Hardware real-time instruction calibration method for robot edge control

Abstract

The application discloses a hardware real-time instruction calibration method for robot edge control and particularly relates to the technical field of robot control, constructs a multivariate polynomial clock prediction model containing the number of memory access, jump and data processing instructions and nonlinear coupling terms and updates the model online; the actual consumption is obtained frame by frame by using a hardware counter, and the instruction time deviation is compared and calculated; when the deviation is out of limit, segmented calibration parameters are generated to perform timing compensation, and when the joint motion change rate is out of limit, first-order linear smoothing truncation is performed to reconstruct the calibration instruction sequence; the calibration sequence is input into a hardware shadow register area in parallel to perform forward and reverse differential simulation verification, and when the differential value meets the standard, the output is latched, otherwise, the state is rolled back and recalibration is adaptively contracted; the application effectively eliminates the microscopic time delay jitter in the edge chip, and significantly improves the multi-axis synchronous precision and safety defense capability of the robot.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and more specifically, to a hardware real-time instruction calibration method for robot edge control. Background Technology

[0002] As industrial robots evolve towards higher speeds, higher precision, and more dynamic collaboration, the real-time performance and synchronization accuracy of edge controllers have become bottlenecks determining robot motion performance. In modern robot control systems, the master control layer typically uses an industrial Ethernet bus to communicate with the joint servo drivers at the edge and issue control commands. To ensure multi-joint coordination, existing technologies often employ clock synchronization techniques to align the communication timing between the master control layer and the edge, thereby resolving communication delays caused by network jitter at the bus level.

[0003] However, in real-world hard real-time control scenarios, simply achieving clock alignment at the bus level is far from sufficient. Once the control command stream reaches the edge control chip, the chip needs to decode the commands, perform inverse trajectory calculations, fuse multi-sensor data, and output closed-loop control algorithms. During this micro-execution process, due to the dynamic combination of different instruction types, such as memory access instructions, jump instructions, and data processing instructions, unpredictable pipeline conflicts, cache misses, and bus arbitration delays can occur within the edge hardware. This results in non-linear dynamic jitter in the actual instruction execution time within the edge hardware. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a hardware real-time instruction calibration method for robot edge control, thereby addressing the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a hardware real-time instruction calibration method for robot edge control, comprising the following steps:

[0006] S1. Identify and extract the number of instructions of different control instruction types in the target control instruction sequence as independent variables. The instruction types include at least memory access instructions, jump instructions, and data processing instructions. Based on the number of instructions and their nonlinear coupling terms, construct a multivariate polynomial clock cycle prediction model. The undetermined coefficients of the multivariate polynomial clock cycle prediction model are identified and updated online based on historical execution data.

[0007] S2. During the execution of the target instruction sequence by the edge control chip, the actual clock cycle consumption is obtained frame by frame online using a hardware counter; the actual clock cycle consumption is compared with the prediction output of the multivariate multinomial clock cycle prediction model to calculate the hardware instruction execution time deviation.

[0008] S3. When the hardware instruction execution time deviation exceeds the preset deviation threshold, segmented timing calibration parameters are generated based on the deviation to dynamically compensate the subsequent instruction sequence to be executed. When the calibrated time parameters cause the expected rate of change of the robot joint to exceed the threshold of the hardware limit register, the calibration amount is truncated by the hardware flow control unit in a first-order linear smooth manner.

[0009] S4. Input the calibrated instruction sequence in parallel into the hardware shadow register area, perform forward execution simulation and reverse backtracking simulation respectively, and calculate the difference value between the forward and reverse simulation results; when the difference value does not exceed the preset difference threshold, latch the calibration instruction sequence into the physical output register and output it; when the difference value exceeds the preset difference threshold, trigger the hardware state to roll back to the previous calibration point and recalibrate.

[0010] Preferably, as a preferred embodiment of the hardware real-time instruction calibration method for robot edge control according to the present invention, S1 specifically includes the following:

[0011] A multivariate polynomial model is constructed to predict the execution time of the target control command sequence within the edge control chip. Its mathematical expression is as follows: ,in, The total number of clock cycles for the predicted target control instruction sequence; The total number of control instruction types; The first in the target control command sequence The statistical number of control instructions is used as the input independent variable of the model. The instruction types include at least memory access instructions, jump instructions, and data processing instructions. For the first The single-term delay sensitivity coefficient of control commands; For the first Class control instructions and the first Nonlinear coupling terms in control instructions; This represents the coupling delay coefficient corresponding to the nonlinear coupling term; This represents the deviation of the system's inherent constants.

[0012] Let the vector of undetermined coefficients be... The corresponding input feature vector is Then the polynomial model can be simplified as follows: Its matrix dimensions are all determined by the total number of instruction types. The decision is made, and the dimension size is... And perform online estimation of the coefficient vector to be determined;

[0013] During the initialization phase, initial identification is performed, and historical execution data is continuously used for online updates during subsequent operation. In the initialization phase, the control edge chip executes a preset calibration test instruction set, and a hardware counter synchronously records the number of each type of instruction in the calibration test instruction set during each execution, constructing the corresponding input feature vector. And simultaneously collect the corresponding actual execution clock cycle consumption. This constitutes the historical execution dataset. ,in, The total number of samples;

[0014] Input the historical execution dataset into the recursive least squares identification algorithm, and determine the coefficient vector. Perform online estimation and iterative updates, with the following update formula:

[0015] ;

[0016] ;

[0017] ;

[0018] in, For the first The gain vector of the next iteration; For the first The covariance matrix after the next iteration For the first The covariance matrix after the next iteration; and The first Second and third The estimated value of the undetermined coefficient vector after the next iteration; Input feature vector transpose; The forgetting factor is set.

[0019] Preferably, as a preferred embodiment of the hardware real-time instruction calibration method for robot edge control according to the present invention, S2 specifically includes the following:

[0020] During the edge control chip initialization phase, the start and stop signals of the internally integrated hardware performance counter are hardwired and bound to the communication frame reception hardware interrupt of the robot edge bus and the start and end flags of the kernel instruction decoding pipeline; the counting frequency of the timing channel is configured to be synchronized with the main frequency of the edge control chip, and the hardware performance counter is configured in the debugging domain of the edge control chip, enabling the shadow reading mechanism based on hardware flow control; so that the start, stop, and value reading actions of the hardware performance counter are directly triggered by the hardware control signals of the instruction pipeline.

