Robot external force detection method, system, device and medium based on double-layer observer

Abstract

The application discloses a robot external force detection method and system based on double-layer observers, equipment and a medium, relates to the technical field of robot detection, and first adopts Newton-Euler method to establish a robot dynamics model, and identifies dynamics parameters through an excitation trajectory and a least square method; then two generalized proportional integral observers with the same structure but different inputs are constructed for each joint, external force detection is carried out based on the observers; and the state estimation values of the two layers of observers are subtracted, and the external torque estimation value is inversely solved in combination with an observer dynamic equation. The application aims at the problem that the existing dynamics modeling error of the robot itself interferes with robot external force detection, causes inaccurate or ineffective external force detection, adopts double-layer observers to simultaneously observe, eliminates internal interference caused by the dynamics model error through the difference between the two layers of observation values, obtains the required external torque value, and simultaneously utilizes the high-order observation characteristics of the momentum generalized proportional observer itself, so that high-precision robot external force detection can be realized.

Description

Technical Field

[0001] This invention relates to the field of robot inspection technology, and in particular to a robot external force detection method, system, equipment and medium based on a dual-layer observer. Background Technology

[0002] As robotics technology accelerates its evolution towards intelligence and flexibility, its applications have expanded from traditional large-scale industrial manufacturing to diverse fields such as healthcare, warehousing and logistics, service interaction, and specialized operations. With human-machine collaboration and intelligent interaction becoming core trends in high-quality industry development, the probability of robots unexpectedly coming into contact with humans, surrounding equipment, or complex and uncertain environments during daily operations has increased significantly. Therefore, the importance of external force detection technology is increasingly prominent. It not only directly relates to the safety and performance of human-machine collaborative operations but is also a key supporting technology for robots to efficiently adapt to diverse and complex scenarios and achieve high-precision operations.

[0003] Currently, force / torque sensors, as a traditional means of external force detection, can directly acquire force information during contact processes, but they still face multiple constraints in practical engineering applications. First, their manufacturing cost is high, making it difficult to achieve widespread deployment in large-scale robot products. Second, the sensor's load adaptability range is limited, making it difficult to meet the operational needs of different types of robots under various working conditions. Third, the additional elastic deformation introduced after sensor installation can disturb the robot's dynamic characteristics, thereby affecting control accuracy. More importantly, in complex and variable actual working environments such as industrial sites and outdoors, factors such as high temperature, dust, vibration, and electromagnetic interference can easily cause sensor reading drift, decreased stability, or even data distortion, making it difficult to meet the stringent requirements of dynamic and high-precision external force sensing.

[0004] To overcome the aforementioned bottlenecks, researchers have been actively exploring external force detection methods based on robot dynamics models in recent years, without relying on external physical sensors. For example, patent application CN113459160B discloses a robot collision detection method based on a second-order generalized momentum observer. This method constructs a second-order observer architecture based on the traditional first-order generalized momentum observer and adds a PD regulator in series to increase the adjustable parameter dimension of the system, effectively reducing detection latency and improving collision detection sensitivity. Another example is patent application CN115488895A, which discloses a collaborative robot collision detection scheme based on a momentum observer. It optimizes the inverse kinematics solution by combining generalized inverse kinematics with Newton's downhill method and introduces feedforward compensation and variable damping design into the observer, significantly improving collision detection speed and robustness.

[0005] However, the aforementioned external force detection techniques based on momentum observers still have certain limitations. Essentially, they are still improvements on a first-order generalized momentum observer to construct a second-order observer. While this improves model accuracy, it cannot eliminate the influence of model errors and makes it difficult to achieve accurate observation of higher-order time-varying external forces.

