Methods, equipment and storage media for compensating asymmetric friction forces in rotary joints
By constructing a comprehensive asymmetric friction torque model, decoupling the friction mechanism using independent velocity and power flow symbols, and combining it with temperature response modulation, the asymmetry and temperature influence problems in the friction modeling of rotary joints are solved, thereby improving the torque prediction accuracy and force control smoothness of robot joints.
Patent Information
- Application Number
- CN202611565258.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-10
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies for modeling friction in robot rotary joints suffer from several problems, including the failure to effectively decouple friction asymmetry, inaccurate characterization of friction characteristics across wide speed and temperature ranges, and unstable torque prediction caused by hard switching of the friction model at critical points. These issues result in insufficient trajectory tracking accuracy and force control smoothness.
By constructing a comprehensive asymmetric friction torque model that includes velocity-dependent, load-dependent, and temperature-dependent factors, the friction mechanism is independently decoupled using the velocity direction sign and power flow sign, and the temperature response is modulated by the power factor of the joint angular velocity to achieve feedforward torque compensation.
It improves the torque prediction accuracy, trajectory tracking stability, and force control smoothness of rotary joints under complex working conditions in a wide speed range, wide temperature range, and four quadrants, while reducing torque fluctuations and fatigue damage to transmission components.
Smart Images

Figure CN123103684A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of robotics, and in particular to a method, device and storage medium for compensating asymmetric friction force in a rotary joint. Background Technology
[0002] Accurately constructing a friction model for rotary joints is a core prerequisite for achieving high-precision dynamic modeling, high-sensitivity current and torque sensing, and high-performance force control in robots. In integrated rotary joints of robots, friction exhibits complex nonlinear and asymmetric behavior due to the physical characteristics of reducers, bearings, and lubricating media. The accuracy of the underlying friction model directly determines the trajectory tracking accuracy and force control smoothness of the control system under low-speed reversing and medium-to-high-speed operation. However, related technologies still have the following main defects and shortcomings when modeling and compensating for friction in rotary joints: Over-reliance on the idealized zero-point symmetry assumption: Many related dynamic modeling methods are based on multi-degree-of-freedom serial robotic arms, heavily relying on the simplified assumption that friction is completely symmetrical about the zero point to decouple load torque from friction torque. However, in actual joints, due to physical factors such as the micro-geometric asymmetry of the reducer tooth profile and the directionality of bearing assembly preload, the friction torque generated by forward and reverse motion exhibits significant inherent offsets and asymmetric characteristics. This makes it difficult for traditional symmetrical models to accurately fit the diverse friction behaviors under actual working conditions.
[0003] There is a lack of independent decoupling methods for the asymmetry of different physical mechanisms: even when related technologies attempt to consider asymmetry, they typically use only a single forward and reverse velocity sign to uniformly switch all friction parameters. In actual joints, load friction characteristics are affected not only by the velocity direction but also by the change in power flow direction under motor mode (doing work externally) and generator mode (doing work with external force). Using a single velocity sign cannot accurately characterize the jump in load friction characteristics caused by the change in power flow direction.
[0004] The strong nonlinear coupling effect across wide speed and temperature ranges has not been effectively characterized: traditional modeling methods often assume that the effect of temperature on friction is a linear relationship independent of speed. However, experiments show that under medium- and high-speed conditions, the decrease in friction torque due to temperature increase exhibits a more pronounced nonlinear curvature change with speed. Related technologies lack effective modeling methods for this strong nonlinear coupling effect between speed and temperature.
[0005] Hard switching at critical points can cause jumps and interference in parameter identification: When relevant models directly use sign functions to hard switch parameters near the critical lines where speed or power crosses zero, the torque prediction value may be affected by sensor noise. In addition, the sudden acceleration change at the moment of commutation during parameter identification can introduce additional dynamic components, affecting the stability of parameter estimation.
[0006] Therefore, designing a high-precision rotary joint asymmetric friction modeling and compensation method that can independently decouple the speed and power asymmetric mechanism, accurately characterize the speed-temperature strong nonlinear coupling characteristics, and maintain predictive smoothness under all working conditions is a technical problem that urgently needs to be solved in the field of robot dynamics control. Summary of the Invention
[0007] This application provides a method, device, and storage medium for asymmetric friction compensation of rotary joints, which helps to improve the torque prediction accuracy, trajectory tracking stability, and force control smoothness of rotary joints under complex working conditions in a wide speed range, wide temperature range, and four quadrants.
[0008] On one hand, this application provides a method for compensating for asymmetric frictional forces in a rotary joint, the method comprising: Acquire real-time operating status data of the rotary joint; the real-time operating status data includes joint angular velocity, external effective load torque, and real-time joint temperature; The velocity direction symbol and power flow direction symbol are determined based on the direction of the joint angular velocity and the direction of work done by the external effective load torque; the velocity direction symbol and the power flow direction symbol are independent triggering conditions; Based on the pre-constructed comprehensive asymmetric friction torque model and the real-time operating status data, the predicted value of the comprehensive friction torque at the current moment is calculated; the comprehensive asymmetric friction torque model includes speed-related friction components, load-related friction components, and temperature-related friction components. Specifically, when calculating the speed-related friction component, the corresponding forward / reverse speed friction parameter set is independently switched using the speed direction sign; when calculating the load-related friction component, the corresponding load friction parameter set of the motor / generator mode is independently switched using the power flow direction sign; when calculating the temperature-related friction component, the joint angular velocity is used as the main modulation factor to construct a nonlinear amplified response to the speed sensitivity to temperature. Based on the comprehensive friction torque prediction value, feedforward torque compensation and motion control are performed on the rotary joint.
[0009] On the other hand, this application provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the methods described in the above aspects.
[0010] In another aspect, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the methods described in the above aspects.