[0021] During the execution of the target control instruction sequence by the edge control chip, when the kernel instruction decoding unit receives the start instruction of the target instruction sequence, the hardware trigger source immediately activates the hardware performance counter to start counting; when the end instruction of the target instruction sequence is executed and the result is latched, the hardware trigger source immediately stops the hardware performance counter, reads the value in the hardware performance counter, and records it as the actual clock cycle consumption value of the target instruction sequence in the current frame. During the execution of the target control instruction sequence, the edge control chip uses a hardware-level task isolator to lock interrupt requests and concurrent bus applications from non-target control tasks.

[0022] By utilizing the hardware adder array built into the edge control chip, the statistical quantity vector of each type of instruction in the target instruction sequence of the current frame is substituted into the multivariate polynomial clock cycle prediction model in S1 to calculate the predicted clock cycle consumption value of the current frame. , is represented as: ;in, For the first The estimated value of the undetermined coefficient vector after the second iteration transpose, This is the input feature vector corresponding to the target instruction sequence in the current frame;

[0023] The actual clock cycle consumption is calculated using the hardware subtractor of the edge control chip. Compared with the predicted clock cycle consumption value absolute deviation between ;

[0024] Based on the current operating frequency of the edge control chip The absolute deviation is converted by a physical time converter. Converted into microscopic physical time deviation The specific formula is as follows ; to measure the microscopic physical time deviation The hardware instruction execution time deviation of the current frame is output and sent to S3 for timing compensation judgment.

[0025] Preferably, as a preferred embodiment of the hardware real-time instruction calibration method for robot edge control according to the present invention, S3 specifically includes the following:

[0026] The current frame hardware instruction execution time deviation output in S2 The deviation threshold is passed to the hardware comparator built into the edge control chip and compared with the preset system tolerance deviation threshold. Perform a comparison;

[0027] when Segmented timing calibration parameters Set to 0;

[0028] when The edge control chip operates according to the current communication bus cycle. With the remaining execution window of the instruction sequence to be executed, generate segmented timing calibration parameters. Its mathematical piecewise function expression is as follows:

[0029] ;

[0030] in, For piecewise scaling factors, It is a symbolic function;

[0031] The hardware flow control unit of the edge control chip receives segmented timing calibration parameters. The physical time of the next control command update will be adjusted to... ,in, To specify the nominal update time, timing calibration parameters are broadcast via a multi-axis hardwired bus. The data is synchronously distributed to the control timing registers corresponding to all collaborative joints of the robot arm; and the phase lock signal of each joint control timer is set to perform microscopic time axis alignment.

[0032] Using the kinematics solving accelerator built into the edge control chip, based on the compensated update time Calculate the expected angular velocity of each joint of the robotic arm, as the expected rate of change of physical motion. ; indicates as: ;in, This represents the joint position command value at the target timing point. This refers to the previous actual update time. This represents the actual joint position value at the previous moment.

[0033] Read the preset maximum safe rate of change of joint motion threshold from the hardware-built-in limit register. The hardware comparator is used to determine whether the condition is met. :

[0034] If the conditions are not met, the truncation protection is not activated; the current timing calibration parameters are directly retained, and the target time-domain compensation calibration amount is set. ;

[0035] When the conditions are met, the hardware first-order linear slope limiter is activated to truncate and smooth the actual output compensation calibration amount, and the timing compensation calibration amount after safe truncation is calculated. , is represented as: ; Set the target time-domain compensation calibration value ;in, For symbolic functions, This represents the joint position command value at the target timing point. This refers to the previous actual update time. This represents the actual joint position value at the previous moment. This is the nominal update time;

[0036] Calculate the unfinished compensation time difference caused by the clipping using a hardware difference register. The specific formula is as follows: and the time difference The accumulated data is temporarily stored in the segmented timing calibration calculation for the next bus cycle;

[0037] After locking the rate of change of physical motion through a series circuit of a hardware comparator and a first-order linear slope limiter, the instruction reconstruction module of the edge control chip is used to calibrate the amount based on timing compensation. Modify the control timing register value to reconstruct and generate a calibration instruction sequence containing the execution timing and joint position instructions after calibration. It is then passed to S4 as the final calibration output of S3.

[0038] Preferably, as a preferred embodiment of the hardware real-time instruction calibration method for robot edge control according to the present invention, S4 specifically includes the following:

[0039] The hardware flow control unit is used to process the calibration command sequence output from S3. The image is copied and loaded in parallel into the first and second image channels of the hardware shadow register area, which is physically isolated from the physical output registers used to control the servo driver. The first and second image channels within the hardware shadow register area serve as the physical hardware carriers for executing the forward execution simulation and the reverse backtracking simulation, respectively.

[0040] Within the same hardware clock cycle, the calibration instruction sequence in the first and second image channels is processed by a dual-channel hardware accelerator. Parallel differential simulation is performed, and the specific simulation process is as follows:

[0041] Forward execution simulation uses the angle vector of the robot's joints at the current moment. With angular velocity vector Build system status Starting from the input point, the expected end-effector pose state after the control command is executed is calculated forward using the forward kinematics simulation hardware unit. ;

[0042] The reverse backtracking simulation is based on the expected end-target pose. As the endpoint, the minimum physical drive timing required for the robot to reach the target state is verified by reverse backtracking through the inverse dynamics verification hardware unit. ;

[0043] The outputs of the forward kinematics simulation hardware unit and the inverse kinematics verification hardware unit are connected to a hardware pipeline alignment buffer. By inserting a hardware no-operation cycle, the expected end-effector pose state is achieved. With minimum physical drive timing Align the outputs on the same hardware synchronous clock edge;

[0044] The hardware differential is used to calculate the joint difference between the analog outputs of the two image channels and the system reference state in real time. ,in, Represents the L2 norm distance; and The preset forward state deviation weights and backward backtracking deviation weights are used to quantitatively evaluate the calibration command sequence from two dimensions: spatial trajectory deviation and historical state continuity. Reliability during physical execution;

[0045] The joint difference value is determined by a hardware comparator. Does it exceed the preset differential security threshold? :

[0046] when At the next bus synchronization clock edge, the calibration instruction sequence in the hardware shadow register area is latched by a hardware latch signal. The data is latched, updated in the physical output register, and then output to the servo driver.