[0006] Therefore, traditional force / torque sensors suffer from inherent drawbacks such as high cost, narrow load adaptability, susceptibility to interference with robot dynamics, and poor stability in complex environments, making it difficult to meet the demands for dynamic, high-precision external force sensing. While existing methods based on dynamic models (such as second-order generalized momentum observers) can improve model accuracy and detection speed, they cannot eliminate the influence of model errors and are also insufficient for accurate observation of high-order time-varying external forces. Therefore, there is an urgent need to design an external force detection technology that can effectively separate model errors from external torques and achieve high-precision detection of high-order time-varying external forces to overcome current technological bottlenecks. Summary of the Invention

[0007] To address the technical problems existing in the prior art, this invention discloses a robot external force detection method and system based on a two-layer generalized proportional-integral observer. First, a robot dynamic model is established using the Newton-Euler method, and dynamic parameters are identified through excitation trajectories and the least squares method. Then, for each joint, two generalized proportional-integral observers with identical structures but different inputs are constructed. The first-layer observer inputs the robot's actual control torque, while the second-layer observer inputs the robot's output torque. By subtracting the state estimates from the two layers of observers and combining this with the inverse solution of the observer's dynamic equations, the estimated external torque is obtained, thereby eliminating the influence of dynamic model errors and achieving high-precision detection of high-order time-varying external forces. This solves the technical problems in existing robot external force detection technologies, such as the high cost and susceptibility to complex environments that interfere with dynamic characteristics of traditional force / torque sensors, and the inability of momentum observer-based methods to eliminate model-specific errors and accurately observe high-order time-varying external forces.

[0008] The solution adopted by this invention to solve its technical problem is as follows:

[0009] A robot external force detection method based on a two-layer observer includes the following steps:

[0010] Step S1: Establish the robot dynamics model.

[0011] The dynamic model of the robot is established using the Newton-Euler method, and is expressed as follows:

[0012] ;

[0013] In the formula: This refers to the position of the robot's joints. For the speed of the robot joints, The acceleration of the robot's joints, The inertia matrix, The correlation matrix between Coriolis force and centrifugal force; This is the term related to gravity. For joint drive control torque; External force; This refers to the joint friction torque;

[0014] Step S2: Identification of dynamic parameters.

[0015] The dynamic model established in step S1 is linearized, the excitation trajectory is designed, joint data is collected and an overdetermined set of equations is constructed, and the minimum identifiable dynamic parameter set is solved by the least squares method.

[0016] Step S3: Establish a two-layer observer.

[0017] First-level observer G1:

[0018] ;

[0019] Second-layer observer G2:

[0020] ;

[0021] Where G1 is the first-layer observer and G2 is the second-layer observer. Used to estimate the position of robot joints Used to estimate the speed of robot joints. These are used to estimate the perturbation and its (n-1)th derivative, respectively. Indicates joint position, defines Define the position estimation error of observer G1 as follows: The position estimation error of observer G2, For the gain of the observer to be designed, The system control variables are respectively the first-layer observer G1 and the second-layer observer G2. ,

[0022] Step S4: Define and calculate system control variables. ,

[0023] Define system control variables The first layer of observers utilizes the robot's control torque. calculate The second-layer observer uses the robot's output torque. calculate ;

[0024] Step S5: Estimate external forces based on a two-layer observer.

[0025] Define the difference between the state estimates of the two-layer observers. By differentiating the difference and combining it with the observer dynamic equations, we obtain the estimate of the robot's external torque:

[0026] .

[0027] As another embodiment of the present invention

[0028] Step S2 includes the following steps:

[0029] Step S21: Perform a linear transformation on the robot dynamics model established in step S1 to obtain its linear representation:

[0030] ;

[0031] In the formula: For robot joint control torque, For the regression matrix, The minimum identifiable set of dynamic parameters;

[0032] Step S22: Design and optimize the excitation trajectory to continuously stimulate the dynamic characteristics of the robot joints;

[0033] Step S23: Send trajectory data to make the robot move according to the excitation trajectory, collect robot joint data and perform filtering processing to form an overdetermined system of equations:

[0034] ;

[0035] In the formula: It is a torque vector. The observation matrix;

[0036] Step S24: Solve the overdetermined system of equations using the least squares method, and obtain: .

[0037] As another embodiment of the present invention

[0038] In step S4, the control torque during robot operation With the robot's output torque satisfy: ,

[0039] Will Substitute the control items respectively The system control terms for the first-level observer G1 and the second-level observer G2 are as follows:

[0040] .

