A robot collision detection and compliance control method based on joint current feedback

Abstract

The application discloses a kind of robot collision detection and compliance control method based on joint current feedback, comprising: by constructing joint space dynamics model, with high sampling rate acquisition each joint motor three-phase current feedback signal and calculate current rate of change, combined with inverse dynamics model operation or directly read controller instruction to obtain joint real-time acceleration, design dynamic threshold value containing static reference threshold, acceleration correlation component and adjustable experience bias term, after threshold comparison triggers primary alarm, through duration window determination and multi-joint information fusion confirm effective collision and classification.The method takes current rate of change as core observation, moves detection opportunity to the initial stage of current mutation, dynamic threshold value can be adjusted adaptively with motion acceleration, with multi-dimensional verification mechanism, accurately distinguish the normal fluctuation of inertia force and the abnormal change caused by collision, improve the safety performance and operation stability of robot in human-robot symbiosis scene.

Description

Technical Field

[0001] This invention relates to the field of robot collision detection and control technology, and in particular to a robot collision detection and compliant control method based on joint current feedback. Background Technology

[0002] As industrial automation upgrades towards intelligence, human-robot collaborative work scenarios are becoming increasingly common. The frequency of robot contact with humans and the environment is increasing significantly, making collision detection and safe interaction core requirements for ensuring operational safety. In unstructured environments, robots need to quickly identify collision risks and respond promptly to avoid personnel injury and equipment damage. This need is driving continuous iteration of collision detection technology, and detection solutions that balance low cost and high reliability have become a focus of industry attention. Optimizing related technologies is of great significance for the safe and efficient operation of various fields such as industrial production and medical services.

[0003] Existing collision detection technologies mainly fall into two categories. One is based on external sensors, which directly captures external force signals to achieve collision recognition by installing specialized sensing devices at the robot's end effector or joints. The other is based on joint current feedback, which uses the correlation between joint motor current and output torque to indirectly sense the current signal and determine whether a collision has occurred by monitoring current changes. The former relies on additional hardware to build the sensing system, while the latter utilizes feedback signals from the robot's own control system, eliminating the need for numerous external devices. These two approaches are adopted in different application scenarios.

[0004] Existing technologies suffer from two main drawbacks: First, they exhibit significant response lag. Current feedback-based solutions often rely on comparing the absolute value of the current with a fixed threshold. However, inertia, friction, and control system delays during mechanical transmission make it difficult for the current signal to rise rapidly in the initial stages of a collision. This means that the collision may take some time to be detected, hindering early warning. Second, they suffer from a high false alarm rate. During acceleration and deceleration, inertial forces cause normal fluctuations in the current. Fixed thresholds cannot distinguish between these normal fluctuations and current changes caused by collisions. In complex motion conditions, this can easily trigger unnecessary alarms, affecting the continuity and stability of robot operations. Summary of the Invention

[0005] In order to overcome the shortcomings and deficiencies of the existing technology, the present invention provides a robot collision detection and compliant control method based on joint current feedback.

[0006] The technical solution adopted in this invention is a robot collision detection and compliant control method based on joint current feedback, comprising the following steps: S1, constructing a joint space dynamics model of the controlled robot, determining the correlation between joint angle vectors, mass matrix, Coriolis force and centrifugal force terms, gravity term and friction term, providing support for theoretical analysis of joint torque and current and preliminary setting of dynamic threshold; S2, acquiring the three-phase current feedback signals of each joint motor in real time at a high sampling rate through the servo driver of the robot control system; S3, performing numerical differentiation on the acquired current signals to obtain the current change rate of each joint in real time; S4, analyzing the joint position signals through the robot inverse dynamics model. Perform second-order differential calculations or directly read the joint expected acceleration command output by the controller planning module to obtain the real-time acceleration information of each joint; S5, design a dynamic threshold associated with the robot's real-time motion state. This threshold includes a static reference current change rate threshold, a component associated with the real-time joint acceleration, and an adjustable empirical bias term. Compare the absolute value of the real-time calculated current change rate with the dynamic threshold to trigger a primary contact or collision alarm; S6, introduce a duration window determination mechanism and a multi-joint information fusion mechanism to verify the primary alarm. After confirming a valid collision event, classify the collision event into levels based on the current change rate overshoot amplitude, abnormal duration, and the number of joints involved.

[0007] Furthermore, the joint space dynamics model satisfies: ,in This is the joint torque vector. Joint angle vector, The mass matrix depends on the joint angle. This is the matrix of Coriolis force and centrifugal force coefficients. The joint angular velocity vector. The joint angular acceleration vector. This is a gravity term vector that depends on the joint angle. The dynamic threshold is a frictional force vector dependent on joint angular velocity; the dynamic threshold satisfies: ,in This is a dynamic threshold vector that changes over time. The threshold for the rate of change of the static reference current. For acceleration influence coefficient, This is the real-time acceleration vector of the joint. This is an adjustable empirical bias term.

[0008] Furthermore, the rate of change of current is calculated using the following formula: ,in For the first The first joint The rate of change of current at time t. For the first The first joint The current value collected at each moment, For the first The first joint The current value collected at any given time. The sampling period is the current signal period; the joint acceleration, when calculated using the inverse dynamics model, satisfies the following: ,in The joint angular acceleration vector. For the mass matrix The inverse matrix, This is the joint torque vector. This is the matrix of Coriolis force and centrifugal force coefficients. The joint angular velocity vector. The gravity term vector, This is the vector of the friction force term.

[0009] Furthermore, the collision level classification satisfies: ,in This is a quantification value for the collision level. This is the overshoot weighting coefficient for the rate of change of current. For the first The first joint The rate of change of current at time t. For dynamic threshold vectors, The weighting coefficient is the duration of the anomaly. For the duration of the abnormality, As the baseline time threshold, For the weighting coefficients involving the number of joints, The number of joints exhibiting abnormality; the acceleration influence coefficient in the dynamic threshold satisfies: ,in For acceleration influence coefficient, This is the proportionality coefficient. For the mass matrix The traces.