[0047] when The hardware flow control unit blocks the update enable signal of the physical output register, rolls back the system state pointer to the previous level safety calibration point state stored in the spare area of ​​the hardware shadow register through the hardware state machine, and updates the segmented timing calibration parameters. Piecewise scaling factor Perform adaptive halving shrinkage and re-execute the calibration and verification process from S3 to S4;

[0048] During the recalibration process, the number of retries is accumulated using a spin safety monitoring counter integrated within the hardware flow control unit; when the retry count reaches a preset hard limit threshold... When this occurs, the hardware state machine forcibly interrupts the recalibration process, locks and resets the robot's current joint command speed to zero, and outputs a hardware-level emergency brake lock signal to the outside.

[0049] On the other hand, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements a functional module of a hardware real-time instruction calibration method for robot edge control as described above.

[0050] On the other hand, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements a hardware real-time instruction calibration method for robot edge control as described above.

[0051] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0052] By constructing a multivariate polynomial model that integrates nonlinear coupling terms of the microarchitecture, nanosecond-level accurate prediction of dynamic instruction execution jitter within the edge hardware is achieved. Hardware performance counters are used to imperceptibly capture actual timing, and segmented compensation and multi-axis hardwired broadcasting enable forced alignment of multiple joints on the microscopic time axis, completely eliminating synchronization misalignment caused by pipeline conflicts and bus arbitration. A hardware first-order linear slope limiter and differential rolling buffer mechanism are introduced to lock physical overshoot while eliminating static pose accumulation errors. Based on hardware shadow registers and dual-channel heterogeneous simulation, highly reliable forward and reverse squeeze verification and deadlock adaptive rollback are completed without affecting current control. This invention significantly improves the robot's hard real-time control accuracy, multi-axis collaborative capability, and overall operational safety. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0054] Figure 1 This is a schematic diagram of the overall process of the method of the present invention;

[0055] Figure 2 This is a schematic diagram of the hardware structure of the dual-track parallel calibration pipeline in this invention;

[0056] Figure 3 This is the principle and logic control diagram of the hardware limiting circuit for the first-order linear slope limiter in this invention. Detailed Implementation

[0057] 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, and 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.

[0058] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0059] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0060] Example 1

[0061] This embodiment provides, for example Figure 1 The hardware real-time instruction calibration method for robot edge control, as shown, specifically includes the following steps:

[0062] S1. Identify and extract the number of instructions of different control instruction types in the target control instruction sequence as independent variables. The instruction types include at least memory access instructions, jump instructions, and data processing instructions. Based on the number of instructions and their nonlinear coupling terms, construct a multivariate polynomial clock cycle prediction model. The undetermined coefficients of the multivariate polynomial clock cycle prediction model are identified and updated online based on historical execution data.

[0063] S2. During the execution of the target instruction sequence by the edge control chip, the actual clock cycle consumption is obtained frame by frame online using a hardware counter; the actual clock cycle consumption is compared with the prediction output of the multivariate multinomial clock cycle prediction model to calculate the hardware instruction execution time deviation.

[0064] S3. When the hardware instruction execution time deviation exceeds the preset deviation threshold, segmented timing calibration parameters are generated based on the deviation to dynamically compensate the subsequent instruction sequence to be executed. When the calibrated time parameters cause the expected rate of change of the robot joint to exceed the threshold of the hardware limit register, the calibration amount is truncated by the hardware flow control unit in a first-order linear smooth manner.

[0065] S4. Input the calibrated instruction sequence in parallel into the hardware shadow register area, perform forward execution simulation and reverse backtracking simulation respectively, and calculate the difference value between the forward and reverse simulation results; when the difference value does not exceed the preset difference threshold, latch the calibration instruction sequence into the physical output register and output it; when the difference value exceeds the preset difference threshold, trigger the hardware state to roll back to the previous calibration point and recalibrate.

[0066] Preferably, in step S1, the number of instructions of different control instruction types in the target control instruction sequence is identified and extracted as an independent variable. The instruction types include at least memory access instructions, jump instructions, and data processing instructions. Based on the number of instructions and their nonlinear coupling terms, a multivariate polynomial clock cycle prediction model is constructed. The undetermined coefficients of the multivariate polynomial clock cycle prediction model are identified and updated online based on historical execution data. Specifically, this includes the following:

[0067] A multivariate polynomial model is constructed to predict the execution time of the target control command sequence within the edge control chip. Its mathematical expression is as follows: ,in, The total number of clock cycles for the predicted target control instruction sequence; The total number of control instruction types; The first in the target control command sequence The statistical number of control instructions is used as the input independent variable of the model. The instruction types include at least memory access instructions, jump instructions, and data processing instructions. For the first The single-term delay sensitivity coefficient of control commands; For the first Class control instructions and the first Nonlinear coupling terms in control instructions; This represents the coupling delay coefficient corresponding to the nonlinear coupling term; This represents the deviation of the system's inherent constants.

[0068] The nonlinear coupling term Corresponding to specific hardware conflict behaviors: When For memory access instructions and When it is a jump instruction, the nonlinear coupling term This is used to characterize the nonlinear cumulative delay introduced by the instruction pipeline when the edge control chip misses the cache during jump target addressing, resulting in bubbles.

[0069] Let the vector of undetermined coefficients be... The corresponding input feature vector is Then the polynomial model can be simplified as follows: Its matrix dimensions are all determined by the total number of instruction types. The decision is made, and the dimension size is... And perform online estimation of the coefficient vector to be determined;

[0070] The kernel instruction decoding unit in the edge control chip is used to decode the target control instruction sequence, identify and extract the number of each type of instruction; the extracted instruction number is sent to the instruction statistics array to construct the input feature vector; the hardware adder array is used to calculate the predicted clock cycle number based on the transpose of the feature vector and the current coefficient vector.