[0041] As another embodiment of the present invention

[0042] In step S5, when the robot collides, the system satisfies: ,

[0043] Right now: .

[0044] As another embodiment of the present invention

[0045] Step S5 includes the following steps:

[0046] Step S51: Define the difference between the two-layer observer estimates as:

[0047] ;

[0048] Step S52: Take the derivative of the difference between the estimated values ​​to obtain:

[0049] ;

[0050] Based on the two-layer observer designed in step S3, the difference between the observer estimates is obtained by subtracting the observation value of the second-layer observer G2 from the observation value of the first-layer observer G1:

[0051] ;

[0052] The above formula is rearranged to obtain the estimated value of the external force on the robot.

[0053] As another embodiment of the present invention

[0054] In step S23 The expression is: ,in, This indicates the number of joints in the robot. Indicates the number of sampling points. Indicates the sampling time The torque transpose at that time, similarly Indicates the sampling time Torque transposition at time;

[0055] In step S23 The expression is: ;

[0056] In the formula: This indicates the number of parameters to be identified.

[0057] As another embodiment of the present invention

[0058] The dual-layer observer established in step S3 adopts a generalized proportional-integral observer.

[0059] A robot external force detection system based on a two-layer observer includes:

[0060] Dynamics model building module: used to build robot dynamics models using the Newton-Euler method;

[0061] Dynamic parameter identification module: used to design excitation trajectory so that the robot moves along the trajectory, collect joint position, velocity, acceleration and control torque data, construct overdetermined equation system after filtering, and identify the minimum identifiable dynamic parameter set using weighted least squares method for model feedforward compensation;

[0062] Two-layer observer building module: used to build two generalized proportional-integral observers with the same structure but different inputs for each joint;

[0063] System control quantity calculation module: used to calculate the inputs of the first-level observer and the second-level observer. The input of the first-level observer is the actual control torque of the robot joint, and the input of the second-level observer is the output torque of the joint.

[0064] External force estimation module: It is used to subtract the corresponding state estimates of two layers of observers to obtain the difference vector. Based on the dynamic equation of the observer, it derives the differential relationship between the differences, thereby solving for the estimated value of the external torque.

[0065] The dynamic model construction module provides a model structure for the dynamic parameter identification module, and the dynamic parameter identification module feeds back the identified parameter set to the dynamic model construction module to complete the model assignment. The dual-layer observer construction module obtains the inertia matrix, Coriolis force matrix, gravity term, and friction force model from the dynamic model construction module, and calculates the observer gain and the parameters required for system control quantities accordingly. The system control quantity calculation module reads the actual control torque and output torque in each control cycle, calls the parameters provided by the dynamic model construction module to calculate the system control quantities of the first-layer observer and the second-layer observer, and injects them into the two observers of the dual-layer observer construction module respectively. The external force estimation module receives the two-layer state estimation values ​​output by the dual-layer observer construction module, calculates the difference vector, and obtains the external torque estimation value based on the observer dynamic equation. This estimation value is output to the upper-layer safety control system on the one hand, and fed back to the system control quantity calculation module on the other hand, to update the input torque of the second-layer observer, forming a closed-loop iteration.

[0066] A computer device is characterized by comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements a robot external force detection method based on a two-layer observer.

[0067] A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, which, when executed by a processor, implements a robot external force detection method based on a two-layer observer.

[0068] Compared with the prior art, the present invention has the following beneficial effects:

[0069] 1. This invention designs a first-layer observer and a second-layer observer to input control torque and output torque respectively, and uses the difference between their state estimates to effectively separate the internal modeling error of the robot dynamics model from the external torque. Even with a moderately accurate dynamics model, high-fidelity external force estimation results can still be obtained, significantly reducing the accuracy requirements for robot dynamics parameter identification and enhancing the system's robustness to model uncertainties.

[0070] 2. The generalized proportional-integral observer used in this invention can make full use of the information of disturbance and its derivatives. By increasing the order of the observer, the observer can still maintain accurate estimation of external forces that change rapidly with time and have high-order derivative characteristics with low delay and small amplitude decay, thus overcoming the inherent limitation of traditional momentum observers that can only effectively track constant or slowly changing external forces.

