specifically, after determining the joint motion observation data for the current period within the observation dataset, the estimated motion state of intermediate joints can be further determined.

[0066] As an example, the discrete Kalman filter is used to update the correction logic, calculate the residual between the joint motion observation data and the joint predicted running state estimate in the current cycle, that is, to allocate correction weights through Kalman gain, and simultaneously correct the three parameters of angle, angular velocity and angular acceleration to generate the intermediate joint motion state estimate.

[0067] Step a3 involves repeatedly performing the steps of estimating the joint motion state based on the temporal evolution law, state transition relationship, joint command value, and the optimal joint motion state of the previous cycle, and then deriving the joint predicted running state estimate for the current cycle based on the joint motion observation data of the current cycle within the observation data set. The steps of correcting the joint predicted running state estimate and obtaining the intermediate joint motion state estimate are repeated until the fluctuation amplitude of the intermediate joint motion state estimate is less than the preset error threshold, thus obtaining the optimal joint motion state estimate.

[0068] The preset error threshold can be understood as a pre-set threshold.

[0069] The fluctuation amplitude can be understood as the difference between the values ​​of joint angle, angular velocity, and angular acceleration in the motion state estimates of intermediate joints generated in two adjacent iterations. The fluctuation amplitude is used to measure the degree of convergence of the iterations.

[0070] Specifically, repeat steps a1 to a2 until the fluctuation amplitude of the intermediate joint motion state estimation is less than the preset error threshold, and the optimal joint motion state estimation can be obtained.

[0071] Step S2024: Based on the optimal joint motion state estimation, predict the future multi-cycle joint angle sequence in multiple steps.

[0072] Specifically, after determining the optimal joint motion state estimate, a multi-cycle joint angle sequence can be further determined.

[0073] As an example, the optimal joint motion state estimate from the discrete Kalman filter output is used as the starting point for deduction. The real-time motion speed of the robotic arm end effector is read in real time to dynamically adjust the total number of recursive steps: 6 control cycles for low-speed operation, 8 control cycles for medium-speed operation, and 10 control cycles for high-speed operation. The joint angle is iteratively updated according to the temporal evolution law and state transition relationship at each step, and the joint angle sequence of future multi-cycles is generated by integrating all predicted cycle angles.

[0074] The self-collision prediction method for serial robotic arms provided in this application solves for the joint angular velocity and joint angular acceleration corresponding to each joint based on the joint motion observation data of multiple consecutive frames in the observation data set. It can restore the real-time dynamic motion characteristics of the robotic arm. Then, by combining the joint motion observation data, joint angular velocity and joint angular acceleration, the temporal evolution law and state transition relationship between joint angle, angular velocity and angular acceleration can be determined. It can establish a temporal constraint and state iteration mapping logic that fits the real motion of the robotic arm. By utilizing temporal evolution patterns, state transition relationships, and joint command values, the system iteratively updates the state and obtains the optimal joint motion state estimate. It integrates measured motion data with theoretical command values ​​to correct motion deviations caused by various working conditions. Finally, based on the optimal joint motion state estimate, it recursively predicts future multi-cycle joint angle sequences in multiple steps, effectively reducing trajectory prediction errors and improving the accuracy of link spacing calculation and self-collision risk prediction.

[0075] In one possible implementation, the observation dataset also includes external force observation data of the robotic arm under force control conditions, such as... Figure 3 As shown, the multi-step recursive prediction of future multi-cycle joint angle sequences based on the observation data set and motion planning joint command values ​​in step S102 above may include the following steps: Step S3021: Based on the joint motion observation data of multiple consecutive frames in the observation data set and the joint command values ​​of the motion planning, determine the joint angle sequence to be processed.

[0076] The joint angle sequence to be processed can be understood as the angle sequence that needs to be processed.

[0077] Specifically, after determining the joint motion observation data and joint command values ​​of motion planning for multiple consecutive frames within the observation dataset, the joint angle sequence to be processed can be further determined.

[0078] Step S3022: Based on the external force observation data of the robotic arm under force control conditions, the joint angle to be processed corresponding to the current prediction cycle is perturbed and compensated to obtain the corrected joint angle.