[0071] During the initialization phase, initial identification is performed, and historical execution data is continuously used for online updates during subsequent operation. In the initialization phase, the control edge chip executes a preset calibration test instruction set, and a hardware counter synchronously records the number of each type of instruction in the calibration test instruction set during each execution, constructing the corresponding input feature vector. And simultaneously collect the corresponding actual execution clock cycle consumption. This constitutes the historical execution dataset. ,in, The total number of samples;

[0072] The historical execution dataset is input into the recursive least squares identification algorithm deployed at the edge, to determine the coefficient vector. Perform online estimation and iterative updates, with the following update formula:

[0073] ;

[0074] ;

[0075] ;

[0076] in, For the first The gain vector of the next iteration; For the first The covariance matrix after the next iteration For the first The covariance matrix after the next iteration; and The first Second and third The estimated value of the undetermined coefficient vector after the next iteration; Input feature vector transpose; The forgetting factor is set, and This is used to enable the model to prioritize learning the latest hardware execution features;

[0077] The initial matrix of the covariance matrix is: ,in, For dimension The identity matrix, The amplification factor is a preset value, and the engineering value range of the amplification factor is set to... This is used to give the model sufficient parameter optimization convergence speed in the early stage of identification, so as to ensure that the parameter convergence of the clock prediction model is completed quickly during the power-on initialization stage of the robot edge controller.

[0078] Preferably, in step S2, during the execution of the target instruction sequence by the edge control chip, the actual clock cycle consumption is obtained frame-by-frame online using a hardware counter; the actual clock cycle consumption is compared with the prediction output of the multivariate polynomial clock cycle prediction model to calculate the hardware instruction execution time deviation; specifically, this includes the following:

[0079] During the initialization phase of the edge control chip, the start and stop signals of the internally integrated hardware performance counter are hard-wired and bound to the communication frame receiving hard interrupt of the robot edge bus and the start and end flags of the kernel instruction decoding pipeline; the counting frequency of the timing channel is configured to be synchronized with the main frequency of the edge control chip, and the actual clock cycle number is obtained through the clock management unit to achieve nanosecond-level hardware clock cycle resolution.

[0080] When configuring the timing channel, the method also includes: configuring the hardware performance counter in the debugging domain of the edge control chip and enabling the shadow reading mechanism based on hardware flow control to capture the actual clock cycle number of the target instruction sequence in a zero-interference / shadow reading manner, so that the start, stop and value reading actions of the hardware performance counter are directly triggered by the hardware control signal of the instruction pipeline, without the CPU executing additional read and write instructions, thereby achieving zero interference of measurement overhead on the execution time of the target instruction sequence.

[0081] During the execution of the target control instruction sequence by the edge control chip, when the kernel instruction decoding unit receives the start instruction of the target instruction sequence, the hardware trigger source immediately activates the hardware performance counter to start counting; when the end instruction of the target instruction sequence is executed and the result is latched, the hardware trigger source immediately stops the hardware performance counter, reads the value in the hardware performance counter, and records it as the actual clock cycle consumption value of the target instruction sequence in the current frame. ;

[0082] During the execution of the target control instruction sequence, the edge control chip uses a hardware-level task isolator to lock interrupt requests and concurrent bus applications from non-target control tasks, ensuring that the measured value of the actual clock cycle consumption corresponds only to the hardware pipeline time of the target control instruction sequence.

[0083] By utilizing the hardware adder array built into the edge control chip, the statistical quantity vector of each type of instruction in the target instruction sequence of the current frame is substituted into the online-updated multivariate multinomial clock cycle prediction model in S1 to calculate the predicted clock cycle consumption value of the current frame. , is represented as: ;in, For the first The estimated value of the undetermined coefficient vector after the second iteration transpose, This is the input feature vector corresponding to the target instruction sequence in the current frame;

[0084] The actual clock cycle consumption is calculated using the hardware subtractor of the edge control chip. Compared with the predicted clock cycle consumption value absolute deviation between ;

[0085] Based on the current operating frequency of the edge control chip The absolute deviation is converted by a physical time converter. Converted into microscopic physical time deviation The specific formula is as follows ; to measure the microscopic physical time deviation The hardware instruction execution time deviation of the current frame is output and sent to S3 for timing compensation judgment.

[0086] Preferably, in step S3, when the hardware instruction execution time deviation exceeds a preset deviation threshold, segmented timing calibration parameters are generated based on the deviation to dynamically compensate the subsequent instruction sequence to be executed; when the calibrated timing parameters cause the expected rate of change of the robot joint to exceed the hardware limiting register threshold, the calibration amount is truncated using a first-order linear smoothing mechanism via a hardware flow control unit; specifically, this includes the following:

[0087] The current frame hardware instruction execution time deviation output in S2 The deviation threshold is passed to the hardware comparator built into the edge control chip and compared with the preset system tolerance deviation threshold. Perform a comparison;

[0088] when If the current hardware latency is determined to be within a safe margin, the segmented timing calibration parameters will be adjusted. Set to 0 to maintain the original control instruction execution timing;

[0089] when The segmented calibration parameter calculation unit is triggered; the edge control chip calculates the parameters according to the current communication bus cycle. With the remaining execution window of the instruction sequence to be executed, generate segmented timing calibration parameters. Its mathematical piecewise function expression is as follows:

[0090] ;

[0091] in, It is a piecewise scaling factor, and This is used to gradually eliminate large deviations over several adjacent bus control cycles, preventing excessive timing compensation in a single operation from causing jumps in the output control signal. It is a symbolic function;

[0092] The hardware flow control unit of the edge control chip receives segmented timing calibration parameters. The control timing register value sent to the servo driver is directly modified to adjust the physical time of the next control command update. ,in, To determine the nominal update time, lag compensation is applied to the edge control timing in the time dimension to eliminate the impact of internal chip instruction timing jitter on joint synchronization accuracy.