[0010] Furthermore, the multi-joint information fusion mechanism satisfies: ,in This is the fusion determination value. The total number of joints. For the first The weighting coefficient of each joint, For the first The first joint The rate of change of current at time t. For the first Dynamic values ​​of each joint. The sign function is used; the duration window determination satisfies: ,in The value is determined within the time window. The moment when the alarm is first triggered. The number of control cycles included in the time window. For the first The first joint The rate of change of current at time t. For the first Dynamic threshold of each joint It is a symbolic function.

[0011] Furthermore, the servo driver employs a multi-channel synchronous acquisition module, supporting parallel acquisition of three-phase current signals. The sampling rate is adjustable from 1kHz to 10kHz, the resolution of the acquisition channel is no less than 16 bits, and it features a shielded structure design to resist electromagnetic interference. The numerical differential operation of the current signal adopts a central difference algorithm, with the operation step size consistent with the sampling period. The control cycle number of the duration window ranges from 2 to 5. When fusing multi-joint information, the weight coefficient of each joint is allocated according to the force sensitivity of the joint during robot motion, with a value range from 0.1 to 1.0. In the dynamic threshold, the static reference current change rate threshold is determined based on the current fluctuation characteristics of the robot during no-load operation. The adjustable empirical bias term is adjusted online according to the collision detection sensitivity requirements of different operating scenarios.

[0012] Further, step S3 includes the following sub-steps: S31, receiving the collected three-phase current feedback signals of each joint motor, classifying and organizing the current signals according to the joint number, and establishing a time sequence of current signals for each joint; S32, determining the interval between adjacent sampling points for numerical differentiation, and extracting the current sampling values ​​of two consecutive adjacent moments for each joint based on the time sequence of the current signals; S33, processing the extracted adjacent current sampling values ​​according to preset differentiation rules to obtain the current change amount at the corresponding moment for each joint; S34, storing the calculated current change amount in association with the joint number according to the time sequence to form the time sequence data of the current change rate of each joint, providing observations for collision detection.

[0013] Further, S4 includes the following sub-steps: S41, if the inverse dynamics model is used to calculate joint acceleration, first acquire the real-time position signals of each joint of the robot, and continuously sample the position signals at preset time intervals to form position timing data; S42, perform first-order differential operation on the acquired position timing data to obtain the angular velocity timing signals of each joint, and then perform second-order differential operation on the angular velocity timing signals to obtain the initial value of the joint angular acceleration; S43, if the desired acceleration command is directly read, establish a data communication link with the controller planning module and receive the desired acceleration command data of each joint output by the module in real time; S44, perform timing alignment processing on the acceleration data obtained by the two methods to remove the delay deviation in the data transmission process and form a set of real-time acceleration information for each joint.

[0014] Further, S5 includes the following sub-steps: S51, obtaining the mass matrix correlation parameters in the robot dynamics model, determining the initial value of the acceleration influence coefficient in combination with the joint motion characteristics, and setting the static reference current change rate threshold based on the current noise level when the system has no external contact; S52, collecting the acceleration information of each joint in real time, calculating the absolute value of acceleration and the acceleration influence coefficient to obtain the dynamic adjustment component associated with the motion; S53, setting the specific value of the adjustable empirical bias term according to the collision detection requirements of the work scenario, superimposing the static reference threshold, the dynamic adjustment component, and the empirical bias term to obtain the real-time dynamic threshold; S54, extracting the absolute value of the real-time current change rate of each joint, comparing it one by one with the corresponding real-time dynamic threshold, and generating and outputting a primary contact or collision alarm signal when the absolute value of the current change rate exceeds the dynamic threshold.

[0015] A robot collision detection and compliant control method based on joint current feedback is disclosed. This method is implemented through a robot collision detection and compliant control system based on joint current feedback, comprising: a joint dynamics model construction unit, used to establish the joint space dynamics model of the controlled robot, determine the correlation between various mechanical parameters, and provide a theoretical basis for analysis; a multi-channel current signal acquisition unit, connected to the robot servo driver, to acquire the three-phase current feedback signals of each joint motor in real time at a high sampling rate, and to classify and store them; and a current change rate calculation unit, which receives the data output by the current signal acquisition unit and obtains the current of each joint through numerical differentiation. The system includes: a rate-of-change time-series data acquisition unit (POC) for acquiring and processing real-time acceleration information of each joint via inverse dynamics model calculation or direct reading of controller commands; a dynamic threshold calculation and primary collision determination unit for receiving acceleration information and current rate-of-change data, calculating and comparing real-time dynamic thresholds, and triggering a primary alarm; and a collision level confirmation and control response unit for receiving alarm signals from the primary determination unit, confirming valid collision events and classifying them into levels through duration window determination and multi-joint information fusion, and outputting corresponding compliant control commands based on the level. All units establish a bidirectional communication connection via a data bus for real-time signal transmission and processing.