[0079] External force observation data can be understood as data such as external contact forces, environmental disturbance forces, and load changes experienced by the robotic arm during the operation of the target working condition. Specifically, it can include dimensional information such as the magnitude, direction, and point of application of the force.

[0080] Specifically, during the dynamic operation of the robotic arm, the joint observation angles and external force observation data of the robotic arm under force control conditions can be acquired. Then, based on the external force observation data of the robotic arm under force control conditions, the joint angle to be processed corresponding to the current prediction cycle is perturbation compensation correction to obtain the corrected joint angle. The specific method for determining the corrected joint angle is described below.

[0081] Specifically, step S3022 above may include the following steps: Step b1: Determine the degree of external disturbance of the robotic arm based on the external force observation data of the robotic arm under force control conditions.

[0082] The degree of external disturbance to the robotic arm can characterize the magnitude and degree of influence of external environment, load change, contact interference and other factors on the robotic arm's motion trajectory within the current prediction period.

[0083] Specifically, after determining the external force observation data, the degree of external disturbance to the robotic arm corresponding to the current prediction cycle can be further determined.

[0084] As an example, real-time external force observation data corresponding to the current prediction cycle is retrieved. Under force control conditions, the Kalman process noise matrix is ​​adaptively adjusted based on the external force observation values ​​collected by the end six-dimensional force sensor to compensate for the state estimation deviation caused by the flexible deformation of the connecting rod and external contact disturbances. By comparing with the preset external force disturbance classification standard, the external disturbance is classified into three levels: slight disturbance, moderate disturbance, and severe disturbance based on the overall fluctuation amplitude and force magnitude of the external force observation data.

[0085] Step b2: Based on the degree of external disturbance to the robotic arm, adjust the uncertainty parameters to obtain the correction iteration parameters; wherein, the uncertainty parameters indicate the correction coefficients used to correct the temporal evolution amplitude and the state transition amplitude.

[0086] Uncertainty parameters can be understood as dynamic correction coefficients used to correct the temporal evolution and state transition amplitude of the robotic arm joint motion.

[0087] The first correction iteration parameter can be understood as the iterative correction parameter that adapts to the current disturbance condition after dynamically adjusting the uncertainty parameter according to the degree of external disturbance in the current period.

[0088] Specifically, after determining the degree of external disturbance in the current prediction period, the uncertainty parameter can be adjusted according to the degree of external disturbance to obtain the first correction iteration parameter.

[0089] As an example, based on the level of external disturbance corresponding to the current prediction period, a preset parameter adjustment rule is matched. Different disturbance levels correspond to different uncertainty parameter adjustment ranges. The uncertainty parameter values ​​used for time series evolution range correction and state transition range correction are synchronously corrected according to the disturbance level to generate the first correction iteration parameter adapted to the current disturbance condition.

[0090] Step b3: Based on the correction iteration parameters, perform perturbation compensation correction on the joint angle to be processed corresponding to the current prediction cycle to obtain the corrected joint angle.

[0091] Correcting the joint angle can be understood as the joint angle of the current prediction cycle being corrected by disturbance compensation to eliminate external disturbance deviations and adapt to external force conditions.

[0092] Specifically, after determining the correction iteration parameters, the predicted joint angle corresponding to the current prediction cycle can be further perturbed and compensated to obtain the corrected joint angle.

[0093] As an example, the correction iteration parameters corresponding to the current prediction period are retrieved, and the correction iteration parameters are substituted into the trajectory correction operation logic to perform perturbation compensation correction on the original prediction joint angle of the current prediction period. The trajectory error caused by the time series evolution deviation and the state transition deviation is corrected simultaneously to obtain the corrected joint angle.

[0094] Step S3023: Using the corrected joint angle as the input for the next prediction cycle, repeatedly execute the step of perturbation compensation correction of the joint angle to be processed corresponding to the current prediction cycle based on the external force observation data of the robotic arm under force control conditions, until all the joint angles to be processed in the joint angle sequence have completed perturbation compensation correction, and obtain the joint angle sequence of future multi-cycles.