[0093] During dynamic time compensation, the hardware flow control unit calculates the timing calibration parameters based on the multi-axis cooperative control relationship. The data is synchronously distributed to the control timing registers corresponding to all collaborative joints of the robot via a multi-axis hard-wired broadcast bus, and the phase lock signal of each joint control timer is forcibly set to achieve forced alignment of multiple joints on the micro time axis, avoiding inter-axis motion misalignment caused by independent timing compensation of a single joint.

[0094] Using the kinematics solving accelerator built into the edge control chip, based on the compensated update time Calculate the expected angular velocity of each joint of the robotic arm, as the expected rate of change of physical motion. ; indicates as: ;in, This represents the joint position command value at the target timing point. This refers to the previous actual update time. This represents the actual joint position value at the previous moment.

[0095] Read the preset maximum safe rate of change of joint motion threshold from the hardware-built-in limit register. The hardware comparator is used to determine whether the condition is met. :

[0096] If the conditions are not met, the truncation protection is not activated; the current timing calibration parameters are directly retained, and the target time-domain compensation calibration amount is set. ;

[0097] When the conditions are met, the hardware first-order linear slope limiter is activated to truncate and smooth the actual output compensation calibration amount, and the timing compensation calibration amount after safe truncation is calculated. , is represented as: ; Set the target time-domain compensation calibration value ;in, This is a sign function, taking values ​​of +1 or -1, used to preserve segmented timing calibration parameters. direction, This represents the joint position command value at the target timing point. This refers to the previous actual update time. This represents the actual joint position value at the previous moment. This is the nominal update time;

[0098] The timing compensation calibration amount after the safety truncation is calculated. Following this, it also includes a rolling temporary storage step for compensation differences: using a hardware difference register to calculate the unfinished compensation time difference caused by the limiting. and the difference The accumulated and temporarily stored data is used in the segmented timing calibration calculation of the next bus cycle to eliminate the long-term static pose accumulation error of the servo control system caused by the amplitude limiting cutoff.

[0099] The first-order linear smoothing truncation is performed in series by a hardware comparator, a first-order linear slope limiter, and a hardware convergence selector. The first input of the hardware convergence selector is connected to the direct output channel, and the second input is connected to the output of the hardware first-order linear slope limiter. Its control terminal is connected to the output judgment level of the hardware comparator. When the hardware comparator determines that the limit is not exceeded, the hardware convergence selector selects the first input and outputs the original timing calibration parameters directly. When the hardware comparator determines that the limit is exceeded, the hardware convergence selector selects the second input and outputs the safe calibration amount after truncation and smoothing by the first-order linear slope limiter. The above selection operation is completed within a single hardware clock cycle without introducing additional software branch delay.

[0100] By using a series circuit of a hardware comparator and a first-order linear slope limiter, the rate of change of the joint's physical motion is forcibly locked within a safe limit, thus preventing resonance in the robot's mechanical structure caused by excessive timing compensation from the bottom layer.

[0101] The instruction reconfiguration module of the edge control chip calibrates based on timing compensation. The control timing register value sent to the servo driver is directly modified, and the physical control update timing of the next moment and the target position data of the corresponding joint are repackaged and combined to reconstruct a calibration instruction sequence containing the calibration execution timing and joint position instructions. And pass it to S4 as the final calibration output of S3;

[0102] When the operating system or upper-level control software running in the edge control chip malfunctions, or when a task is suspended or a high-priority interrupt preempts, causing the software calibration logic to fail, the hardware comparator, the first-order linear slope limiter, and the hardware convergence selector all use pure combinational logic hardware circuits independent of the CPU software instruction stream to form a hardware safety protection channel. The hardware safety protection channel directly controls the hardware convergence selector to perform physical channel switching within a single clock cycle based on the judgment level output by the hardware comparator, forcibly truncating and locking the excessive target time-domain compensation calibration amount within a preset safety limit, so as to eliminate the robot joint control loss caused by software failure from the physical hardware level.

[0103] Preferably, in step S4, the calibrated instruction sequence is input in parallel to the hardware shadow register area, and forward execution simulation and reverse backtracking simulation are performed respectively, and the difference value between the forward and reverse simulation results is calculated; when the difference value does not exceed a preset difference threshold, the calibration instruction sequence is latched into the physical output register and output; when the difference value exceeds the preset difference threshold, the hardware state is triggered to roll back to the previous calibration point and recalibrate; specifically, this includes the following:

[0104] The edge control chip has a built-in independent hardware shadow register area, which is divided into a first image channel and a second image channel, respectively used as the physical hardware carriers for executing the forward execution simulation and the reverse backtracking simulation. The hardware shadow register area is physically isolated from the physical output registers that directly control the servo driver. The hardware flow control unit outputs the calibration instruction sequence in S3. The first and second image channels of the hardware shadow register area are copied and loaded in parallel so that the simulation verification process of the calibration instructions is completely unaffected by the robot's current physical control before the actual output.

[0105] Within the same hardware clock cycle, the edge control chip activates the dual-channel hardware accelerator to process the calibration command sequences in the first and second image channels. Parallel differential simulations were performed separately. The specific simulation process is as follows:

[0106] Forward execution simulation is based on the current robot state Starting from the input, the calibration command sequence is... The input is fed into the forward kinematics simulation hardware unit to calculate the expected end-effector pose state after the control command is executed. ;in, and These are the actual angle vectors and actual angular velocity vectors of each joint of the robot at the current moment;

[0107] The reverse backtracking simulation is based on the expected end-target pose. As the endpoint, the calibration instruction sequence will be... The input is fed into the inverse dynamics verification hardware unit, which backtracks and calculates the minimum physical drive timing required for the robot to reach the target state under the current physical constraints. ;

[0108] The output of the dual-channel hardware accelerator is connected to a hardware pipeline alignment buffer; the pipeline alignment buffer adjusts the expected end-effector pose state by inserting hardware idle cycles. With the minimum physical drive timing Aligning outputs on the same hardware synchronous clock edge eliminates hardware contention risks caused by heterogeneous dual-channel computing latency.