[0016] Beneficial Effects: This invention proposes a robot collision detection and compliant control method based on joint current feedback. Using the joint current change rate as the core observation, it replaces traditional absolute current detection, significantly advancing the collision detection trigger time and solving the response lag problem of existing solutions. This shifts collision recognition from the current accumulation exceeding the threshold stage to the initial stage of current mutation, significantly shortening the safety response time. Simultaneously, a dynamic threshold strongly correlated with the robot's real-time motion acceleration is introduced. Combined with multi-joint information fusion and a duration window determination mechanism, it accurately distinguishes between normal current fluctuations caused by motion inertia and abnormal current changes caused by external collisions, fundamentally reducing the system's false alarm rate. A high sampling rate, high resolution multi-channel synchronous acquisition module and anti-interference design are employed. Optimized numerical calculation algorithms and parameter configuration strategies further improve detection accuracy and stability. This invention does not rely on additional external sensors. While controlling costs, it achieves accurate responses to collisions of different risk levels through collision classification and corresponding compliant control strategies. This ensures human-robot collaboration safety, avoids unnecessary task interruptions, and improves the robot's safety performance and operational reliability under complex working conditions. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the overall process of the method of the present invention. Figure 2 This is a flowchart of method step S3 of the present invention; Figure 3 This is a flowchart of method step S4 of the present invention; Figure 4 This is a flowchart of step S5 of the method of the present invention; Figure 5 This is a visual diagram of the system interface for the method of the present invention. Figure 6 The diagram shows the execution of the backend code of the system running the method of the present invention. Detailed Implementation

[0018] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] like Figure 1 As shown, a robot collision detection and compliant control method based on joint current feedback includes the following steps: S1, constructing a joint space dynamics model of the controlled robot, determining the correlation between joint angle vectors, mass matrix, Coriolis force and centrifugal force terms, gravity term and friction term, providing support for theoretical analysis of joint torque and current and preliminary setting of dynamic threshold; S2, acquiring the three-phase current feedback signals of each joint motor in real time at a high sampling rate through the servo driver of the robot control system; S3, performing numerical differentiation on the acquired current signals to obtain the current change rate of each joint in real time; S4, performing secondary differentiation on the joint position signals through the robot inverse dynamics model. S5. The system performs calculations or directly reads the joint expected acceleration command output by the controller planning module to obtain real-time acceleration information for each joint; S6. A dynamic threshold associated with the robot's real-time motion state is designed. This threshold includes a static reference current change rate threshold, a component associated with the real-time joint acceleration, and an adjustable empirical bias term. The absolute value of the real-time calculated current change rate is compared with the dynamic threshold to trigger a primary contact or collision alarm; S7. A duration window determination mechanism and a multi-joint information fusion mechanism are introduced to verify the primary alarm. After confirming a valid collision event, the collision event is classified according to the overshoot amplitude of the current change rate, the abnormal duration, and the number of joints involved.

[0020] Step S1 involves constructing a joint space dynamics model of the controlled robot. This process requires a comprehensive analysis of the mechanical characteristics and interactions of each joint. During implementation, the mechanical parameters related to the joint angles are first determined based on the robot's structural design parameters, material properties, and joint connection methods. The quantitative correlation between the joint angle vector and the mass matrix, Coriolis force, centrifugal force, gravity, and friction terms is then established. The mass matrix needs to be accurately calculated based on the moment of inertia of each joint, the mass distribution of the links, and the relative positions of the joints. The Coriolis force and centrifugal force terms need to consider the influence of joint angular velocity on the mechanical characteristics. The gravity term needs to be dynamically adjusted in conjunction with the robot's posture changes during operation. The friction term is determined by fitting frictional resistance data from experimental tests at different joint angular velocities. This model construction provides a foundation for the subsequent theoretical conversion of joint torques and currents, and also provides crucial mechanical parameter support for the initial setting of dynamic thresholds, ensuring a high degree of matching between the subsequent detection mechanism and the robot's actual physical characteristics. The model's accuracy directly affects the sensitivity and accuracy of collision detection. Through multiple simulation iterations and actual test calibrations, the model parameters can be further optimized to improve the model's adaptability to the robot's dynamic motion states.

[0021] Step S2 involves real-time acquisition of the three-phase current feedback signals from the robot's joint motors. This is achieved through the servo drivers of the robot control system, using a multi-channel synchronous acquisition module to achieve parallel signal acquisition. The acquisition module must have an anti-electromagnetic interference shielding structure to prevent electromagnetic noise in the industrial environment from affecting signal quality. The sampling rate is adjustable from 1kHz to 10kHz, with the specific value determined based on the robot's movement speed and collision detection response requirements. Higher sampling rates can be selected for high-speed motion scenarios to ensure the capture of instantaneous changes in the current signal. The resolution of the acquisition channel is no less than 16 bits to ensure the accuracy of the current signal acquisition and to accurately identify minute current fluctuations. During the acquisition process, the three-phase current signals of each joint motor are synchronously captured, categorized and stored according to joint number, forming an independent current signal dataset for each joint. Simultaneously, an association index between the acquisition timestamp and the joint number is established to ensure accurate location of the current information of each joint at each moment during subsequent data processing. The acquired raw current signals are directly transmitted to the subsequent processing unit, providing a high-quality data source for calculating the current change rate.

[0022] Step S3 performs numerical differentiation on the acquired current signals to obtain the current change rate of each joint. First, the time-series data of the current signals of each joint, acquired and categorized in step S2, are received. The corresponding current time-series sequence is retrieved one by one according to the joint number to ensure the orderly processing of data. The numerical differentiation operation uses the central difference algorithm, with the operation step size consistent with the sampling period. The sampling period is determined according to the sampling rate set in step S2; for example, when the sampling rate is 5kHz, the sampling period is 200 microseconds, and the operation step size is synchronously set to 200 microseconds. During implementation, the current sample values ​​of two consecutive adjacent moments for each joint are extracted in chronological order, strictly adhering to the time sequence to avoid data misalignment. The current sample values ​​of adjacent moments are calculated and processed using the preset central difference operation rules to obtain the current change amount of each joint at the corresponding moment. After the calculation is completed, the current change of each joint is associated and stored in chronological order and joint number to form a complete time series dataset of current change rate. This dataset serves as the core observation for collision detection and can accurately reflect the instantaneous change trend of the current signal, providing key data support for early collision recognition. During the operation, it is necessary to ensure the real-time performance of data transmission and calculation to avoid affecting the response speed of collision detection due to delay.