[0095] Specifically, using the corrected joint angle completed in the current cycle as the iterative benchmark, the process iterates sequentially to the next prediction cycle, continuously performing external disturbance degree determination, dynamic adjustment of uncertainty parameters, solution of correction iteration parameters, and prediction joint angle disturbance compensation correction. The process optimizes all prediction parameters in the initial future joint angle sequence sequentially and cycle by cycle, correcting the differential disturbance deviations in different cycles, until the prediction joint angles of all time-series nodes in the initial future joint angle sequence have completed adaptive disturbance compensation correction, thus determining the future multi-cycle joint angle sequence.

[0096] As an example, taking the corrected joint angles completed in the current cycle as the basis for iteration, the next time node is taken as the current prediction cycle. The steps of determining the degree of disturbance of external force observation data, adjusting uncertainty parameters, solving for correction iteration parameters, and compensating for disturbance of predicted joint angles are repeatedly executed. The trajectory parameters of the initial future joint angle sequence are iteratively optimized cycle by cycle, and the external disturbance deviation of each cycle is continuously compensated until all predicted joint angles contained in the initial future joint angle sequence have completed disturbance compensation correction, generating a future multi-cycle joint angle sequence that adapts to the external force disturbance conditions.

[0097] The self-collision prediction method for serial robotic arms provided in this application determines the degree of external disturbance of the robotic arm corresponding to the current prediction cycle through external force observation data. It can determine the actual magnitude of external force interference in each prediction cycle based on external force monitoring data, thereby identifying dynamic disturbance conditions. Then, based on the degree of external disturbance in the current prediction cycle, it dynamically adjusts the uncertainty parameters used to correct the time-series evolution amplitude and state transition amplitude to obtain correction iteration parameters. It can adaptively match and adapt the correction coefficients according to different disturbance intensity, solving the problem that fixed correction parameters cannot adapt to dynamic disturbance changes, and making the correction parameters highly matched with the real-time disturbance conditions. Then, based on the first correction iteration parameters, the predicted joint angle corresponding to the current prediction cycle is perturbation compensation correction to obtain the corrected joint angle. This can specifically offset the temporal evolution deviation and state transition deviation caused by external disturbances, effectively correcting the trajectory offset error of the original predicted joint angle. Finally, based on the corrected joint angle, the steps of disturbance degree judgment, parameter adjustment and disturbance compensation correction are repeated for the next prediction cycle until all predicted joint angles in the initial future joint angle sequence have completed disturbance compensation correction and the future multi-cycle joint angle sequence is determined, improving the realism and reliability of the robot arm's future motion trajectory prediction.

[0098] In one possible implementation, such as Figure 4 As shown, step S103 above, which calculates the minimum link spacing from the pre-stored effective collision pairing set based on the future multi-cycle joint angle sequence, may include the following steps: Step S4031: Based on the future joint angle sequence, determine the world coordinate system pose of the robotic arm link at the target prediction time; wherein, the target prediction time is any prediction time corresponding to the future joint angle sequence.

[0099] The target prediction time can be understood as any discrete prediction time within the future time range covered by the future joint angle sequence.

[0100] The links of a robotic arm can indicate the independent link units that constitute the rigid body motion structure of a multi-degree-of-freedom robotic arm.

[0101] The world coordinate system pose indicates the set of spatial position coordinates and attitude angles of each link under the world absolute coordinate system established with the robotic arm's working reference space.

[0102] Specifically, after determining the future joint angle sequence, the world coordinate system pose of the robotic arm link at the target prediction time can be further determined.

[0103] As an example, the predicted angles of each joint at any target prediction time in the future joint angle sequence are extracted, the forward kinematics solution algorithm of the robotic arm is called, coordinate transformation and attitude matrix solution are completed in sequence, and the coordinates are uniformly transformed to the world coordinate system of the robotic arm. The world coordinate system poses of all links of the robotic arm at the target prediction time are output in batches.

[0104] As an example, based on the standard forward kinematics solution, combined with the trajectory deviation characteristics of the previous disturbance compensation correction, and superimposed with the attitude compensation amount corresponding to the small flexible deformation of the robotic arm, assembly error, and transmission clearance, the calculated pose of each link at the target prediction time is finely adjusted and corrected; based on the temporal change characteristics of the future joint angle sequence, the link attitude offset law under dynamic working conditions is adapted to obtain the compensated link world coordinate system pose.