[0011] The beneficial effects of the technical solution provided in this application include at least the following: This solution constructs a comprehensive asymmetric friction torque model containing three components: speed-related, load-related, and temperature-related. It utilizes independent velocity direction signs and power flow signs to decouple and switch between two types of asymmetric mechanisms with different physical sources. Furthermore, it modulates the temperature polynomial with the power exponent of the absolute value of the joint angular velocity, thus overcoming the accuracy bottleneck of traditional symmetric assumptions and linear temperature modeling. Based on this, feedforward torque compensation is implemented, which helps to improve the torque prediction accuracy, trajectory tracking stability, and force control smoothness of rotary joints under complex working conditions in a wide speed range, wide temperature range, and four quadrants. Attached Figure Description
[0012] Figure 1 This is a flowchart of the asymmetric friction compensation method for rotary joints provided in the embodiments of this application; Figure 2 This is a detailed algorithm flowchart of the asymmetric friction compensation method for rotary joints provided in the embodiments of this application; Figure 3 This is a flowchart of the calculation of the three-dimensional asymmetric and nonlinear coupled friction model provided in the embodiments of this application; Figure 4 This is a time-domain response curve of the friction torque of the rotary joint under constant temperature and no-load conditions in the embodiments of this application. Figure 5 This is a flowchart of the hysteresis latch and continuous smooth transition control under the critical switching state in the embodiments of this application; Figure 6 This is a diagram showing the asymmetric mapping relationship between the speed and frictional torque of the rotary joint under constant temperature and no-load conditions in the embodiments of this application. Figure 7 This is a diagram showing the state division and friction torque relationship of the rotary joint under load in four quadrant operation in the embodiments of this application; Figure 8 This is a cluster of nonlinear relationship curves between temperature and frictional torque of the rotary joint at different angular velocities in the embodiments of this application; Figure 9 This is a flowchart of the segmented continuous excitation trajectory design and two-stage parameter identification in the embodiments of this application; Figure 10 This is a structural block diagram of a computer device provided in an exemplary embodiment of this application. Detailed Implementation
[0013] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.
[0014] In this article, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.
[0015] In the field of high-precision dynamic control of robots, accurately constructing the friction model of rotary joints is the core prerequisite and key link for achieving high-precision dynamic modeling, high-sensitivity current and torque sensing, and high-performance force control operations (such as gravity compensation, full-arm friction feedforward compensation, and zero-force drag). In an integrated rotary joint, the reducer, bearings, and lubricating medium together constitute a strongly nonlinear and strongly coupled friction system. However, most mainstream modeling techniques are based on the idealized assumption that friction is symmetrical about the zero point of the coordinate system, compressing completely different physical processes such as forward and reverse rotation, and electric and electric generation into the same set of model parameters. This results in a persistent and unconvergent systematic deviation between the model's predicted values and the actual friction behavior of the joint.
[0016] These deviations translate into specific engineering bottlenecks in actual production and applications. For example, when robots perform high-precision assembly, constant-force grinding, or zero-force drag teaching, torque jumps and trajectory crawling frequently occur during reversals. Under medium-to-high-speed, heavy-load, continuous operation, joint temperature rise causes a non-linear decrease in the viscosity of the lubricating medium. Models using linear temperature assumptions cannot follow the attenuation trend of frictional torque, resulting in severe undercompensation or overcompensation. In four-quadrant operation, the change in the force path of the gear meshing surface between motor and generator modes further amplifies the prediction residuals, making it difficult for the positioning accuracy and force control smoothness of the end effector to meet the requirements of high-end manufacturing. If these problems are not effectively resolved, the application of robots in precision manufacturing will continue to be limited to conservative operating conditions of low speed and light load. Frequent torque fluctuations will also accelerate fatigue damage to transmission components, shorten joint life, and increase maintenance costs.
[0017] At its root, the joint exhibits two completely different asymmetric mechanisms: one is the "velocity-direction-related asymmetry" caused by mechanical factors such as the micro-geometric asymmetry of the reducer teeth and the directionality of bearing assembly preload; the other is the "power flow-direction-related asymmetry" caused by the fundamental change in the gear meshing force surface and energy transfer path under motor mode (performing work externally) and generator mode (performing work with external force). The triggering conditions for these two are completely different. For example, although the velocity direction is the same for unloaded forward rotation and loaded forward rotation, the power flow direction and load parameter state are drastically different. Furthermore, under medium- and high-speed operating conditions, the amplitude of the joint angular velocity actually acts as a nonlinear amplifier of the temperature effect. The attenuation of frictional torque caused by temperature increases is drastically amplified, exhibiting a strong nonlinear downward trend. Traditional modeling methods often assume that the effect of temperature on friction is a linear relationship independent of velocity, resulting in severe fitting failures in wide-speed ranges and variable-temperature conditions.
[0018] To address this, this embodiment provides a method for compensating for asymmetric friction in a rotary joint. By independently decoupling and distinguishing the velocity direction and power flow direction through two channels, and by nonlinearly modulating the temperature effect by the power exponent of the velocity amplitude, a high-fidelity comprehensive asymmetric friction torque model is constructed, and accurate feedforward torque compensation is achieved under all operating conditions.
[0019] Figure 1 This is a flowchart of the asymmetric friction compensation method for rotary joints provided in the embodiments of this application, which includes the following steps: Step S1: Acquire real-time operating status data of the rotary joint; the control system and rotary joint of this solution can be humanoid robots, bionic robots, industrial robots, etc., and other systems and mechanical joints. Real-time operating status data includes parameters such as joint angular velocity, external effective load torque, and real-time joint temperature. The system completes the synchronous acquisition and preprocessing of these three types of physical quantities through a hardware sensing layer.
[0020] The joint angular velocity is obtained by a high-resolution encoder mounted at the rear of the motor. The encoder outputs raw position pulses, which are counted by the controller's capture unit and converted into an angular position sequence. The angular velocity is then estimated through differential calculation or an observer algorithm. To ensure the real-time performance and smoothness of the angular velocity signal, the sampling frequency should be no less than an integer multiple of the control loop frequency, and the raw position signal is low-pass filtered to suppress quantization noise and measurement interference. When a transient anomaly occurs in the data of a certain channel (such as sensor spike interference), median filtering or moving average filtering is used for preprocessing to ensure the validity of the input data.
[0021] The external effective load torque can be directly measured by a torque sensor installed at the joint or reducer output end, or it can be obtained through forward calculation based on the rigid body dynamics model of the joint. In the implementation using forward calculation, the controller calls upon pre-identified nominal dynamic parameters such as the link mass, center of mass position, and moment of inertia, and combines them with the current joint position, angular velocity, and angular acceleration to calculate the combined result of gravitational torque and inertial torque in real time, thereby obtaining an estimated load torque value without relying on expensive torque sensors. For a single-link load rotating about the joint axis, the load torque can be expressed as:
[0022] in, For the mass of the connecting rod, It is the acceleration due to gravity. The distance from the load center of mass to the joint rotation center. This refers to the joint angle position. The initial phase offset angle, Let be the moment of inertia of the connecting rod about the joint axis. This refers to joint angular acceleration. This calculation requires high floating-point computing power and is usually performed by a motion control chip or field-programmable gate array (FPGA), with zero bias eliminated through automatic calibration during the initialization phase.