[0023] Step S4 acquires real-time acceleration information for each joint, providing two selectable implementation methods for different application scenarios. If the inverse dynamics model calculation method is used, firstly, the real-time position signals of each joint are acquired through the robot's position sensors. The position signals are continuously sampled at preset time intervals of 1ms to 5ms to form position timing data for each joint. First-order differentiation is performed on the acquired position timing data to obtain the angular velocity timing signals for each joint. Then, second-order differentiation is performed on the angular velocity timing signals to obtain the initial values ​​of the joint angular acceleration. During the calculation, a smoothing algorithm is used to reduce noise introduced by the differentiation operation. If the desired acceleration command is read directly, a high-speed data communication link must be established with the controller planning module, with a communication rate of no less than 1Mbps, to ensure real-time reception of the desired acceleration command data for each joint output by the module. The command data must include key information such as the joint number, timestamp, and acceleration value. Regardless of the method used, the obtained acceleration data must ultimately undergo time-series alignment processing. This involves using timestamp calibration to remove delays and biases during data transmission, eliminating time differences between acceleration data from different joints, and forming a unified time series of real-time acceleration information for each joint. This provides accurate motion state parameters for subsequent dynamic threshold calculations.

[0024] Step S5 involves first constructing and calculating dynamic thresholds in real time, then triggering a primary alarm through threshold comparison. The construction of dynamic thresholds requires combining robot dynamics model parameters with the actual operating scenario. First, the mass matrix correlation parameters in the dynamics model are obtained. The initial value of the acceleration influence coefficient is determined based on joint motion characteristics. Based on the current noise level test data during robot no-load operation, a static reference current change rate threshold is set to ensure that the threshold covers the normal noise range of the system. Subsequently, acceleration information of each joint is collected in real time. The absolute value of the acceleration is multiplied by the acceleration influence coefficient to obtain a dynamic adjustment component strongly correlated with the motion state. According to the collision detection sensitivity requirements of different operating scenarios, the specific value of the adjustable empirical bias term is set. The static reference threshold, dynamic adjustment component, and empirical bias term are then superimposed to obtain the real-time changing dynamic threshold for each joint. Simultaneously, the absolute value of the real-time current change rate of each joint is extracted and compared with the corresponding real-time dynamic threshold one by one according to the joint number. The comparison process must be real-time, and the comparison time of each frame of data shall not exceed 100 microseconds. When the absolute value of the current change rate of a certain joint exceeds its corresponding real-time dynamic threshold, a primary contact or collision alarm signal is immediately generated. The alarm signal includes key information such as joint number, trigger time, current change rate value and dynamic threshold value, providing detailed basis for subsequent collision confirmation.

[0025] Step S6 verifies the primary alarm and completes the collision level classification to ensure the accuracy of collision detection and the targeted nature of subsequent responses. Implementation begins with a duration window determination mechanism. The control cycle number of the duration window ranges from 2 to 5, specifically determined based on the robot's motion characteristics and operational scenario. When a joint triggers a primary alarm, the current change rate of that joint is continuously monitored within the set control cycle number, comparing it with the dynamic threshold, and the duration of the threshold exceeding the limit is recorded. Simultaneously, a multi-joint information fusion mechanism is activated. Weight coefficients are assigned to each joint based on its force sensitivity during robot motion. The weight coefficient ranges from 0.1 to 1.0, with higher weights assigned to joints with higher force sensitivity. The primary alarm states of all joints are weighted and summed to obtain a multi-joint fusion determination value. Only when the abnormal duration within the duration window reaches the set standard, or the multi-joint fusion determination value exceeds the preset threshold, is the event confirmed as a valid collision event. After confirming a valid collision, the collision event is classified into three levels based on three core indicators: overshoot amplitude of current change rate, abnormal duration, and number of joints involved. Overshoot amplitude is calculated by the proportion of current change rate exceeding the dynamic threshold, abnormal duration is measured by the number of control cycles, and the number of joints involved is the total number of joints that are abnormal at the same time. Based on the quantitative results of the three indicators, the collision event is classified into Level 1 collision with slight contact and Level 2 collision with safety risks, providing a basis for the subsequent invocation of compliant control strategies.

[0026] Preferably, the joint space dynamics model satisfies: ,in This is the joint torque vector. Joint angle vector, The mass matrix depends on the joint angle. This is the matrix of Coriolis force and centrifugal force coefficients. The joint angular velocity vector. The joint angular acceleration vector. This is a gravity term vector that depends on the joint angle. The dynamic threshold is a frictional force vector dependent on joint angular velocity; the dynamic threshold satisfies: ,in This is a dynamic threshold vector that changes over time. The threshold for the rate of change of the static reference current. For acceleration influence coefficient, This is the real-time acceleration vector of the joint. This is an adjustable empirical bias term.

[0027] Specifically, the derivation of the joint space dynamics model is based on the Newton-Euler equations. By analyzing the force state of each joint of the robot, the joint torque is decomposed into four parts: inertial torque, Coriolis force and centrifugal force, gravity and friction force corresponding to the mass matrix. Each mechanical term is directly related to the joint motion parameters. This derivation process fully considers the dynamic mechanical characteristics of the robot during motion, ensuring that the model can accurately reflect the quantitative relationship between joint torque and motion parameters, providing theoretical support for subsequent correlation analysis of current and torque. The derivation of the dynamic threshold formula addresses the shortcomings of existing fixed thresholds. Combining the influence of inertial force caused by acceleration during robot motion on the current, dynamic adjustment terms and empirical bias terms related to acceleration are added to the static reference threshold. The static reference threshold value is determined based on the current noise level of the robot during no-load operation, usually between 1.2 and 1.5 times the peak value of the system noise. The acceleration influence coefficient is determined through experimental calibration based on the mass matrix parameters, with a value range of 0.05 to 0.3. The adjustable empirical bias term is set between 0.1 and 0.5 according to the sensitivity requirements of the operation scenario. During implementation, the correlation coefficient between joint torque and current is first calculated through a dynamic model, and then acceleration data is collected in real time to calculate the dynamic threshold. The two formulas work together to provide a mechanical theoretical basis for collision detection and to achieve dynamic self-adaptation of the threshold, thus ensuring the accuracy of detection.