[0109] The hardware differential within the edge control chip calculates the joint difference between the analog outputs of the two image channels and the system reference state in real time. ,in, Represents the L2 norm distance; and The preset forward state deviation weights and backward backtracking deviation weights are used to quantitatively evaluate the calibration command sequence from two dimensions: spatial trajectory deviation and historical state continuity. Reliability during physical execution;

[0110] The joint difference value is determined by a hardware comparator. Does it exceed the preset differential security threshold? :

[0111] when This indicates that the limit has not been exceeded, and the calibration instruction sequence is determined. At the physical layer, the edge control chip is secure and reliable. On the next bus synchronization clock edge, it directly latches the calibration instruction sequence from the hardware shadow register area via a hardware latch signal. The data is latched and updated in the physical output register, and then output as a hard real-time instruction to the servo driver.

[0112] when This indicates that the limit has been exceeded and a risk of calibration distortion has been identified. The hardware flow control unit immediately blocks the update enable signal of the physical output register to maintain the current robot joint output state without distortion. The system state pointer is triggered by the hardware state machine to roll back to the previous level safe calibration point state stored in the spare area of ​​the hardware shadow register, and the segmented timing calibration parameters are updated. Piecewise scaling factor Perform adaptive halving shrinkage and re-execute the calibration and verification process from S3 to S4;

[0113] During the recalibration process, the spin safety monitoring counter integrated within the hardware flow control unit synchronously increments the retry count once; when the retry count reaches a preset hard limit threshold... When the recalibration process is interrupted, the hardware state machine forcibly interrupts the recalibration process, directly locks and resets the robot's current joint command speed to zero, and outputs a hardware-level emergency brake lock signal externally, causing the robot arm to enter a physically safe suspension state.

[0114] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0115] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the steps of the hardware real-time instruction calibration method for robot edge control as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0116] Example 2

[0117] The following is another embodiment of the present invention, which provides a hardware real-time instruction calibration method for robot edge control. In order to verify the beneficial effects of the present invention, a simulation experiment is conducted for scientific demonstration.

[0118] This experiment aims to verify the effectiveness of a real-time hardware instruction calibration method for robot edge control. Through techniques such as instruction feature extraction, online hardware counter acquisition, timing deviation comparison, timing compensation and limiting, dual-channel differential verification, and state rollback, the method enhances the early warning and real-time calibration capabilities for instruction execution jitter in edge control chips. The experiment uses simulated and actual edge control instruction execution data, including the number of instructions under different instruction combinations, predicted clock cycle, actual clock cycle, hardware delay deviation value, and multi-axis joint physical state characteristics. By analyzing the consistency between the model's clock prediction results and the actual hardware operation labels, the accuracy and robustness of this invention in identifying and predicting delay jitter are verified.

[0119] The simulation experiment steps are implemented according to the content of the hardware real-time instruction calibration method for robot edge control provided in Example 1, and the specific steps include:

[0120] Collect static and dynamic data of the edge control chip when executing control command streams, including command stream type, bus cycle, operating temperature, and main frequency parameters. The collection frequency is set to once per command frame.

[0121] Within the framework of a multivariate polynomial clock cycle prediction model, a prediction and evaluation model for edge control clock cycles is constructed. The influence of various instruction types and nonlinear coupling terms on internal hardware pipeline conflicts is considered, and their weights are evaluated. A recursive least squares algorithm is used for identification and analysis. All collected data is then integrated, and the following steps are performed:

[0122] a. Determine the evaluation criteria and weights for each type of control instruction;

[0123] b. Use a multivariate polynomial model for clock cycle prediction and deviation ranking;

[0124] c. Assign weights to the linear and nonlinear coupling terms of each feature factor to obtain the clock cycle prediction model;

[0125] A multivariate polynomial prediction model is trained using instruction features and actual cycle data from historical execution data of the edge control chip. The measured data is aligned in real time using a hardware performance counter to extract the latency jitter features during chip operation. The prediction model is then reconstructed by updating the coefficients. The trained polynomial model can update the instruction latency prediction results of the edge control chip in real time.

[0126] During the actual operation of the control chip, the trained prediction model is used to update the clock prediction results in real time. The clock deviation extracted by the prediction model is combined with the hardware state machine to perform dynamic compensation decision analysis. Combined with first-order linear smoothing truncation, the following operations are performed:

[0127] 1) Dynamically identify and predict instruction delay jitter, and take corresponding timing compensation measures in a timely manner;

[0128] 2) Adjust instruction execution time parameters in real time to optimize the real-time control performance of the edge control chip;

[0129] 3) During operation, the dual-channel differential verification results are continuously updated to provide real-time support for the health and safety management of the robot control system.

[0130] The specific data from the above simulation experiment are shown in Table 1.

[0131] Table 1 Data Recording Table of Simulation Experiment of the Invention

[0132] Time / minute Instruction sequence type Operating frequency / MHz Ambient temperature / °C Operating parameters (pressure / flow rate / number of commands) Predict clock cycles / times Actual clock cycles / Deviation / Truncation / Rollback Status Risk management (time-sequence compensation) measures 0-5 Basic memory access instruction stream 400 45 Memory access: 10; Jump: 2; Data processing: 50 500 502 No / No / No Start regular monitoring 5-10 Basic jump instruction stream 400 47 Memory access: 12; Jumps: 5; Data processing: 48 512 515 No / No / No Maintain routine monitoring 10-15 Hybrid highly coupled instruction stream 400 50 Memory access: 35; Redirects: 15; Data processing: 80 1200 1215 Yes / No / No Triggering segmented timing compensation 15-20 Strong burst dense command stream 400 55 Memory access: 60; Redirects: 30; Data processing: 120 1850 1920 Yes / Yes / No First-order smooth truncation limiting protection 20-25 Strong interference abnormal command stream 400 58 Memory access: 80; Redirects: 45; Data processing: 150 2300 2480 Yes / Yes / Yes Triggering hardware rollback and parameter recalibration 25-30 Restore normal instruction flow 400 52 Memory access: 15; Jumps: 3; Data processing: 60 620 621 No / No / No Restore normal monitoring

[0133] Experimental Analysis:

[0134] By comparing the conformity between the model prediction results and the actual hardware operation labels, the accuracy and robustness of the present invention in identifying and predicting instruction execution delay deviations are verified. Experiments show that the hardware real-time instruction calibration method for robot edge control has high accuracy and reliability, and can effectively improve the real-time control performance and physical motion safety of the robot throughout its entire life cycle.