[0105] Step S4032: Based on the link world coordinate system pose at the target prediction time and the pre-stored effective collision pairing set, determine the minimum Euclidean distance of the risk link pairing corresponding to the target prediction time.

[0106] The minimum spacing between links can be understood as the minimum Euclidean spacing between risky link pairs at the moment of single target prediction, representing the minimum safe spacing under the overall spatial attitude of the robotic arm at that moment.

[0107] Specifically, after determining the world coordinate system pose of the link at the target prediction time, the minimum Euclidean distance of the risk link pairing corresponding to the target prediction time can be determined by combining the pre-stored effective collision pairing set, thus obtaining the minimum link distance.

[0108] As an example, the pre-stored set of robot arm collision pairs is retrieved, all risk link pairs within the set are traversed, and the three-dimensional Euclidean distance between risk link pairs is calculated group by group based on the world coordinate system pose of each link at the target prediction time. The minimum Euclidean distance of all pairs at the current time is then summarized.

[0109] As an example, the pre-stored set of robot arm collision pairs is retrieved, all risk link pairs within the set are traversed, and the three-dimensional Euclidean distance between risk link pairs is calculated group by group using the GJK convex hull iterative distance solution algorithm, combined with the world coordinate system pose of each link at the target prediction time. The distance data of all pairs at the current time are summarized, and the minimum distance value is selected by numerical comparison and determined as the minimum distance between links at the target time.

[0110] Step S4033: Determine the smallest Euclidean distance among multiple minimum Euclidean distances as the minimum link distance.

[0111] Specifically, after determining the minimum Euclidean spacing of all pairs, the sizes of the minimum Euclidean spacing of all pairs can be compared, and the minimum spacing of the smallest link can be determined as the minimum link spacing.

[0112] The self-collision prediction method for a serial robotic arm provided in this application determines the world coordinate system pose of each link of the robotic arm at the target prediction time based on the future joint angle sequence at any prediction time corresponding to the future joint angle sequence. This can convert the temporal dimension of joint angle motion data into three-dimensional spatial dimension of link pose data, solving the defect that a single joint angle parameter cannot characterize the spatial distribution state of the robotic arm links. Then, based on the link world coordinate system pose at the target prediction time and the pre-stored effective collision pairing set, the minimum Euclidean distance of the risk link pairing corresponding to the target prediction time is specifically solved to obtain the minimum link distance. By pre-setting the risk link pairing set, high-risk detection objects are locked. Then, the minimum link distance with the smallest Euclidean distance is selected from the minimum link distances of multiple target times and determined as the minimum link distance corresponding to the trajectory, avoiding the risk omission problem caused by single-time or local time period detection.

[0113] In one possible implementation, such as Figure 5 As shown, step S104 above, predicting whether the robot arm's motion trajectory has a self-collision risk based on the minimum link spacing, may include the following steps: Step S5041: Detect whether the minimum link spacing is less than the first preset safety threshold.

[0114] The first preset safety threshold can be a pre-defined safety threshold. Detecting whether the minimum link spacing is less than the first preset safety threshold can include two scenarios.

[0115] Case 1: The minimum link spacing is less than the first preset safety threshold; Case 2: The minimum link spacing is not less than the first preset safety threshold.

[0116] Step S5042: If the minimum link spacing is less than the first preset safety threshold, then the robot arm's motion trajectory during the preset time period is predicted to have a risk of self-collision.

[0117] In case of scenario one, it can be determined that the predicted trajectory of the robotic arm during the preset time period carries a risk of self-collision.

[0118] Step S5043: If the minimum link spacing is not less than the first preset safety threshold, then the predicted movement trajectory of the robotic arm during the preset time period does not have the risk of self-collision.

[0119] In case two, it can be determined that the predicted trajectory of the robotic arm during the preset time period does not pose a risk of self-collision.