[0028] Preferably, the rate of change of current is calculated using the following formula: ,in For the first The first joint The rate of change of current at time t. For the first The first joint The current value collected at any given time. For the first The first joint The current value collected at any given time. The sampling period is the current signal period; the joint acceleration, when calculated using the inverse dynamics model, satisfies the following: ,in The joint angular acceleration vector. For the mass matrix The inverse matrix, This is the joint torque vector. This is the matrix of Coriolis force and centrifugal force coefficients. The joint angular velocity vector. The gravity term vector, This is the vector of the friction force term.

[0029] Specifically, the derivation of the current change rate formula adopts the differential principle. Considering the temporal characteristics of the current signal, the ratio of the current difference between two adjacent moments to the sampling period characterizes the rate of current change. This derivation method can quickly capture instantaneous current changes, with low computational load and strong real-time performance, making it suitable for rapid detection in the initial stage of a collision. The sampling period is matched with the current acquisition sampling rate; when the sampling rate is between 1kHz and 10kHz, the sampling period corresponds to 1ms to 100 microseconds. The derivation of the joint acceleration inverse dynamics calculation formula involves matrix inversion of the forward dynamics model. By subtracting the Coriolis force, centrifugal force, gravity, and friction terms from the joint torque, and then multiplying by the inverse of the mass matrix, the joint acceleration is solved in reverse. This derivation process fully utilizes the mechanical relationships of the forward dynamics model, ensuring the accuracy of acceleration calculation. During implementation, the current values ​​at continuous moments are first collected by the servo driver, and then the core observations are calculated by substituting them into the current change rate formula. If the acceleration is obtained by inverse dynamics, the mechanical parameters need to be determined by the dynamic model first, and then substituted into the formula for calculation. The implementation of both formulas must ensure the real-time performance of data transmission and calculation, and the calculation delay must be controlled within 100 microseconds to provide accurate data support for subsequent collision detection.

[0030] Preferably, the collision level classification satisfies: ,in This is a quantification value for the collision level. This is the overshoot weighting coefficient for the rate of change of current. For the first The first joint The rate of change of current at time t. For dynamic threshold vectors, The weighting coefficient is the duration of the anomaly. For the duration of the abnormality, As the baseline time threshold, For the weighting coefficients involving the number of joints, The number of joints exhibiting abnormality; the acceleration influence coefficient in the dynamic threshold satisfies: ,in For acceleration influence coefficient, This is the proportionality coefficient. For the mass matrix The traces.

[0031] Specifically, the derivation of the collision level classification formula takes into account the multi-dimensional influencing factors of collision detection. It weights and sums three core indicators: current rate of change overshoot amplitude, abnormal duration, and number of involved joints. Each indicator is normalized and multiplied by its corresponding weight coefficient to ensure a reasonable contribution of each indicator to the level classification. This derivation method comprehensively reflects the severity of the collision, avoiding the one-sidedness of judging by a single indicator. The weight coefficient for current rate of change overshoot ranges from 0.4 to 0.6, the weight coefficient for abnormal duration is 0.2 to 0.3, and the weight coefficient for the number of involved joints is 0.2 to 0.3. The reference time threshold is set to 2 to 5 control cycles. The derivation of the acceleration influence coefficient formula is based on the correlation between the trace of the mass matrix and the acceleration influence. The trace of the mass matrix reflects the overall inertial characteristics of the robot joints. It is correlated with the acceleration influence coefficient through a scaling factor, which is experimentally calibrated to be between 0.02 and 0.1. During implementation, the quantitative values ​​of the three collision level evaluation indicators are first calculated in real time, and then substituted into the weighted formula to obtain the level quantitative value. At the same time, the acceleration influence coefficient is calculated through the mass matrix parameters. The two formulas work together to achieve accurate classification of collision levels and provide key parameters for dynamic thresholds, thereby improving the pertinence of detection and response.

[0032] Preferably, the multi-joint information fusion mechanism satisfies: ,in This is the fusion determination value. The total number of joints. For the first The weighting coefficient of each joint, For the first The first joint The rate of change of current at time t. For the first Dynamic values ​​of each joint. The sign function is used; the duration window determination satisfies: ,in The value is determined within the time window. The moment when the alarm is first triggered. The number of control cycles included in the time window. For the first The first joint The rate of change of current at time t. For the first Dynamic threshold of each joint It is a symbolic function.

[0033] Specifically, the derivation of the multi-joint information fusion formula adopts a weighted summation method. Considering the differences in force sensitivity of different joints during robot motion, different weight coefficients are assigned to each joint. A sign function is used to determine whether the joint triggers the threshold, and then a weighted summation is performed to obtain the fusion judgment value. This derivation method can integrate the detection information of multiple joints, reduce the impact of false triggering by a single joint, and improve the reliability of the judgment. The weight coefficients of each joint are allocated between 0.1 and 1.0 according to the force sensitivity, with higher weight coefficients for force-sensitive joints. The derivation of the duration window judgment formula is based on the stability verification principle of timing signals. By statistically analyzing the number of times the joint triggers the threshold within the set time window, if the number of triggers reaches the number of control cycles included in the time window, the anomaly is confirmed as valid. This derivation method can filter transient interference signals and avoid false alarms. The number of control cycles included in the time window is between 2 and 5. The sign function outputs 1 when the current change rate exceeds the threshold, and 0 otherwise. During implementation, the current change rate and dynamic threshold data of each joint are first collected and substituted into the multi-joint information fusion formula to calculate the fusion judgment value; at the same time, the duration window timing is started and substituted into the formula to count the number of triggers. The results of the two formulas are used together as the basis for judging the effective collision event. During the implementation process, it is necessary to ensure that the data is processed synchronously, and the delay of fusion calculation and time window statistics are both controlled within 50 microseconds.