[0071] 3. This invention can detect external forces based on the robot's own dynamic model, joint encoder, and feedback signals, without the need for additional force / torque sensors. This saves on expensive sensor purchase costs and avoids additional expenses such as structural modifications, signal conditioning, calibration, and maintenance required for sensor installation, which is conducive to large-scale application in low-cost robots, collaborative robots, and service robots.

[0072] 4. When traditional force sensors are installed on robot joints or end effectors, they introduce additional mass, stiffness, and damping, altering the robot's original dynamic characteristics and thus affecting control accuracy and dynamic response. This invention employs a pure software observer scheme, without altering the robot's structure, thereby completely avoiding the additional elastic deformation and inertial disturbances introduced by sensors and ensuring the robot's original operational performance.

[0073] 5. This invention does not rely on physical sensors that are susceptible to temperature, humidity, dust, vibration, and electromagnetic interference. Its observer algorithm runs in a digital controller, exhibiting strong environmental adaptability. In complex environments where sensors are prone to failure, such as industrial welding, outdoor operations, nuclear radiation detection, and underwater operations, it can still stably output accurate external force estimates, significantly improving the engineering applicability of the external force detection system.

[0074] 6. Physical force sensors are prone to zero-point drift and decreased sensitivity after long-term use, requiring periodic recalibration or even replacement. The observer method of this invention requires no calibration, relying solely on the algorithm parameters within the controller. It does not exhibit mechanical fatigue or electrical aging failure modes, and theoretically can operate for an extended period, extending with the robot's lifespan, significantly reducing maintenance frequency and overall costs. Attached Figure Description

[0075] Figure 1 This is a flowchart illustrating a robot external force detection method based on a dual-layer observer proposed in this invention.

[0076] Figure 2 This is a comparison diagram of step signal detection in a simulation experiment of a robot external force detection method based on a dual-layer observer proposed in this invention;

[0077] Figure 3 This is a comparison diagram of step signal detection when there is a modeling error in the robot external force detection method based on a dual-layer observer proposed in this invention;

[0078] Figure 4 This is a torque comparison diagram of a robot external force detection experiment conducted on a certain joint in a robot external force detection method based on a dual-layer observer proposed in this invention.

[0079] Figure 5 This is a torque comparison diagram of a certain joint in a robot external force detection experiment based on a dual-layer observer proposed in this invention. Detailed Implementation

[0080] The specific embodiments of the present invention are described below with reference to the accompanying drawings and examples:

[0081] It should be noted that the structures, colors, proportions, sizes, etc. shown in the accompanying drawings are only used to complement the content disclosed in the specification, so that those skilled in the art can understand and read them, and are not intended to limit the conditions under which the present invention can be implemented. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0082] like Figures 1-5 As shown, this invention discloses a robot external force detection method based on a two-layer observer, specifically including the following steps:

[0083] Step S1: Establish the robot dynamics model.

[0084] The robot dynamics model is the foundation for realizing external force detection. Therefore, the first step is to establish the robot dynamics model. When the robot is subjected to external forces, the Newton-Euler method is used to establish the robot dynamics model:

[0085] ;

[0086] In the formula: This refers to the position of the robot's joints. For the speed of the robot joints, The acceleration of the robot's joints, The inertia matrix, The correlation matrix between Coriolis force and centrifugal force; This is the term related to gravity. For joint drive control torque; External force; This represents the joint friction torque.

[0087] Step S2: Identify dynamic parameters, including the following steps:

[0088] Step S21: The robot dynamics model established in step S1 has the characteristics of nonlinear strong coupling, making it very difficult to directly identify the dynamic parameters. Therefore, it needs to be linearized to obtain its linear representation.

[0089] ;

[0090] In the formula: For robot joint control torque, For the regression matrix, This is the minimum identifiable set of dynamic parameters.

[0091] Step S22: Design the excitation trajectory. By designing and optimizing the excitation trajectory, the dynamic characteristics of the robot joints are continuously stimulated, and the identification accuracy of the dynamic parameters is improved.