[0023] Real-time joint temperature is measured using temperature sensors built into the joint. Measurement points are located near key heat sources such as the motor stator windings or the inner side of the reducer housing to reflect the true temperature level of the friction-generating area inside the joint. Considering the slow temperature change and thermal hysteresis, the temperature signal undergoes extremum removal and moving average processing before entering the model. If necessary, hysteresis compensation can be performed using a thermal network model to more accurately reflect the thermodynamic state of the friction interface. The sampling frequency of the temperature signal can be lower than that of the speed and torque channels, typically between 1Hz and 10Hz.
[0024] Before entering the subsequent processing flow, the above three types of real-time operating status data need to be synchronized in time and standardized in format. For example, the angular velocity signal, load torque signal, and temperature signal are uniformly converted to the same time base and cached according to a preset data frame format so that subsequent steps can synchronously read the data from each channel. Since the three types of signals come from different sensors and sampling channels, there are inherent sampling delays and clock deviations between the channels. In this embodiment, a unified timestamp is applied to the data of each channel in the data acquisition system, and an interpolation resampling method is used to align the data of each channel to the same time base.
[0025] Step S2: Determine the sign of the velocity direction based on the direction of the joint angular velocity, and determine the sign of the power flow direction based on the work direction of the joint angular velocity and the external effective load torque; See Figure 2The detailed algorithm flowchart of the asymmetric friction compensation method for rotary joints is shown. In this step, two completely decoupled discrimination channels are established, and the velocity direction symbol and the power flow direction symbol are independent triggering conditions.
[0026] The first channel determines the sign of the velocity direction based on the algebraic sign of the joint angular velocity. A positive angular velocity indicates a positive state, and a negative angular velocity indicates a negative state. The physical meaning of this symbol is that, due to mechanical factors such as the micro-geometric asymmetry of the reducer teeth and the directionality of bearing assembly preload, the joint exhibits different frictional characteristics during positive and negative movements. Therefore, it is necessary to independently switch the corresponding velocity friction parameter set using the velocity direction symbol. To avoid state jitter caused by measurement noise near zero, a dead zone or hysteresis mechanism can be introduced in the engineering implementation to keep the symbol state stably latched within the zero-velocity neighborhood. See subsequent implementation details for specific implementation methods.
[0027] The second channel determines the power flow sign based on the sign of the product of the joint angular velocity and the external effective load torque. A product greater than zero indicates that the joint is doing work on the external force and converting electrical energy into mechanical energy, indicating a motor mode; a product less than zero indicates that external force is doing work on the joint and converting mechanical energy into electrical energy, indicating a generator mode. The physical meaning of this symbol is that the force direction and energy transfer path on the gear meshing surface fundamentally change between motor and generator modes, leading to significant differences in load-related friction characteristics. Therefore, it is necessary to independently switch the corresponding load friction parameter set using the power flow symbol.
[0028] It is important to emphasize that the triggering conditions of the two symbols are independent and do not interfere with each other. Even if the joint maintains the same direction of rotation, the power flow direction symbol will still flip when the load nature changes from resisting motion to assisting motion; conversely, the velocity direction can also be reversed when the power flow direction remains unchanged. This independent decoupling design stems from the distinction between the physical nature of friction—velocity-related asymmetry is caused by the micro-geometry of the reducer tooth profile and the directionality of the bearing assembly preload, while power flow-related asymmetry is caused by the fundamental change in the force direction of the gear meshing surface and the energy transfer path in motor and generator modes. The two have completely different physical origins and cannot be uniformly managed by a single forward / reverse symbol. For example, although the velocity direction is the same in no-load forward rotation and in loaded forward rotation, the power flow direction may be different, and the parameter states of the load friction term are also completely different. Therefore, it is impossible to uniformly manage the two asymmetric mechanisms using a single forward / reverse symbol; decoupling modeling must be achieved through independent switching of the two symbols; dual-channel discrimination provides the logical basis for the subsequent precise switching of friction parameters.
[0029] Step S3: Based on the pre-built integrated asymmetric friction torque model and real-time operating status data, calculate the predicted value of the integrated friction torque at the current moment; The comprehensive asymmetric friction torque model includes velocity-dependent friction components, load-dependent friction components, and temperature-dependent friction components. These three components are physically independent but are superimposed in real time during computation.
[0030] in, For velocity-dependent friction components, For load-related friction components, This represents the temperature-dependent friction component.
[0031] One is the velocity-dependent friction component, which is used to describe velocity-dominated friction behaviors such as viscous shear, Coulomb friction, and low-speed Stribeck effect. During the calculation, the corresponding positive or negative velocity friction parameter set is switched independently by using the velocity direction sign, so that the forward and reverse rotations each have a complete set of parameters. The symmetry assumption is abandoned at the model structure level, thereby fitting the difference in forward and reverse friction caused by the inherent deviation of the machine.
[0032] The second is the load-related friction component, which describes the additional friction introduced by the external load through gear contact pressure transmission. During calculation, the corresponding motor mode or generator mode load friction parameter set is independently switched using the power flow direction symbol to capture the friction gain and friction base jump caused by the change in energy transmission path.
[0033] Thirdly, there is the temperature-dependent friction component, which is used to compensate for the influence of temperature changes on the viscosity and material properties of the lubricating medium. During the calculation, the joint angular velocity (specifically in the form of a power exponent of the absolute value) is used as the main modulation factor of the temperature polynomial to construct a nonlinear amplified response of the speed to temperature sensitivity. That is, at low speeds, the temperature effect is approximately linear, while at medium and high speeds, the frictional attenuation caused by temperature rise is amplified sharply by the velocity field, thereby accurately characterizing the thermally induced frictional softening characteristics in a wide speed range.