[0135] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A hardware real-time instruction calibration method for robot edge control, characterized in that: Includes the following steps: S1. Identify and extract the number of instructions of different control instruction types in the target control instruction sequence as independent variables. The instruction types include at least memory access instructions, jump instructions, and data processing instructions. Based on the number of instructions and their nonlinear coupling terms, construct a multivariate polynomial clock cycle prediction model. The undetermined coefficients of the multivariate polynomial clock cycle prediction model are identified and updated online based on historical execution data. S2. During the execution of the target instruction sequence by the edge control chip, the actual clock cycle consumption is obtained frame by frame online using a hardware counter. The actual clock cycle consumption is compared with the predicted output of the multivariate polynomial clock cycle prediction model to calculate the hardware instruction execution time deviation. S3. When the hardware instruction execution time deviation exceeds the preset deviation threshold, segmented timing calibration parameters are generated based on the deviation to dynamically compensate the subsequent instruction sequence to be executed. When the calibrated time parameters cause the expected rate of change of the robot joint to exceed the threshold of the hardware limit register, the calibration amount is truncated by the hardware flow control unit in a first-order linear smooth manner. S4. Input the calibrated instruction sequence in parallel into the hardware shadow register area, perform forward execution simulation and reverse backtracking simulation respectively, and calculate the difference between the forward and reverse simulation results; When the difference value does not exceed the preset difference threshold, the calibration instruction sequence is latched into the physical output register and output. When the difference value exceeds the preset difference threshold, the hardware state is triggered to roll back to the previous calibration point and recalibrate.

2. The hardware real-time instruction calibration method for robot edge control according to claim 1, characterized in that: In step S1, a multivariate polynomial clock cycle prediction model is constructed, which specifically includes the following: A multivariate polynomial model is constructed to predict the execution time of the target control command sequence within the edge control chip. Its mathematical expression is as follows: ,in, The total number of clock cycles for the predicted target control instruction sequence; The total number of control instruction types; The first in the target control command sequence The statistical number of control instructions is used as the input independent variable of the model. The instruction types include at least memory access instructions, jump instructions, and data processing instructions. For the first The single-term delay sensitivity coefficient of control commands; For the first Class control instructions and the first Nonlinear coupling terms in control instructions; This represents the coupling delay coefficient corresponding to the nonlinear coupling term; This represents the deviation of the system's inherent constants. Let the vector of undetermined coefficients be... The corresponding input feature vector is Then the polynomial model can be simplified as follows: Its matrix dimensions are all determined by the total number of instruction types. The decision is made, and the dimension size is... And perform online estimation of the coefficient vector to be determined.

3. The hardware real-time instruction calibration method for robot edge control according to claim 2, characterized in that: The online estimation of the coefficient vector to be determined specifically includes: During the initialization phase, initial identification is performed, and historical execution data is continuously used for online updates during subsequent operation. In the initialization phase, the control edge chip executes a preset calibration test instruction set, and a hardware counter synchronously records the number of each type of instruction in the calibration test instruction set during each execution, constructing the corresponding input feature vector. And simultaneously collect the corresponding actual execution clock cycle consumption. This constitutes the historical execution dataset. ,in, The total number of samples; Input the historical execution dataset into the recursive least squares identification algorithm, and determine the coefficient vector. Perform online estimation and iterative updates, with the following update formula: ； ； ； in, For the first The gain vector of the next iteration; For the first The covariance matrix after the next iteration For the first The covariance matrix after the next iteration; and The first Second and third The estimated value of the undetermined coefficient vector after the next iteration; Input feature vector transpose; The forgetting factor is set.

4. The hardware real-time instruction calibration method for robot edge control according to claim 3, characterized in that: In step S2, during the execution of the target instruction sequence by the edge control chip, the actual clock cycle consumption is obtained frame-by-frame online using a hardware counter; specifically, this includes the following: During the initialization phase of the edge control chip, the start and stop signals of the internally integrated hardware performance counter are hardwired and bound to the communication frame receiving hard interrupt of the robot edge bus and the start and end flags of the kernel instruction decoding pipeline; the counting frequency of the timing channel is configured to be synchronized with the main frequency of the edge control chip, and the hardware performance counter is configured in the debugging domain of the edge control chip to enable the shadow reading mechanism based on hardware flow control. During the execution of the target control instruction sequence by the edge control chip, when the kernel instruction decoding unit receives the start instruction of the target instruction sequence, the hardware trigger source immediately activates the hardware performance counter to start counting; when the end instruction of the target instruction sequence is executed and the result is latched, the hardware trigger source immediately stops the hardware performance counter, reads the value in the hardware performance counter, and records it as the actual clock cycle consumption value of the target instruction sequence in the current frame. ; During the execution of the target control instruction sequence, the edge control chip uses a hardware-level task isolator to lock interrupt requests and concurrent bus applications from non-target control tasks.

5. The hardware real-time instruction calibration method for robot edge control according to claim 4, characterized in that: In step S2, the actual clock cycle consumption is compared with the predicted output of the multivariate polynomial clock cycle prediction model to calculate the hardware instruction execution time deviation, specifically including: By utilizing the hardware adder array built into the edge control chip, the statistical quantity vector of each type of instruction in the target instruction sequence of the current frame is substituted into the multivariate multinomial clock cycle prediction model to calculate the predicted clock cycle consumption value of the current frame. , represented as: ;in, For the first The estimated value of the undetermined coefficient vector after the second iteration transpose, This is the input feature vector corresponding to the target instruction sequence in the current frame; The actual clock cycle consumption is calculated using the hardware subtractor of the edge control chip. Compared with the predicted clock cycle consumption value absolute deviation between ; Based on the current operating frequency of the edge control chip The absolute deviation is converted by a physical time converter. Converted into microscopic physical time deviation The specific formula is as follows ; to measure the microscopic physical time deviation The hardware instruction execution time deviation of the current frame is output and sent to S3 for timing compensation judgment.