[0034] Preferably, the servo driver adopts a multi-channel synchronous acquisition module, which supports parallel acquisition of three-phase current signals. The sampling rate is adjustable from 1kHz to 10kHz, the resolution of the acquisition channel is not less than 16 bits, and it has a shielding structure design to resist electromagnetic interference. The numerical differential operation of the current signal adopts the central difference algorithm, and the operation step size is consistent with the sampling period. The control cycle number of the duration window ranges from 2 to 5. When the multi-joint information is fused, the weight coefficient of each joint is allocated according to the force sensitivity of the joint in the robot's movement, and the value ranges from 0.1 to 1.0. The static reference current change rate threshold in the dynamic threshold is determined according to the current fluctuation characteristics of the robot when it is running without load. The adjustable empirical bias term is adjusted online according to the collision detection sensitivity requirements of different working scenarios.

[0035] Preferred, such as Figure 2As shown, step S3 includes the following sub-steps: S31, receiving the collected three-phase current feedback signals of each joint motor, classifying and organizing the current signals according to the joint number, and establishing a time sequence of current signals for each joint; S32, determining the interval between adjacent sampling points for numerical differentiation operation, and extracting the current sampling values ​​of two consecutive adjacent moments for each joint based on the time sequence of the current signals; S33, processing the extracted adjacent moment current sampling values ​​according to preset differentiation operation rules to obtain the current change amount at the corresponding moment for each joint; S34, storing the calculated current change amount in association with the joint number according to the time sequence to form the time sequence data of the current change rate of each joint, providing observations for collision detection.

[0036] Specifically, the current change rate calculation process in step S3 ensures accuracy and real-time performance through four sub-steps. In step S31, the three-phase current feedback signals from each joint motor are received from the multi-channel current signal acquisition unit. The signals are then categorized and organized according to the joint number, eliminating any invalid data that may have occurred during transmission. An independent current signal time sequence for each joint is established, including sampling timestamps and corresponding current values, providing an ordered data source for subsequent calculations. In step S32, the interval between adjacent sampling points for numerical differentiation is determined based on the current acquisition sampling rate. When the sampling rate is between 1kHz and 10kHz, the interval corresponds to 1ms to 100 microseconds. Current sampling values ​​from two consecutive adjacent moments are extracted from the current signal sequence of each joint strictly according to the time sequence, ensuring the temporal continuity and accuracy of the correspondence between the sampling values. In step S33, the central difference algorithm is used as the preset differentiation operation rule. This algorithm has low computational complexity and high accuracy. The extracted current sampling values ​​from adjacent moments are substituted into the algorithm for processing to obtain the current change at each joint at the corresponding moment. During the operation, the delay of a single operation must be controlled to not exceed 50 microseconds to meet real-time detection requirements. S34 stores the calculated changes in current at each joint in chronological order and by joint number, generating a time-series dataset that includes joint identifiers, timestamps, and current change rate values. This dataset serves as the core observation for collision detection and can accurately capture the sudden changes in current at the initial stage of a collision, providing high-quality data support for subsequent comparison with dynamic thresholds. The entire step-by-step implementation process must ensure the continuity of data processing, and the connection delay between each step must be controlled within 100 microseconds.

[0037] Preferred, such as Figure 3As shown, step S4 includes the following sub-steps: S41, if the inverse dynamics model is used to calculate joint acceleration, first acquire the real-time position signals of each joint of the robot, and continuously sample the position signals at preset time intervals to form position timing data; S42, perform first-order differential operation on the acquired position timing data to obtain the angular velocity timing signals of each joint, and then perform second-order differential operation on the angular velocity timing signals to obtain the initial value of the joint angular acceleration; S43, if the desired acceleration command is directly read, establish a data communication link with the controller planning module and receive the desired acceleration command data of each joint output by the module in real time; S44, perform timing alignment processing on the acceleration data obtained by the two methods to remove the delay deviation in the data transmission process and form a set of real-time acceleration information for each joint.

[0038] Specifically, the joint acceleration acquisition process in step S4 ensures the accuracy of acceleration information through standardized implementation of two paths. S41: If the inverse dynamics model calculation method is used, the real-time position signals of each joint are first acquired through the robot's position sensors. The position signals are continuously sampled at preset time intervals of 1ms to 5ms, with the sampling process synchronized with the current acquisition, forming position timing data for each joint. This data includes the joint position information and corresponding timestamp at each sampling moment. S42: The acquired position timing data undergoes a first-order differential operation to obtain the angular velocity timing signals of each joint. A smoothing algorithm is used during the operation to reduce noise interference. Then, a second differential operation is performed on the angular velocity timing signals to obtain the initial values ​​of the joint angular acceleration. Both differential operations strictly follow numerical calculation rules to ensure calculation accuracy. S43: If the direct reading method is used, a high-speed data communication link with a rate of no less than 1Mbps is established with the controller planning module. Through this link, the desired acceleration command data for each joint output by the module is received in real time. The command data must include key information such as the joint number, timestamp, and acceleration value to ensure the real-time performance and integrity of data transmission. S44 performs time-series alignment processing on the acceleration data obtained through the two methods. Using the timestamp of current acquisition as a reference, it calibrates the time deviation of the acceleration data, removes the delay error generated during data transmission and calculation, and finally forms a set of real-time acceleration information for each joint under a unified time sequence, providing accurate motion state parameters for dynamic threshold calculation.