[0034] The three components are superimposed in real time in the controller according to the control cycle, and the comprehensive friction torque prediction value at the current moment is output. The parameter sets of each component can be pre-stored in the system memory and quickly recalled through symbolic state index. All calculations can be completed within a single control cycle, meeting the real-time requirements. The specific mathematical expression and physical mechanism of each component are described in the following embodiments.
[0035] Step S4: Based on the comprehensive friction torque prediction value, perform feedforward torque compensation and motion control on the rotary joint.
[0036] In this step, the predicted value of the combined friction torque is injected as a feedforward control quantity into the current loop or torque loop of the joint servo driver, and algebraically superimposed with the output command of the feedback controller. During joint motion control, the controller calculates the basic control torque based on the desired trajectory, the feedback controller is responsible for handling position and velocity errors, and the friction feedforward term directly cancels the predicted friction resistance, thereby significantly reducing the burden of feedback adjustment and improving the system response bandwidth and tracking accuracy.
[0037] In engineering implementation, amplitude limiting protection must be applied to the feedforward command to constrain its amplitude within a safe proportion of the motor's rated peak torque, preventing excessive compensation torque due to abnormal parameters or sensor malfunctions. At the moment of sign state switching, ramp limiting or low-pass smoothing can be applied to the predicted value to avoid impacting the current loop due to step jumps in the torque command. After feedforward compensation, the equivalent frictional interference of the joint is significantly offset, the crawling phenomenon in low-speed trajectory tracking is significantly eliminated, and the smoothness of high-speed operation and force control interaction is improved, laying the underlying foundation for system-level applications such as gravity compensation and zero-force drag.
[0038] In summary, this scheme constructs a comprehensive asymmetric friction torque model comprising three components: velocity-dependent, load-dependent, and temperature-dependent. It utilizes independent velocity direction and power flow signs to decouple and switch between two asymmetric mechanisms with different physical origins. Furthermore, it modulates the temperature polynomial using the power exponent of the absolute value of the joint angular velocity, thus overcoming the accuracy bottleneck of traditional symmetric assumptions and linear temperature modeling. Based on this, implementing feedforward torque compensation helps improve the torque prediction accuracy, trajectory tracking stability, and force control smoothness of rotary joints under complex conditions in a wide velocity range, wide temperature range, and four quadrants.
[0039] When the compensation framework given in the aforementioned embodiments is put into real-time operation, the robustness of the sign discrimination channel becomes a key weakness in engineering implementation. In actual systems, encoder quantization noise, electromagnetic interference, and mechanical vibration cause high-frequency jitter in the joint angular velocity and external effective load torque signals. If hard switching of the sign function is directly used at the zero-crossing point of velocity and the zero-crossing line of power, the sign state will repeatedly flip at high frequency near the critical value, causing step jumps and sawtooth oscillations in the predicted friction torque value, thus disrupting the stability of the current loop.
[0040] Therefore, this embodiment provides a hysteresis latching and continuous smooth transition control method under critical switching states, see [link to relevant documentation]. Figure 5 As shown, it includes two parts: a speed threshold hysteresis latch mechanism within the zero-speed band and a power confidence smooth transition mechanism near the zero-power line.
[0041] I. Speed Threshold Hysteresis Latch Mechanism within Zero Speed Zone The velocity direction symbol has both positive and negative state markings. See also Figure 4The time-domain curve of friction torque in constant-speed reciprocating motion shown ( Figure 4 (a) is the low-speed operating condition. Figure 4 (b) Under high-speed conditions, the friction torque exhibits a vertical step jump during reversal and fluctuates in a sawtooth pattern at extremely low speeds. This is because direct judgment based on the sign of the angular velocity is unstable in the zero-value neighborhood. Therefore, this solution presets a minimum / maximum speed threshold. To construct the velocity threshold hysteresis interval, The typical range of values is to Construct a width of [missing information] on both sides of zero velocity. hysteresis interval .
[0042] After the encoder pulses are counted by the capture unit, the raw angular velocity is obtained using the M / T method or differential operation. It is first filtered by a first-order low-pass filter to suppress high-frequency quantization noise. Hysteresis detection is implemented by a finite state machine with state memory, containing three branches: forward latch, reverse latch, and state hold. State variables and parameter thresholds reside in the controller memory and share an index bus with the speed and friction parameter set lookup table. The state machine transition process is as follows: When the absolute value of the joint angular velocity rises from zero and exceeds the positive minimum velocity threshold, i.e., angular velocity... Rise from zero and strictly exceed At that time, the velocity direction sign is switched to the positive state flag and latched; When the joint angular velocity from the opposite direction causes its absolute value to exceed the minimum velocity threshold in the negative direction, i.e., the angular velocity... The reversal direction causes the absolute value to exceed At that time, the velocity direction sign is switched to the reverse state flag and latched; When the joint angular velocity is in the velocity threshold hysteresis range During this period, the sign of the previously valid velocity direction remains unchanged.
[0043] Therefore, as long as the small drop caused by noise does not reach the opposite threshold, the latched state will not be broken, and the parameter set will remain stable. The definition of the insensitive zone for direction discrimination is as follows: if it is too small, noise filtering will be insufficient; if it is too large, the low-speed commutation response will be sluggish. This occurs when gravity causes extremely slow drift and prolonged stagnation. When the fault is near, the latching mechanism maintains parameter stability. In case of an anomaly, if the speed signal is lost or constantly abnormal, the state-holding branch maintains the state in the direction before the fault; instantaneous spikes are unlikely to continuously cross the entire hysteresis interval, and this mechanism is naturally anti-spike.
[0044] II. Power Confidence Smoothing Transition Mechanism Near Zero Power Line The load factor of the load-related friction component switches between two sets of values, one for motor mode and one for generator mode, according to the power flow direction. When the instantaneous power... When the load is close to zero, even minute measurement noise can cause the power flow sign to repeatedly flip. Therefore, a power confidence factor is used to smoothly mix the two sets of load factors.
[0045] Power confidence factor The calculation formula is:
[0046] For any load factor Its smooth blending value is:
[0047] In the formula, A coefficient used to control the slope of the transition zone; The preset power threshold; and The first The corresponding load factor in the power quadrant; Values are taken for the power quadrant (values 1-4). For the corresponding coefficient, this embodiment can take... .