[0120] The self-collision prediction method for serial robotic arms provided in this application compares the minimum link spacing with a pre-set first safety threshold to clearly distinguish between two judgment scenarios: the minimum link spacing is less than the first safety threshold and the minimum link spacing is not less than the first safety threshold. When the minimum link spacing is less than the first safety threshold, it is directly determined that the movement trajectory of the robotic arm during a preset time period has a risk of self-collision. When the minimum link spacing is not less than the first safety threshold, it is determined that the movement trajectory of the robotic arm during a preset time period does not have a risk of self-collision, thereby improving the accuracy of self-collision risk prediction.

[0121] In one possible implementation, the above method may also include the following steps: Step c1: Obtain the linkage pairing combination.

[0122] Link pairing combinations can indicate the set of pairs formed by randomly combining all the links of the robotic arm in pairs. These pairing combinations can include parent-child link pairings, permanently interference-free link pairings, and link pairings with a risk of self-collision.

[0123] Specifically, in the offline preprocessing stage, the independent links of the robotic arm are identified based on the parameter model and 3D link model corresponding to the robotic arm; pairwise enumeration and pairing are performed on the independent links, and all pairwise pairing relationships are summarized to generate link pairing combinations.

[0124] Step c2: Based on the parent-child relationship of the links, remove link pairs with direct parent-child connections from the link pair combinations, and remove link pairs that will not interfere with the robot arm in any pose, to obtain a pre-stored set of valid collision pairs.

[0125] After determining the link pairings, link pairs with direct parent-child connections can be removed, as well as link pairs that do not interfere with the robot arm in any pose, to obtain a pre-stored set of valid collision pairs.

[0126] The self-collision prediction method for serial robotic arms provided in this application obtains all combinations of link pairings that cover all pairs of links in the robotic arm, avoiding the omission of link pairings with self-collision risks. Then, based on the parent-child relationship of the links, link pairings with direct parent-child connections are removed from the link pairing combinations. At the same time, link pairings that will not interfere with the robotic arm in any pose are also removed, resulting in a pre-stored set of effective collision pairings. When calculating the minimum link spacing, only the pre-stored set of effective collision pairings needs to be traversed, without having to traverse all link pairing combinations, thus reducing the amount of calculation required for real-time ranging.

[0127] This embodiment provides a self-collision prevention method for serial robotic arms, which can be used in the aforementioned computer equipment, such as computers and servers. Figure 6 This is a flowchart of a self-collision prevention method for a serial robotic arm according to an embodiment of this application, such as... Figure 6 As shown, the process includes the following steps: Step S601: Determine the self-collision risk assessment result of the robotic arm's motion trajectory during the preset time period.

[0128] Step S602: If the movement trajectory of the robotic arm during the preset time period has a risk of self-collision, detect whether the minimum link spacing corresponding to the predicted target time is less than the second preset safety threshold.

[0129] The second preset safety threshold can indicate the preset threshold for determining the collision risk of the robotic arm.

[0130] Specifically, if the robotic arm's movement trajectory during a preset time period carries a risk of self-collision, it can further detect whether the minimum link spacing corresponding to the target prediction time is less than a second preset safety threshold.

[0131] Step S603: If the minimum link spacing corresponding to the target prediction time is not less than the second preset safety threshold, determine the first operation of the robotic arm from the preset correlation relationship.

[0132] The preset association can be understood as a one-to-one mapping relationship between pre-stored working conditions with different risk levels and the safe operation strategy of the robotic arm. Specifically, it can include the corresponding adaptive operation instructions for different working conditions with low risk and high risk.

[0133] The first operation can instruct a gentle safety control operation of the robotic arm to match low-risk approach conditions, corresponding to potential approach risk conditions where the minimum link spacing is between the second preset safety threshold and the first preset safety threshold.

[0134] Specifically, if the minimum link spacing corresponding to the target prediction time is not less than the second preset safety threshold, the first operation of the robotic arm can be further determined from the preset correlation relationship.