[0092] Step S23: Send trajectory data to make the robot move according to the excitation trajectory, and at the same time collect robot joint data, filter it, and form an overdetermined system of equations:

[0093] ;

[0094] In the formula: Let be the torque vector, and its expression is: ,in, This indicates the number of joints in the robot. Indicates the number of sampling points. Indicates the sampling time The torque transpose at that time, similarly Indicates the sampling time Torque transposition at time To represent the sampling time Torque transposition at time;

[0095] The observation matrix is ​​expressed as follows: ;

[0096] In the formula: This indicates the number of parameters to be identified.

[0097] Step S24: Solve the minimum identifiable parameter set X in the overdetermined system of equations using the least squares method to obtain: .

[0098] Step S3: Establish a two-layer observer.

[0099] External force detection is performed using a generalized proportional-integral observer (GPIO), a high-performance disturbance estimation method that can fully utilize the information of the disturbance's derivatives.

[0100] After a collision, the robot system experiences both internal and external disturbances. Internal disturbances are defined as errors in the robot's dynamics model, while external disturbances are defined as external torques. A difference exists between the system's control input and its feedback output. This difference is utilized to design a two-layer observer to observe both internal and external disturbances separately. The influence of the internal disturbance is eliminated by calculating the difference between the observed values, thus obtaining the required external torque value. The designed two-layer observer is as follows:

[0101] First-level observer G1:

[0102] ;

[0103] Second-layer observer G2:

[0104] .

[0105] Where G1 is the first-layer observer and G2 is the second-layer observer. Used to estimate the position of robot joints Used to estimate the speed of robot joints. ( These are used to estimate the perturbation and its (n-1)th derivative, respectively. Indicates joint position, defines Define the position estimation error of observer G1 as follows: The position estimation error of observer G2, It is necessary to design the observer gain. The system control variables are respectively the first-layer observer G1 and the second-layer observer G2. .

[0106] Step S4: Define and calculate system control variables This includes the following steps:

[0107] Step S41: Define system control variables .

[0108] Step S42: Calculate the system control quantity :

[0109] The first-level observer G1 calculates the control input of the system. When using the system's control torque ,

[0110] The second-level observer G2 calculates the control input of the system. The robot output torque is used at this time .

[0111] During robot operation, control torque With the robot's output torque It should meet the following requirements: .

[0112] Will Substitute the control items respectively The system control terms for the first-level observer G1 and the second-level observer G2 are as follows:

[0113] ;

[0114] Step S5: Estimate external forces based on the established two-layer observer.

[0115] When a robot collides with another robot, its system should satisfy the following: ;Right now: .

[0116] Step S51: Define the difference between the two-layer observer estimates as:

[0117] ;

[0118] Step S52: Taking the derivative of the difference between the estimated values, we get:

[0119] ;

[0120] Based on the observer designed in step S3, the difference between the observer estimates is obtained by subtracting the observation value of the second-layer observer G2 from the observation value of the first-layer observer G1:

[0121] ;

[0122] Simplifying the above equation, we can obtain the estimated external force on the robot as follows:

[0123] .

[0124] A robot external force detection system based on a two-layer observer includes a dynamic model construction module, a dynamic parameter identification module, a two-layer observer construction module, a system control quantity calculation module, and an external force estimation module. Wherein:

[0125] The dynamics model building module is used to build robot dynamics models using the Newton-Euler method.

[0126] The dynamic parameter identification module is used to generate an excitation trajectory based on Fourier series, enabling the robot to move along the trajectory. It collects joint position, velocity, acceleration, and control torque data, constructs an overdetermined set of equations after filtering, and identifies the minimum identifiable dynamic parameter set using the weighted least squares method for model feedforward compensation.

[0127] The dual-layer observer building block is used to build two generalized proportional-integral observers (GPIOs) with the same structure but different inputs for each joint.

[0128] The system control quantity calculation module is used to calculate the inputs of the first-layer observer and the second-layer observer. The input of the first-layer observer is the actual control torque of the robot joint, and the input of the second-layer observer is the output torque of the joint.