[0048] In the project, the controller stores two sets of load factors simultaneously, and each control cycle is based on... Dynamically synthesize the current parameter value; Functions can be obtained using a lookup table. Defining the boundary of the hybrid transition zone: when hour, When the output tends to 1 or 0, the hybrid formula degenerates into a hard switch to ensure model accuracy under steady-state conditions; when hour, The parameters vary continuously between 0 and 1, and the two sets of parameters are linearly interpolated according to the confidence level. This mechanism extends the motor mode and generator mode into a continuous confidence state space, making the derivatives of the load-related friction components continuously bounded. In case of anomalies, the limiting logic will... Constraints should be kept within safe numerical ranges to prevent overflow during division by zero; The saturation properties naturally Limited to Within the interval, even if the input signal changes abruptly, the mixing coefficients remain valid, avoiding computational crashes caused by parameter overflow.
[0049] In summary, this embodiment eliminates high-frequency flipping of the speed direction sign through hysteresis latching and achieves soft switching between motor mode and generator mode through power confidence hybridization, relying solely on... , With slope coefficient Three engineering parameters can be adjusted to adapt to different hardware platforms and mechanical joints.
[0050] Within the aforementioned compensation framework, the upper limit of torque prediction accuracy is determined by the mathematical structure of the integrated asymmetric friction torque model. Traditional models suffer from three significant drawbacks in this area: first, they use a single symbolic function to uniformly switch all friction parameters, failing to distinguish between velocity asymmetry caused by mechanical geometry and power asymmetry caused by changes in the energy transfer path; second, load friction is simplified to a single linear proportional relationship, failing to describe the coupling law of the meshing surface contact state changing with the velocity and power quadrants; and third, the temperature term is assumed to be a linear relationship independent of velocity, failing to explain the phenomenon of frictional force rapidly decreasing with temperature rise at medium to high speeds. Therefore, this embodiment provides analytical expressions for the three components—velocity, load, and temperature—and their engineering calculation logic, see [link to relevant documentation]. Figure 3 As shown.
[0051] In this embodiment, the comprehensive friction torque prediction value is designed to be composed of three superimposed components, namely... The design of each component is as follows: To avoid ambiguity of symbols, in this specification This represents the equivalent frictional compensation torque used for feedforward compensation; the actual physical frictional resistance torque is in the opposite direction to the corresponding compensation torque.
[0052] I. Based on the sign of velocity direction velocity-related friction components Velocity-related friction components of The calculation formula is:
[0053] In the formula, Joint angular velocity; The sign for the direction of velocity; 、 、 The coefficient of viscous friction is given under different velocity directions. The coefficient of static friction; Coulomb friction coefficient; and Stribeck characteristic velocity; and For shape factor; It is a symbolic function.
[0054] See Figure 6 The asymmetric mapping curve of friction torque versus speed under constant temperature and no-load conditions is shown. The physical mechanism of this formula is reflected in the following four aspects: First, the complete independence of the forward and reverse parameter sets: superscript By giving each of the forward and reverse motions a complete set of viscous, static friction, Coulomb and Stribeck parameters, the model mathematically fits the inherent offset caused by the micro-geometric asymmetry of the reducer tooth profile and the bearing preload directionality, enabling the model to directly characterize the phenomenon that the forward and reverse steady-state friction torque amplitudes are unequal and not completely mirror images of the origin.
[0055] Second, the direction of the second-order viscosity term remains unchanged: (Equation follows) Equivalent to , and cubic terms As an odd function of velocity, it ensures that the frictional force reverses its sign with respect to velocity while absorbing the nonlinear growth of residuals caused by lubricant shear thinning and other effects in the medium-to-high speed region, thus overcoming the limitations of traditional linear viscous friction models.
[0056] Third, static friction and Coulomb friction transition using independent characteristic velocities and shape factors: and The two items respectively 、 For their respective characteristic velocities, with 、 Each is a shape factor, which independently describes the decay process of static friction and the establishment process of Coulomb friction, thus enabling the reproduction of the Stribeck local concave effect in the low-speed region where the friction torque decreases with a slight increase in speed.
[0057] Fourth, the physical constraints of the shape factor: shape factor 、 The steepness of the transition curve is controlled. A value below 1 will make the curve too flat to describe pre-slip instability, while a value above 2 will be close to a step function and not in line with the actual physical system. Therefore, in parameter identification, it is constrained to the range of 1 to 2.
[0058] At the engineering calculation level, the velocity signal can first be low-pass filtered and a small dead zone set; the forward and reverse parameter sets are indexed and retrieved according to the currently latched velocity direction sign. The model contains multiple exponentiation / power operations, which can be implemented by table lookup or numerical approximation depending on the controller's computing power, and is not limited to "calculating the exponential function only once per cycle".
[0059] II. Based on power flow direction symbol Load-related friction components Load-related friction components The calculation formula is:
[0060] in, Joint angular velocity; External effective load torque; Symbol for power flow direction; to These are the load model coefficients for the first to fourth power quadrants.
[0061] The specific rules for determining the power flow direction sign are as follows: when When it is determined that the rotary joint is performing work, the power flow direction symbol switches to the motor mode; when When the external force does work on the rotating joint, the power flow sign switches to generator mode. When the product approaches zero, the load-related friction term itself... and The factor approaches zero. The precise attribution of the power mode has minimal impact on the model output. In engineering practice, the previous effective power mode can be kept unchanged, or the power confidence factor can be used for smooth mixing.
[0062] See Figure 7 The four-quadrant operating framework and load dependency curve shown are as follows: Figure 7 (a) is a schematic diagram of the four-quadrant operation of a rotary joint; Figure 7 (b) A graph showing the mapping relationship between load torque and friction torque under different velocity directions. This formula can be decomposed into two parts: Part One This represents the load friction gain term that varies with velocity. Wherein, As a constant gain basis, it characterizes the velocity-independent rigid component in load friction; It is an exponentially decaying term. The characteristic velocity of this term is used to characterize the dynamic process at low speeds where the interaction between micro-protrusions on the contact surface is significant, and the additional effect gradually diminishes after the oil film forms as the speed increases. This results in different additional frictions introduced by the same load at different speeds. The existence of this exponential term allows the model to capture the slight Stribeck transition effect observed in the low-speed region when the load changes, i.e., the load friction gain increases slightly at very low speeds and gradually stabilizes as the speed increases.