[0135] As an example, when the minimum link spacing at the target prediction time is detected to be less than the first preset safety threshold and not less than the second preset safety threshold, the robotic arm is determined to be in a low-risk approach condition. The preset correlation is retrieved, the first operation corresponding to the low-risk condition is matched, and the robotic arm is subjected to a uniform deceleration and a slight trajectory adjustment optimization operation to slow down the movement rhythm of the robotic arm and fine-tune the movement trajectory.

[0136] Step S604: If the minimum link spacing corresponding to the target prediction time is less than the second preset safety threshold, determine the second operation of the robotic arm from the preset correlation relationship.

[0137] The second operation can be understood as an emergency safety protection operation of the robotic arm to match high-risk critical collision conditions, corresponding to high-risk conditions where the minimum linkage distance is less than the second preset safety threshold.

[0138] Specifically, after determining that the minimum link spacing corresponding to the target prediction time is less than the second preset safety threshold, the second operation of the robotic arm can be further determined from the preset correlation relationship.

[0139] As an example, when the minimum link spacing at the target prediction time is less than the second preset safety threshold, the robotic arm is determined to be in a high-risk critical collision condition. The preset correlation is retrieved to match the corresponding second operation, and the robotic arm is controlled to perform emergency deceleration and instantaneous braking protective operations.

[0140] As an example, for high-risk critical working conditions, the second operation of the robotic arm’s safe retraction is triggered by a preset correlation. This not only performs an emergency braking operation, but also controls the robotic arm to retract slightly in the opposite direction to a safe distance range based on the approach direction of the linkage.

[0141] Prevention strategies can indicate differentiated safety control strategies matched to different collision risk levels, specifically including low-risk flexible protection strategies and high-risk emergency protection strategies.

[0142] Specifically, after determining the minimum link spacing, the robot arm's prevention strategy can be determined based on the comparison between the minimum link spacing and the target threshold.

[0143] For example, the minimum link spacing and warning thresholds (such as the first warning threshold mentioned above) and emergency stop thresholds (such as the second warning threshold mentioned above) at each prediction time are compared in layers. The specific numerical scenarios that can be implemented are as follows: the preset warning threshold is 15mm and the preset emergency stop threshold is 5mm. When the minimum link spacing is greater than or equal to 15mm, it is determined that there is no collision risk for the current robotic arm, and the preset operating speed and motion trajectory are maintained for normal operation. When the minimum link spacing is between 5mm and 15mm, it is determined to be a low-risk collision hazard, and a flexible warning strategy of scaling the joint speed by 30% to 50% and replanning the local trajectory is implemented to avoid risks without interrupting the operation. When the minimum link spacing is less than 5mm, it is determined to be a high-risk collision hazard, and an emergency protection strategy of immediately cutting off the trajectory, safely retracting the robotic arm to a safe position, and switching to a compliant buffer mode is implemented.

[0144] The self-collision prevention method for a serial robotic arm provided in this application embodiment detects whether the minimum link spacing corresponding to the target prediction time is less than a second preset safety threshold, which is smaller, when the robotic arm's movement trajectory during a preset time period poses a risk of self-collision. Based on the two-level threshold gradient structure of the first and second preset safety thresholds, the method achieves a graded judgment of the robotic arm's approach risk, effectively distinguishing between low-risk approach conditions and high-risk critical collision conditions. This solves the problems of single threshold judgment failing to identify risk gradients and coarse protection strategies. In low-risk approach conditions where the minimum link spacing corresponding to the target prediction time is not less than the second preset safety threshold, the first operation of the robotic arm is determined through a preset correlation relationship. This allows for gentle early intervention and control of minor potential risks, suppressing risk escalation while ensuring the continuity and efficiency of robotic arm operations. In high-risk critical collision conditions where the minimum link spacing corresponding to the target prediction time is less than the second preset safety threshold, the second operation of the robotic arm is determined through a preset correlation relationship. This allows for emergency safety protection operations to be performed for high-risk collisions, quickly blocking dangerous movement trends and avoiding collision hazards.