[0129] The external force estimation module is used to subtract the corresponding state estimates of the two observers to obtain the difference vector. Based on the observer dynamic equations, the differential relationship between the differences is derived, thereby solving for the estimated value of the external torque.

[0130] The dynamic model construction module provides the model structure for the dynamic parameter identification module, which then feeds back the identified parameter set to the dynamic model construction module to complete the model assignment. The dual-layer observer construction module obtains the inertia matrix, Coriolis force matrix, gravity term, and friction force model from the dynamic model construction module, and calculates the observer gain and the parameters required for system control quantities based on these. The system control quantity calculation module reads the actual control torque and output torque in each control cycle, calls the parameters provided by the dynamic model construction module to calculate the system control quantities of the first-layer observer and the second-layer observer, and injects them into the two observers of the dual-layer observer construction module respectively. The external force estimation module receives the two-layer state estimates output by the dual-layer observer construction module, calculates the difference vector, and obtains the external torque estimate based on the observer dynamic equation. This estimate is output to the upper-layer safety control system on the one hand, and fed back to the system control quantity calculation module on the other hand, to update the input torque of the second-layer observer, forming a closed-loop iteration.

[0131] A computer device includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to implement the above-mentioned robot external force detection method based on a two-layer observer.

[0132] A computer-readable storage medium in which the methods described above according to embodiments of the present invention can be implemented in hardware, firmware, or implemented as recordable on a storage medium, or implemented as computer code originally stored on a remote storage medium or a non-transitory machine-readable storage medium and to be stored on a local storage medium after being downloaded over a network, so that the methods described herein can be processed by such software stored on a storage medium using a general-purpose computer, a special-purpose processor, or programmable or special-purpose hardware.

[0133] The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; furthermore, the storage medium can also include combinations of the above types of memory. It is understood that a computer, processor, microprocessor controller, or programmable hardware includes storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0134] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0135] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

[0136] Many other changes and modifications can be made without departing from the concept and scope of this invention. It should be understood that this invention is not limited to the specific embodiments, and the scope of this invention is defined by the appended claims.

Claims

1. A robot external force detection method based on a two-layer observer, Its features are, Includes the following steps: Step S1: Establish the robot dynamics model. The dynamic model of the robot is established using the Newton-Euler method, and is expressed as follows: ； In the formula: This refers to the position of the robot's joints. For the speed of the robot joints, The acceleration of the robot's joints, The inertia matrix, The correlation matrix between Coriolis force and centrifugal force; This is the term related to gravity. For joint drive control torque; External force; This refers to the joint friction torque; Step S2: Identification of dynamic parameters. The dynamic model established in step S1 is linearized, the excitation trajectory is designed, joint data is collected and an overdetermined set of equations is constructed, and the minimum identifiable dynamic parameter set is solved by the least squares method. Step S3: Establish a two-layer observer. First-level observer G1: ； Second-layer observer G2: ； Where G1 is the first-layer observer and G2 is the second-layer observer. Used to estimate the position of robot joints Used to estimate the speed of robot joints. These are used to estimate the perturbation and its (n-1)th derivative, respectively. Indicates joint position, defines Define the position estimation error of observer G1 as follows: The position estimation error of observer G2, For the gain of the observer to be designed, The system control variables are respectively the first-layer observer G1 and the second-layer observer G2. , Step S4: Define and calculate system control variables. , Define system control variables The first layer of observers utilizes the robot's control torque. calculate The second-layer observer uses the robot's output torque. calculate ; Step S5: Estimate external forces based on a two-layer observer. Define the difference between the state estimates of the two-layer observers. By differentiating the difference and combining it with the observer dynamic equations, we obtain the estimate of the robot's external torque: 。 2. The robot external force detection method based on a dual-layer observer as described in claim 1, Its features are, Step S2 includes the following steps: Step S21: Perform a linear transformation on the robot dynamics model established in step S1 to obtain its linear representation: ； In the formula: For robot joint control torque, For the regression matrix, The minimum identifiable set of dynamic parameters; Step S22: Design and optimize the excitation trajectory to continuously stimulate the dynamic characteristics of the robot joints; Step S23: Send trajectory data to make the robot move according to the excitation trajectory, collect robot joint data and perform filtering processing to form an overdetermined system of equations: ； In the formula: It is a torque vector. The observation matrix; Step S24: Solve the overdetermined system of equations using the least squares method, and obtain: .