[0063] Part Two, This represents the nonlinear term of the load amplitude. Wherein, Used to describe nonlinear accelerated growth under heavy load; Describes the magnitude term of a first-order term. Multiply by the end. This causes the equivalent friction compensation torque to reverse synchronously with the direction of motion; the corresponding actual physical friction resistance torque is in the opposite direction.
[0064] Regarding the physical constraints for parameter identification, and It must be a positive value to ensure that the load friction increases monotonically with the increase of the load torque; It can be positive or negative to flexibly fit the direction of the low-speed additional effect; It must be a positive value, and its range of values can be limited to 1 / 2. to Between these, to ensure that the characteristic velocity of the exponential decay term falls within a physically reasonable range.
[0065] III. Based on the speed power exponent Modulated temperature-dependent friction component The temperature-related friction component The calculation formula is:
[0066] in, Real-time joint temperature; The main modulation factor of temperature effect ( ); to These are the coefficients of the temperature polynomial; Used to determine the direction of the compensation torque.
[0067] See Figure 8 The set of friction torque curves shown represents different constant speeds within a temperature range of 20°C to 40°C. The formula uses the power of the speed amplitude. As the master modulation factor of the entire temperature polynomial, it is the core structure characterizing the nonlinear amplification of the velocity-temperature sensitivity. When the modulation term increases sublinearly with speed, when It grows linearly over time; when It exhibits superlinear growth over time. In this embodiment... This is a global constant; the specific parameters should be selected based on the actual scenario.
[0068] The temperature polynomial within parentheses describes the curve shape of friction torque as a function of temperature at a specific speed amplitude. At low speeds... With relatively small values, higher-order terms contribute limitedly, and the model exhibits approximately linear temperature drift; at medium to high speeds... The value increases significantly, higher-order terms are amplified, and a strongly nonlinear downward curve is observed. When When approaching zero, As the velocity approaches zero, the temperature term automatically decays to avoid unreasonable influences in the zero-speed region. At extremely low speeds, the model degenerates into one dominated by velocity-dependent friction components; at high speeds, temperature-dependent friction components become an important part of the total friction torque.
[0069] In engineering calculations, exponentiation can be implemented using lookup tables or Taylor series expansion; when When the temperature approaches zero, a lower limit is set for the base value to ensure that the output of the temperature term in a static state is zero; the temperature signal must undergo thermal hysteresis compensation before entering the formula. Regarding fault tolerance: when the temperature sensor malfunctions or the signal is missing, the temperature-related friction component is automatically zeroed, and the model is downgraded to the sum of the velocity and load terms; when the temperature reading exceeds a reasonable range, the previous valid temperature is used; if there are continuous abnormalities, the temperature is zeroed and an alarm is triggered; when the velocity signal is subject to noise interference, [further measures are taken]. Low-pass filtering before participation Calculations are performed in real time even during transient temperature changes, and transient deviations are eliminated after thermal steady-state conditions are reached.
[0070] This embodiment decouples mechanical geometric asymmetry from energy transfer path asymmetry by using independent switching of dual symbols and speed power modulation, accurately characterizing the slope, curvature, and direction characteristics of four-quadrant load friction, and solving the problem of linear temperature modeling failure in a wide speed domain.
[0071] When the joint angular velocity is extremely low and near reversal, the friction interface has not yet fully entered the sliding state, and the contact micro-protrusions are in the pre-sliding stage. The friction torque depends on the displacement history and exhibits hysteresis behavior. The aforementioned three components are all static mapping models, which are difficult to characterize the micro-hysteresis at this stage. Therefore, the comprehensive asymmetric friction torque model also includes a low-speed hysteresis friction component. The calculation formula is as follows:
[0072]
[0073] In the formula, This is the hysteresis stiffness coefficient; These are internally hysteretic state variables; for The derivative with respect to time; This is a relaxation function related to velocity.
[0074] The low-speed hysteresis friction component takes effect when the absolute value of the joint angular velocity is below a preset activation threshold. In engineering implementation, internal state variables... The differential equation is updated discretely and recursively within each control cycle, with the initial value taken as the final value of the previous cycle; when the absolute value of the angular velocity is lower than the activation threshold, The value is superimposed on the overall predicted friction torque; when it exceeds the activation threshold, Set the state variables to zero and stop updating them. Relaxation function. A positive function that monotonically increases with speed can be adopted, reflecting that the higher the speed, the easier it is for the internal state to be "washed out" and the weaker the hysteresis effect. This mechanism can suppress torque fluctuations and trajectory crawling during extremely low-speed reversal, and improve the smoothness of zero-force drag and micro-motion positioning.
[0075] Furthermore, considering that grease aging, wear accumulation, and long-term temperature drift can cause some parameters to change slowly, adjusting all parameters online could lead to parameter drift divergence and model structure damage. Therefore, during feedforward torque compensation operation, real-time acquired torque feedback data is used, employing algorithms such as recursive least squares with a forgetting factor or extended Kalman filtering, to fine-tune the slowly drifting viscous friction coefficient and temperature polynomial constants online, while keeping the shape factor and the velocity modulation exponent of the temperature effect locked constant.
[0076] Online updates only apply to the coefficient of viscous friction. 、 、 With temperature polynomial constant term Slow-drift parameters can be updated at a period shorter than the control period to reduce computational overhead and avoid coupling with fast dynamic processes. Shape factor 、 With speed modulation index The nonlinear morphology of the model is determined by the fact that online adjustments can easily lead to model structural distortion and identification divergence; therefore, the parameter is locked. When the torque feedback excitation is insufficient, the system pauses online updates and maintains the previous effective parameters. This mechanism adaptively compensates for the slow parameter drift caused by grease aging and wear without changing the model structure.
[0077] The comprehensive asymmetric friction torque model includes multiple parameters such as speed, load, and temperature, resulting in a high-dimensional parameter space. Conventional excitation trajectories may introduce additional dynamic components near the commutation point; directly optimizing all parameters together can easily lead to results influenced by initial values and parameter coupling. Therefore, this embodiment employs a parameter acquisition method combining a smooth excitation trajectory with step-by-step identification, such as... Figure 9 As shown, the specific steps include the following: I. Piecewise Continuous Excitation Trajectory for Eliminating Abrupt Acceleration Changes When acquiring experimental data, the rotary joint was controlled to operate according to a speed-motion mode, and a commutation segment with continuous deceleration to zero was set at the end of each unidirectional motion segment. During the deceleration and commutation time... The internal velocity is continuously varied to eliminate disturbances caused by sudden acceleration changes. See also Figure 4 The time-domain curve of friction torque in constant-speed reciprocating motion, as shown, exhibits a drastic vertical step jump in friction torque during the forward and reverse switching of joint speeds in a conventional constant-speed reciprocating motion trajectory. This jump originates from the high-frequency dynamic component introduced by the sudden acceleration change at the moment of reversal. Directly using the data from this time period for parameter identification would severely interfere with the stable extraction of parameters in the low-speed region of the Stribeck curve. Therefore, this embodiment eliminates this interference at its source at the trajectory design level.