[0145] This embodiment also provides a self-collision prediction device for a serial robotic arm, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0146] This embodiment provides a self-collision prediction device for a serial robotic arm. The device includes: an acquisition module for acquiring an observation data set, wherein the observation data set includes joint motion observation data of the robotic arm in each control cycle; a first prediction module for recursively predicting the joint angle sequence of future multi-cycles based on the observation data set and the joint command values ​​of the motion plan; a calculation module for calculating the minimum link spacing from a pre-stored set of valid collision pairs based on the future multi-cycle joint angle sequence; and a second prediction module for predicting whether there is a risk of self-collision in the motion trajectory of the robotic arm during a preset time period based on the minimum link spacing.

[0147] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0148] In this embodiment, the robotic arm collision prevention device is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0149] This application also provides a computer device having the above-described robotic arm collision prevention device.

[0150] Please see Figure 7 , Figure 7 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of this application, such as... Figure 7 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 7 Take a processor 10 as an example.

[0151] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0152] The memory 20 stores instructions executable by at least one processor 10 to cause at least one processor 10 to perform the method shown in the above embodiments.

[0153] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0154] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0155] The computer device also includes an input device 30 and an output device 40. The processor 10, memory 20, input device 30, and output device 40 can be connected via a bus or other means. Figure 7 Taking the example of a connection between China and Israel via a bus.

[0156] Input device 30 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the computer device, such as a touchscreen, keypad, mouse, trackpad, touchpad, joystick, one or more mouse buttons, trackball, joystick, etc. Output device 40 may include display devices, auxiliary lighting devices (e.g., LEDs), and haptic feedback devices (e.g., vibration motors). The aforementioned display devices include, but are not limited to, liquid crystal displays, light-emitting diodes, displays, and plasma displays. In some alternative embodiments, the display device may be a touchscreen.

[0157] The computer device also includes a communication interface for communicating with other devices or communication networks.

[0158] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the methods shown in the above embodiments are implemented.

[0159] A portion of this application can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to this application through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0160] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A method for predicting self-collision in a serial robotic arm, characterized in that, The method includes: Acquire an observation data set; wherein, the observation data set includes joint motion observation data of the robotic arm in each control cycle; Based on the observed data set and the joint command values ​​of the motion planning, the future multi-cycle joint angle sequence is predicted in a multi-step recursive manner. Based on the future multi-cycle joint angle sequence, the minimum link spacing is calculated from the pre-stored effective collision pair set; Based on the minimum link spacing, predict whether the robotic arm's motion trajectory during a preset time period carries a risk of self-collision.

2. The self-collision prediction method for serial robotic arms according to claim 1, characterized in that, The multi-step recursive prediction of future multi-cycle joint angle sequences based on the observed data set and motion planning joint command values ​​includes: Based on continuous multi-frame joint motion observation data within the observation dataset, determine the joint angular velocity and joint angular acceleration corresponding to each joint. Based on the joint motion observation data, the joint angular velocity, and the joint angular acceleration, the temporal evolution law and state transition relationship between the joint angle, angular velocity, and angular acceleration are determined. Based on the temporal evolution pattern, the state transition relationship, and the joint command value, perform state iterative update to obtain the optimal joint motion state estimate; Based on the optimal joint motion state estimation, the future multi-cycle joint angle sequence is predicted in a multi-step recursive manner.

3. The self-collision prediction method for serial robotic arms according to claim 2, characterized in that, The step of performing state iterative updates based on the temporal evolution pattern, the state transition relationship, and the joint command value to obtain the optimal joint motion state estimate includes: Based on the aforementioned temporal evolution pattern, the aforementioned state transition relationship, the aforementioned joint command value, and the optimal joint motion state estimate of the previous cycle, the joint predicted running state estimate for the current cycle is derived. Based on the joint motion observation data of the current period in the observation dataset, the joint predicted running state estimate is corrected to obtain the intermediate joint motion state estimate. The process of repeatedly executing the steps of estimating the joint movement state based on the time-series evolution law, the state transition relationship, the joint command value, and the optimal joint motion state estimate of the previous cycle, and then estimating the joint movement state based on the joint motion observation data of the current cycle within the observation data set, and correcting the joint movement state estimate to obtain the intermediate joint motion state estimate, continues until the fluctuation amplitude of the intermediate joint motion state estimate is less than a preset error threshold, thus obtaining the optimal joint motion state estimate.