[0039] Preferred, such as Figure 4As shown, step S5 includes the following sub-steps: S51, obtaining the mass matrix correlation parameters in the robot dynamics model, determining the initial value of the acceleration influence coefficient based on the joint motion characteristics, and setting the static reference current change rate threshold based on the current noise level when the system has no external contact; S52, collecting the acceleration information of each joint in real time, calculating the absolute value of acceleration and the acceleration influence coefficient to obtain the dynamic adjustment component associated with the motion; S53, setting the specific value of the adjustable empirical bias term according to the collision detection requirements of the work scenario, superimposing the static reference threshold, the dynamic adjustment component, and the empirical bias term to obtain the real-time dynamic threshold; S54, extracting the absolute value of the real-time current change rate of each joint, comparing it one by one with the corresponding real-time dynamic threshold, and generating and outputting a primary contact or collision alarm signal when the absolute value of the current change rate exceeds the dynamic threshold.

[0040] Specifically, the dynamic threshold calculation and initial judgment process in step S5 ensures the dynamic adaptability of the threshold and the accuracy of the judgment. During S51, relevant parameters of the mass matrix are obtained from the joint dynamics model construction unit. Combined with the motion characteristics and force analysis results of each joint, the initial value of the acceleration influence coefficient is determined, ranging from 0.05 to 0.3. Simultaneously, through a no-load, no-load operation test of the robot, the current noise level when the system has no external contact is recorded. 1.2 to 1.5 times the noise peak value is taken as the static reference current change rate threshold to ensure that the threshold can cover the normal noise range. In S52, the acceleration information of each joint output by the joint acceleration acquisition unit is received in real time. The absolute value of the acceleration of each joint is extracted and multiplied with the preset acceleration influence coefficient to obtain a dynamic adjustment component strongly correlated with the robot's real-time motion state. This component updates in real time with changes in acceleration, achieving dynamic adaptation of the threshold. S53 sets the specific value of the adjustable empirical bias term according to the collision detection sensitivity requirements of different operating scenarios. The value range is from 0.1 to 0.5. The static reference threshold, dynamic adjustment component and empirical bias term are superimposed and calculated according to preset rules to obtain the dynamic threshold that changes in real time for each joint. The calculation process must ensure real-time performance, and the calculation delay for each frame of data shall not exceed 50 microseconds. S54 extracts the absolute value of the real-time current change rate of each joint from the current change rate calculation unit. It compares the joint number with the corresponding real-time dynamic threshold one by one. The comparison order is consistent with the data acquisition order. When the absolute value of the current change rate of a certain joint exceeds its corresponding real-time dynamic threshold, a primary contact or collision alarm signal including the joint number, trigger time, current change rate value and dynamic threshold value is immediately generated to ensure the timeliness and integrity of the alarm signal and provide detailed basis for subsequent collision confirmation.

[0041] A robot collision detection and compliant control method based on joint current feedback is disclosed. This method is implemented through a robot collision detection and compliant control system based on joint current feedback, comprising: a joint dynamics model construction unit, used to establish the joint space dynamics model of the controlled robot, determine the correlation between various mechanical parameters, and provide a theoretical basis for analysis; a multi-channel current signal acquisition unit, connected to the robot servo driver, to acquire the three-phase current feedback signals of each joint motor in real time at a high sampling rate, and to classify and store them; and a current change rate calculation unit, which receives the data output by the current signal acquisition unit and obtains the current of each joint through numerical differentiation. The system includes: a rate-of-change time-series data acquisition unit (POC) for acquiring and processing real-time acceleration information of each joint via inverse dynamics model calculation or direct reading of controller commands; a dynamic threshold calculation and primary collision determination unit for receiving acceleration information and current rate-of-change data, calculating and comparing real-time dynamic thresholds, and triggering a primary alarm; and a collision level confirmation and control response unit for receiving alarm signals from the primary determination unit, confirming valid collision events and classifying them into levels through duration window determination and multi-joint information fusion, and outputting corresponding compliant control commands based on the level. All units establish a bidirectional communication connection via a data bus for real-time signal transmission and processing.

[0042] like Figure 5The image shows the visual operating interface of the robot collision detection and compliant control system based on joint current feedback. This interface integrates functional modules such as system operating parameters, multi-dimensional signal monitoring, collision level determination, compliant control execution, and historical alarm records, intuitively presenting the actual operating conditions of the technical solution of this invention. The top of the interface displays the system's basic operating indicators. The system's cumulative running time is 7 days, 04:26:38, the current sampling frequency is set to 1.00kHz, the real-time running latency is 2.35ms, and the overall system status is judged to be normal, conforming to the hardware design specifications of this invention with adjustable sampling rate from 1kHz to 10kHz and low latency operation. The core area of ​​the interface sets up multiple monitoring modules to synchronously monitor all joints from J1 to J6 of the robot. The joint current and dynamic threshold comparison curves are generated based on the dynamic threshold algorithm of this invention. This threshold is adaptively adjusted by combining static reference values, joint acceleration components, and adjustable bias terms, which can accurately distinguish between normal current fluctuations caused by robot inertial motion and current anomalies caused by external collisions. The interface synchronously displays motion data such as joint acceleration, position, and attitude, forming data linkage with the two acceleration acquisition methods of this invention: inverse dynamics calculation and controller command reading. The collision level is divided into three levels, strictly adhering to the duration window determination, multi-joint information fusion mechanism, and collision quantification formula of this invention. The compliant control panel displays an adaptive impedance control mode with a target stiffness of 120 N / m and a response time of 12.6 ms. The alarm log below records each collision event, noting the collision time, abnormal joint, collision level, peak current, duration, and graded handling strategies such as automatic recovery, deceleration, and emergency stop. It also provides real-time feedback on hardware status, including bus communication, packet loss rate, and servo driver status. This interface fully demonstrates that this invention effectively reduces false alarm rates, improves collision detection response speed, and provides a comprehensive graded compliant control strategy, meeting the core requirements for safe and stable robot operation in human-robot collaborative scenarios.