[0078] Within the unidirectional motion segment, the target angular velocity of the rotary joint Represented as:
[0079] in, The duration of a unidirectional motion segment. To reduce commutation time, The trajectory coefficients satisfy the boundary continuity constraints:
[0080]
[0081] The first constraint mentioned above guarantees that in At time; the second constraint guarantees that at At time t, the velocity at the end of the commutation segment is zero. Given the desired steady-state velocity... Trajectory Period and commutation time Then, the trajectory coefficients can be uniquely determined. and The value of . Among them, the constant angular velocity in the steady segment. As a signed quantity, adjacent periods take opposite signs, forming a reciprocating motion in both directions.
[0082] Before the reverse motion begins, a symmetrical continuous acceleration segment is set up to allow the speed to transition continuously from 0 to the reverse steady-state speed, avoiding the situation where adjacent unidirectional motion segments are simply spliced together, resulting in a speed of 0 to 1. The instantaneous jump. Data from constant velocity or low acceleration ranges is preferred for parameter identification.
[0083] The trajectory operates at a constant angular velocity for most of the time, with zero acceleration, and the joint is in a steady-state friction state. During the commutation phase, the angular velocity changes linearly with time, and the time derivative is a finite constant. This avoids the introduction of infinite acceleration impacts from velocity step jumps, suppressing flexible joint resonance at the excitation source and providing high signal-to-noise ratio data for parameter identification. (Deceleration commutation time) The selection should consider both identification accuracy and experimental efficiency. Position, velocity, acceleration, temperature, and current / torque information are recorded synchronously during trajectory operation. Data timestamps from each channel are aligned, and only data from the velocity stability segment is used as valid identification samples.
[0084] II. Parameter Acquisition of Velocity-Related Friction Components Initial values for velocity-dependent friction component parameters were identified using experimental data under no-load and isothermal conditions. The joint velocity was set to... to Multiple constant gradients control the reciprocating motion of the joint. Under no-load conditions, the motor output torque is entirely used to overcome frictional resistance, allowing direct extraction of frictional torque information. Experiments show that the forward and reverse curves are not perfectly mirror images of the origin, exhibiting non-negligible asymmetric characteristics; the low-speed region displays Stribeck variation characteristics, while the medium-to-high-speed region shows a non-linear growth trend. Therefore, independent velocity-friction parameter sets are established for forward and reverse motion, with the velocity direction sign switched. The collected frictional torque samples are fitted with the predicted values of the velocity-related friction component model to obtain initial values of the velocity-related friction parameters.
[0085] III. Parameter Acquisition of Load-Related Friction Components The initial values of the load-related friction component parameters are identified using experimental data under load conditions with a fixed velocity-dependent friction component. One existing experimental condition involves connecting a link with an effective load of 10 kg to the joint's output flange. The load torque is changed by adjusting the lever arm of the center of mass relative to the joint axis. The joint reciprocates within a range of ±180° at a constant angular velocity of 0.5° / s at approximately 27°C. A no-load control is performed under the same conditions. The load torque and the velocity / temperature-dependent friction component are subtracted from the total joint torque to separate the load-related friction component.
[0086] During the load test, the joint output torque Includes load torque Speed-related friction torque Temperature-dependent frictional torque and load-related frictional torque The temperature remained constant at 27℃. It can be calculated using a pre-calibrated temperature model. Measured values of load-related friction components:
[0087] Because the experimental temperature is constant, This is a constant offset, which can be eliminated using the unloaded baseline. Experiments have shown that the load-related friction component increases with... The changes exhibit different slopes and curvatures under different power modes, and are not a simple linear relationship: the rate of increase is relatively gentle in motor mode, while the rate of increase is significantly increased and the curve is curved in generator mode. During parameter identification, [the following will be considered]... Compared with model predictions Construct the residual sum of squares cost function and solve it using the nonlinear least squares method. to The data for motor mode and generator mode are naturally separated, and the two sets of parameters can be identified independently.
[0088] IV. Parameter Acquisition of Temperature-Related Friction Components By fixing the aforementioned two components, the initial values of the parameters of the temperature-dependent friction component are identified using experimental data under varying temperature conditions. See [link to relevant documentation]. Figure 8 As shown, the temperature dependence experiment in this embodiment was conducted at an ambient temperature of approximately 17°C. The joint temperature was raised to 20°C as the starting point, and temperature points were selected at intervals of 2°C to 5°C within the temperature range of 20°C to 40°C. At each temperature point, a low-speed... to medium speed to ,high speed to Drive the joint to obtain a data set of friction torque variations with temperature at different speeds.
[0089] All data were collected under no-external-load conditions, with the motor output torque entirely used to overcome internal frictional resistance. The temperature sensor was installed at a location that characterizes the actual temperature of the lubricating medium. At each target temperature point, the joint was run unloaded at a set speed for 10 to 15 minutes until the temperature fluctuation was less than [a certain value]. Valid data was then collected. During parameter identification, the measured friction torque, after deducting the determined velocity-related component, can be used as an observation sample for the temperature-related component, and the nonlinear least squares method can be used for estimation. and to This embodiment does not include experimental values such as fixed heat preservation time and temperature fluctuation threshold that are not given in existing materials.
[0090] V. Joint Optimization and Optional Consistency Check The first stage is step-by-step initial value identification: first, initial values of velocity-related friction component parameters are identified using no-load isothermal data; then, with the velocity-related friction components fixed, initial values of load-related friction component parameters are identified using loaded data; finally, with the first two parameters fixed, initial values of temperature-related friction component parameters are identified using variable-temperature data. Step-by-step identification is used to provide physically reasonable initial iteration points for subsequent joint optimization.