3. The robot external force detection method based on a dual-layer observer as described in claim 1, Its features are, In step S4, the control torque during robot operation With the robot's output torque satisfy: , Will Substitute the control items respectively The system control terms for the first-level observer G1 and the second-level observer G2 are as follows: 。 4. The robot external force detection method based on a dual-layer observer as described in claim 1, Its features are, In step S5, when the robot collides, the system satisfies: , Right now: .

5. The robot external force detection method based on a dual-layer observer as described in claim 3, Its features are, Step S5 includes the following steps: Step S51: Define the difference between the two-layer observer estimates as: ； Step S52: Take the derivative of the difference between the estimated values ​​to obtain: ； Based on the two-layer observer designed in step S3, the difference between the observer estimates is obtained by subtracting the observation value of the second-layer observer G2 from the observation value of the first-layer observer G1: ； The above formula is rearranged to obtain the estimated value of the external force on the robot.

6. The robot external force detection method based on a dual-layer observer as described in claim 2, Its features are, In step S23 The expression is: ,in, This indicates the number of joints in the robot. Indicates the number of sampling points. Indicates the sampling time The torque transpose at that time, similarly Indicates the sampling time Torque transposition at time; In step S23 The expression is: ; In the formula: This indicates the number of parameters to be identified.

7. The robot external force detection method based on a dual-layer observer as described in claim 1, Its features are, The dual-layer observer established in step S3 adopts a generalized proportional-integral observer.

8. A robot external force detection system based on a two-layer observer, The robot external force detection method based on a dual-layer observer as described in any one of claims 1-7 is adopted. Its features are, include: Dynamics model building module: used to build robot dynamics models using the Newton-Euler method; Dynamic parameter identification module: used to design excitation trajectory so that the robot moves along the trajectory, collect joint position, velocity, acceleration and control torque data, construct overdetermined equation system after filtering, and identify the minimum identifiable dynamic parameter set using weighted least squares method for model feedforward compensation; Two-layer observer building module: used to build two generalized proportional-integral observers with the same structure but different inputs for each joint; System control quantity calculation module: used to calculate the inputs of the first-layer observer and the second-layer observer. The input of the first-layer observer is the actual control torque of the robot joint, and the input of the second-layer observer is the output torque of the joint. External force estimation module: It is used to subtract the corresponding state estimates of two layers of observers to obtain the difference vector. Based on the dynamic equation of the observer, it derives the differential relationship between the differences, thereby solving for the estimated value of the external torque. The dynamic model construction module provides a model structure for the dynamic parameter identification module, and the dynamic parameter identification module feeds back the identified parameter set to the dynamic model construction module to complete the model assignment. The dual-layer observer construction module obtains the inertia matrix, Coriolis force matrix, gravity term and friction model from the dynamic model construction module, and calculates the observer gain and the parameters required for system control based on these. The system control calculation module reads the actual control torque and output torque in each control cycle, calls the parameters provided by the dynamic model construction module to calculate the system control quantities of the first-layer observer and the second-layer observer, and injects them into the two observers of the dual-layer observer construction module respectively. The external force estimation module receives the two-layer state estimation values ​​output by the two-layer observer construction module, calculates the difference vector, and obtains the external torque estimation value based on the observer dynamic equation. This estimation value is output to the upper-layer safety control system on the one hand, and fed back to the system control quantity calculation module on the other hand, to update the input torque of the second-layer observer, forming a closed-loop iteration.

9. A computer device, Its features are, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a robot external force detection method based on a dual-layer observer as described in any one of claims 1-7.

10. A computer-readable storage medium, Its features are, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the robot external force detection method based on a dual-layer observer as described in any one of claims 1-7.

Images