6. The hardware real-time instruction calibration method for robot edge control according to claim 5, characterized in that: In step S3, when the hardware instruction execution time deviation exceeds a preset deviation threshold, segmented timing calibration parameters are generated based on the deviation to dynamically compensate the subsequent instruction sequence to be executed; specifically, this includes the following: The current frame hardware instruction execution time deviation output in S2 The deviation threshold is passed to the hardware comparator built into the edge control chip and compared with the preset system tolerance deviation threshold. Perform a comparison; when Segmented timing calibration parameters Set to 0; when The edge control chip operates according to the current communication bus cycle. With the remaining execution window of the instruction sequence to be executed, generate segmented timing calibration parameters. Its mathematical piecewise function expression is as follows: ； in, For piecewise scaling factors, It is a symbolic function; The hardware flow control unit of the edge control chip receives segmented timing calibration parameters. The physical time of the next control command update will be adjusted to... ,in, To specify the nominal update time, timing calibration parameters are broadcast via a multi-axis hardwired bus. The data is synchronously distributed to the control timing registers corresponding to all collaborative joints of the robot arm; and the phase lock signal of each joint control timer is set to perform microscopic time axis alignment.

7. The hardware real-time instruction calibration method for robot edge control according to claim 6, characterized in that: In step S3, when the calibrated time parameters cause the expected rate of change of motion of the robot joint to exceed the threshold of the hardware limiting register, the calibration quantity is truncated by a first-order linear smoothing through the hardware flow control unit, specifically including the following: Using the kinematics solving accelerator built into the edge control chip, based on the compensated update time Calculate the expected angular velocity of each joint of the robotic arm, as the expected rate of change of physical motion. ; indicates as: ;in, This represents the joint position command value at the target timing point. This refers to the previous actual update time. This represents the actual joint position value at the previous moment. Read the preset maximum safe rate of change of joint motion threshold from the hardware-built-in limit register. The hardware comparator is used to determine whether the condition is met. : If the conditions are not met, the truncation protection is not activated; the current timing calibration parameters are directly retained, and the target time-domain compensation calibration amount is set. ; When the conditions are met, the hardware first-order linear slope limiter is activated to truncate and smooth the actual output compensation calibration amount, and the timing compensation calibration amount after safe truncation is calculated. , represented as: ; Let the target time-domain compensation calibration value ;in, For symbolic functions, This represents the joint position command value at the target timing point. This refers to the previous actual update time. This represents the actual joint position value at the previous moment. This is the nominal update time; Calculate the unfinished compensation time difference caused by the clipping using a hardware difference register. The specific formula is as follows: and the time difference The accumulated data is temporarily stored in the segmented timing calibration calculation for the next bus cycle; After locking the rate of change of physical motion through a series circuit of a hardware comparator and a first-order linear slope limiter, the instruction reconstruction module of the edge control chip is used to calibrate the amount based on timing compensation. Modify the control timing register value to reconstruct and generate a calibration instruction sequence containing the execution timing and joint position instructions after calibration. It is then passed to S4 as the final calibration output of S3.

8. The hardware real-time instruction calibration method for robot edge control according to claim 7, characterized in that: In step S4, the calibrated instruction sequence is input in parallel to the hardware shadow register area to perform forward execution simulation and reverse backtracking simulation, and the difference between the forward and reverse simulation results is calculated; specifically including: The hardware flow control unit is used to process the calibration command sequence output from S3. The image is copied and loaded in parallel into the first and second image channels of the hardware shadow register area, which is physically isolated from the physical output registers used to control the servo driver. The first and second image channels within the hardware shadow register area serve as the physical hardware carriers for executing the forward execution simulation and the reverse backtracking simulation, respectively. Within the same hardware clock cycle, the calibration instruction sequence in the first and second image channels is processed by a dual-channel hardware accelerator. Parallel differential simulation is performed, and the specific simulation process is as follows: Forward execution simulation uses the angle vectors of the robot's joints at the current moment. With angular velocity vector Build system status Starting from the input point, the expected end-effector pose state after the control command is executed is calculated forward using the forward kinematics simulation hardware unit. ; The reverse backtracking simulation is based on the expected end-target pose. As the endpoint, the minimum physical drive timing required for the robot to reach the target state is verified by reverse backtracking through the hardware unit of the inverse dynamics verification. ; The outputs of the forward kinematics simulation hardware unit and the inverse kinematics verification hardware unit are connected to a hardware pipeline alignment buffer. By inserting a hardware no-operation cycle, the expected end-effector pose state is achieved. With minimum physical drive timing Align the outputs on the same hardware synchronous clock edge.

9. The hardware real-time instruction calibration method for robot edge control according to claim 8, characterized in that: In step S4, when the difference value does not exceed the preset difference threshold, the calibration instruction sequence is latched into the physical output register and output. When the difference value exceeds a preset difference threshold, the hardware state is triggered to roll back to the previous calibration point and recalibrate; specifically including: The hardware differential is used to calculate the joint difference between the analog outputs of the two image channels and the system reference state in real time. ,in, Represents the L2 norm distance; and The preset forward state deviation weights and backward backtracking deviation weights are used to quantitatively evaluate the calibration command sequence from two dimensions: spatial trajectory deviation and historical state continuity. Reliability during physical execution; The joint difference value is determined by a hardware comparator. Does it exceed the preset differential security threshold? : when At the next bus synchronization clock edge, the calibration instruction sequence in the hardware shadow register area is latched by a hardware latch signal. The data is latched, updated in the physical output register, and then output to the servo driver. when The hardware flow control unit blocks the update enable signal of the physical output register, rolls back the system state pointer to the previous level safety calibration point state stored in the spare area of ​​the hardware shadow register through the hardware state machine, and updates the segmented timing calibration parameters. Piecewise scaling factor Perform adaptive halving shrinkage and re-execute the calibration and verification process from S3 to S4; During the recalibration process, the number of retries is accumulated using a spin safety monitoring counter integrated within the hardware flow control unit; when the retry count reaches a preset hard limit threshold... When this occurs, the hardware state machine forcibly interrupts the recalibration process, locks and resets the robot's current joint command speed to zero, and outputs a hardware-level emergency brake lock signal to the outside.

Images