[0091] The second stage is joint optimization. Using all initial parameter values from the first stage as initial iteration points, a cost function is constructed based on the sum of squared residuals between the measured torque and the model-predicted torque. This cost function is then solved uniformly using the nonlinear least squares method for all friction parameters. Multiple sets of reasonable initial values can be used for repeated solutions, and the parameter estimation results that satisfy the physical constraints and have a smaller sum of squared residuals are selected. For example, the parameter vector to be estimated is... The optimization objective is:
[0092] in, This is a sample vector of friction torque obtained from the collected data. To synthesize the asymmetric friction torque model in the parameter vector The predicted value vector is calculated. In the optimization, the Stribeck shape factor is constrained to a range of 1 to 2, and multiple sets of initial values are set within a reasonable parameter range for repeated solving. The set with the smallest sum of squared residuals and physically reasonable parameters is selected as the final identification result.
[0093] Optionally, data not involved in the identification can be used to perform a consistency check on the model to examine its applicability under different operating trajectories. The specific evaluation method can be selected according to the actual experimental conditions.
[0094] This embodiment reduces the coupling between different parameters and minimizes the impact of commutation dynamics on parameter estimation through a process of "smooth trajectory excitation - step-by-step initial value identification - joint optimization".
[0095] Figure 10 A structural block diagram of a computer device provided in an exemplary embodiment of this application is shown. The device may include a processor and memory, which may be connected via a bus or other means. The processor may be a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a graphics processing unit (GPU), a neural network processing unit (NPU), or other programmable logic devices and combinations thereof.
[0096] This application also discloses a computer-readable storage medium for storing a computer program. When the computer program is executed by a processor, it implements all or part of the processes described in the above method embodiments. The storage medium may include read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), solid-state drive (SSD), optical disk, or other readable storage media.
[0097] This specific embodiment is only used to illustrate this application and does not constitute a limitation on the scope of protection; the scope of protection is determined by the claims.
Claims
1. A method for compensating asymmetric frictional forces in a rotary joint, characterized in that, The method includes: Acquire real-time operating status data of the rotary joint; the real-time operating status data includes joint angular velocity, external effective load torque, and real-time joint temperature; The velocity direction symbol and power flow direction symbol are determined based on the direction of the joint angular velocity and the direction of work done by the external effective load torque; the velocity direction symbol and the power flow direction symbol are independent triggering conditions; Based on the pre-constructed comprehensive asymmetric friction torque model and the real-time operating status data, the predicted value of the comprehensive friction torque at the current moment is calculated; the comprehensive asymmetric friction torque model includes speed-related friction components, load-related friction components, and temperature-related friction components. Specifically, when calculating the speed-related friction component, the corresponding forward / reverse speed friction parameter set is independently switched using the speed direction sign; when calculating the load-related friction component, the corresponding load friction parameter set of the motor / generator mode is independently switched using the power flow direction sign; when calculating the temperature-related friction component, the joint angular velocity is used as the main modulation factor to construct a nonlinear amplified response to the speed sensitivity to temperature. Based on the comprehensive friction torque prediction value, feedforward torque compensation and motion control are performed on the rotary joint.
2. The method according to claim 1, characterized in that, The velocity direction symbol has a positive state mark and a negative state mark; determining the velocity direction symbol based on the direction of the joint angular velocity includes: Preset speed threshold ,by and Constructing the hysteresis interval for velocity direction determination ; When the joint angular velocity When this occurs, the velocity direction symbol is switched to a positive state marker and latched. When the joint angular velocity At that time, the velocity direction symbol is switched to a reverse state marker and latched; when At that time, the sign of the previously valid velocity direction remains unchanged.
3. The method according to claim 1, characterized in that, The velocity-related friction component The calculation formula is: in, The joint angular velocity; The sign for the velocity direction; , , The coefficient of viscous friction is given under different velocity directions. The coefficient of static friction; Coulomb friction coefficient; and Stribeck characteristic velocity; and For shape factor; It is a symbolic function.
4. The method according to claim 1, characterized in that, The load-related friction component The calculation formula is: in, The joint angular velocity; The external effective load torque; The symbol for the power flow direction; to These are the load model coefficients for different power quadrants; The rule for determining the power flow direction sign is: when When it is determined that the rotary joint is performing work, the power flow direction symbol switches to the motor mode; when When it is determined that an external force is doing work on the rotating joint, the power flow direction symbol is switched to the generator mode.
5. The method according to claim 4, characterized in that, When calculating the load-related friction component, for operating conditions near the zero power line, the load coefficients of the motor mode and the generator mode are smoothly mixed based on the power confidence factor: The power confidence factor The calculation formula is: Regarding the load factor Its smooth blending value is: in, A coefficient used to control the slope of the transition zone; The preset power threshold; and The number is respectively the number. The corresponding load factors for the motor mode and the generator mode in the power quadrant; For the corresponding coefficient.
6. The method according to claim 1, characterized in that, The temperature-related friction component The calculation formula is: in, The real-time temperature of the joint; The main modulation factor for the temperature effect; to These are the coefficients of the temperature polynomial.
7. The method according to claim 1, characterized in that, The friction parameters in the comprehensive asymmetric friction torque model are obtained through a combination of offline step-by-step identification and joint optimization, including: The first stage involves identifying the initial parameter values of the velocity-related friction component using experimental data under no-load and constant-temperature conditions; fixing the velocity-related friction component and identifying the initial parameter values of the load-related friction component using experimental data under loaded conditions; fixing the aforementioned two components and identifying the initial parameter values of the temperature-related friction component using experimental data under variable-temperature conditions. The second stage involves using all the initial parameter values obtained in the first stage as initial iteration points to construct a cost function based on the sum of squared residuals between the measured torque and the model-predicted torque, and then using the nonlinear least squares method to jointly optimize all friction parameters.
8. The method according to claim 7, characterized in that, When acquiring the experimental data, the rotary joint is controlled to operate according to the speed motion mode, and a commutation segment that continuously decelerates to zero is set at the end of each unidirectional motion segment. Within the unidirectional motion segment, the target angular velocity of the rotary joint Represented as: in, The duration of the unidirectional motion segment is given by the following formula: To reduce commutation time, The trajectory coefficients satisfy the boundary continuity constraint condition: 。 9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 8.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 8.