4. The self-collision prediction method for serial robotic arms according to claim 1, characterized in that, The observation data set also includes external force observation data of the robotic arm under force control conditions; Based on the observed data set and the joint command values ​​from the motion planning, a multi-step recursive prediction of future multi-cycle joint angle sequences is performed, including: Based on the joint motion observation data of multiple consecutive frames and the joint command values ​​of the motion planning within the observation data set, the joint angle sequence to be processed is determined. Based on the external force observation data of the robotic arm under force control conditions, the angle of the joint to be processed corresponding to the current prediction cycle is perturbated and corrected to obtain the corrected joint angle. Using the corrected joint angle as the input for the next prediction cycle, the process of repeatedly performing disturbance compensation correction on the joint angle to be processed corresponding to the current prediction cycle based on the external force observation data of the robotic arm under force control conditions is repeated until all the joint angles to be processed in the joint angle sequence have been disturbed and compensated, thus obtaining the future multi-cycle joint angle sequence.

5. The self-collision prediction method for serial robotic arms according to claim 4, characterized in that, The step of performing disturbance compensation correction on the joint angle to be processed corresponding to the current prediction cycle based on the external force observation data of the robotic arm under force control conditions to obtain the corrected joint angle includes: Based on the external force observation data of the robotic arm under force control conditions, the degree of external disturbance of the robotic arm is determined; Based on the degree of external disturbance to the robotic arm, the uncertainty parameter is adjusted to obtain the correction iteration parameter; wherein, the uncertainty parameter indicates the correction coefficient used to correct the temporal evolution amplitude and the state transition amplitude; Based on the correction iteration parameters, the joint angle to be processed corresponding to the current prediction cycle is perturbation compensation correction to obtain the corrected joint angle.

6. The self-collision prediction method for serial robotic arms according to claim 1, characterized in that, The calculation of the minimum link spacing from a pre-stored set of valid collision pairs based on future multi-cycle joint angle sequences includes: Based on the future joint angle sequence, the world coordinate system pose of the robotic arm link at the target prediction time is determined; wherein, the target prediction time is any prediction time corresponding to the future joint angle sequence; Based on the link world coordinate system pose at the target prediction time and the pre-stored effective collision pairing set, determine the minimum Euclidean distance of the risk link pairing corresponding to the target prediction time. The smallest of the multiple minimum Euclidean distances is determined as the minimum link distance.

7. The self-collision prediction method for serial robotic arms according to claim 1, characterized in that, The method of predicting whether there is a risk of self-collision in the movement trajectory of the robotic arm during a preset time period based on the minimum link spacing includes: Detect whether the minimum link spacing is less than a first preset safety threshold; If the minimum link spacing is less than the first preset safety threshold, it is predicted that the movement trajectory of the robotic arm during the preset time period has a risk of self-collision. If the minimum link spacing is not less than the first preset safety threshold, then the predicted movement trajectory of the robotic arm during the preset time period does not have a risk of self-collision.

8. The self-collision prediction method for a serial robotic arm according to any one of claims 1-7, characterized in that, The method further includes: Obtain the link pairing combination; Based on the parent-child relationship of the links, links that have a direct parent-child connection are removed from the link pairings, as well as links that will not interfere with the robot arm in any pose, to obtain a pre-stored set of valid collision pairs.

9. A method for preventing self-collision in a series robotic arm, characterized in that, The method includes: Using the self-collision prediction method for serial robotic arms as described in any one of claims 1-8, the self-collision risk assessment result of the robotic arm's motion trajectory during a preset time period is determined; If the self-collision risk determination result indicates that the movement trajectory of the robotic arm during a preset time period has a self-collision risk, detect whether the minimum link spacing is less than the second preset safety threshold. If the minimum link spacing is not less than the second preset safety threshold, the first operation of the robotic arm is determined from the preset association relationship; If the minimum link spacing is less than the second preset safety threshold, the second operation of the robotic arm is determined from the preset association relationship.

10. A computer device, characterized in that, include Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the self-collision prevention method for a serial robotic arm as described in any one of claims 1 to 8 and / or the self-collision prevention method for a serial robotic arm as described in claim 9.

Images