[0043] like Figure 6The image shows the backend code running interface corresponding to the robot collision detection and compliant control method based on joint current feedback of the present invention. The code adopts a modular architecture design, and the code components correspond one-to-one with the functional units of the control system of the present invention. It includes core code files such as main controller, current signal acquisition, threshold determination, compliant control, and joint dynamics model. Each module achieves collaborative operation through a data bus to realize the complete technical process of S1 to S6 of the present invention. The execution log output from the code console fully reconstructs the execution logic of a single detection and control operation: After system initialization, the current acquisition module is called to acquire the three-phase current of joint 3 at a high sampling rate, obtaining current data [2.35, -1.87, -0.48] A, corresponding to the joint current acquisition step of this invention; subsequently, the system combines the joint acceleration parameters to complete dynamic threshold calculation and update the threshold to 1.26; after comparing the current change rate values, the joint current change rate of 1.58 exceeds the threshold, triggering a primary collision alarm; the system initiates the duration window and multi-joint information fusion mechanism to complete verification, determining it to be a level two moderate collision; finally, the compliant control service is activated, reducing the joint stiffness to 38% and adjusting the damping to 4%, successfully issuing the compliant control command. The entire code's computational logic strictly follows the joint dynamics model, current change rate calculation formula, dynamic threshold calculation rules, and collision classification judgment algorithm of this invention. The log feedback shows extremely low overall computational latency, demonstrating the excellent real-time computational performance of the algorithm of this invention. The code architecture is functionally decoupled and logically clear, and can stably complete the entire process of current acquisition, signal differentiation, threshold comparison, collision verification, and compliant control command output. This fully demonstrates that the algorithm logic of the technical solution of this invention is reliable and has strong engineering applicability, and can stably support the real-time operation of collision detection and compliant control of industrial robots for a long time.

[0044] A robot collision detection and compliant control method based on joint current feedback abandons the traditional static amplitude detection approach and instead uses the rate of change of current as the core observable, significantly improving the response speed of collision detection. It can capture sudden current changes in the initial stage of a collision, enabling early warning. Simultaneously, it introduces a dynamic threshold closely related to the robot's real-time motion state, combined with multi-joint information fusion and a duration window determination mechanism, ensuring detection sensitivity during precise operations while effectively avoiding false alarms caused by motion inertia. The hardware employs a high sampling rate, high resolution, multi-channel synchronous acquisition module with anti-electromagnetic interference design. The software utilizes optimized numerical calculation algorithms and adaptive parameter configuration to further improve detection stability and adaptability, eliminating the need for additional external sensors and balancing low cost and high reliability.

[0045] For the problem of response lag, this method replaces the absolute value of the current with the rate of change of the current, shifting the detection triggering time from the current accumulation exceeding the threshold stage to the initial stage of the current change. Combined with high sampling rate data acquisition and fast processing, it significantly shortens the time difference between collision recognition and response. For the problem of false alarms, the dynamic threshold can automatically adjust the judgment boundary according to the real-time acceleration of the joints. The threshold is increased during violent movements such as robot acceleration and deceleration to tolerate current fluctuations caused by normal inertial forces, and the threshold is decreased in uniform or static states to maintain high sensitivity to small external forces. Combined with multi-joint information fusion and duration window verification, transient interference signals are further filtered, fundamentally reducing the probability of false triggering and achieving a balance between detection accuracy and robustness.

[0046] In the description of this invention, it should be noted that, unless otherwise specified and limited, the terms "set," "install," "connect," "link," and "fix" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0047] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A robot collision detection and compliant control method based on joint current feedback, characterized in that, Includes the following steps: S1. Construct a joint space dynamics model of the controlled robot, determine the relationship between joint angle vector, mass matrix, Coriolis force and centrifugal force, gravity and friction terms, and provide support for the theoretical analysis of joint torque and current and the preliminary setting of dynamic threshold. S2, through the servo driver of the robot control system, collects the three-phase current feedback signals of each joint motor in real time with a high sampling rate; S3 performs numerical differentiation on the acquired current signal to obtain the current change rate of each joint in real time. S4, by performing a second differential operation on the joint position signal through the robot inverse dynamics model, or by directly reading the joint expected acceleration command output by the controller planning module, to obtain the real-time acceleration information of each joint; S5, design a dynamic threshold associated with the robot's real-time motion state. This threshold includes a static reference current change rate threshold, a component associated with the real-time joint acceleration, and an adjustable empirical bias term. Compare the absolute value of the real-time calculated current change rate with the dynamic threshold to trigger a primary contact or collision alarm. S6 introduces a duration window determination mechanism and a multi-joint information fusion mechanism to verify the primary alarm. After confirming the valid collision event, the collision event is classified according to the overshoot amplitude of the current change rate, the abnormal duration and the number of joints involved.

2. The robot collision detection and compliant control method based on joint current feedback according to claim 1, characterized in that, The joint space dynamics model satisfies: ,in This is the joint torque vector. Joint angle vector, The mass matrix depends on the joint angle. This is the matrix of Coriolis force and centrifugal force coefficients. The joint angular velocity vector. The joint angular acceleration vector. This is a gravity term vector that depends on the joint angle. The dynamic threshold is a frictional force vector dependent on joint angular velocity; the dynamic threshold satisfies: ,in This is a dynamic threshold vector that changes over time. The threshold for the rate of change of the static reference current. For acceleration influence coefficient, This is the real-time acceleration vector of the joint. This is an adjustable empirical bias term.

3. The robot collision detection and compliant control method based on joint current feedback according to claim 1, characterized in that, The rate of change of current is calculated using the following formula: ,in For the first The first joint The rate of change of current at time t. For the first The first joint The current value collected at any given time. For the first The first joint The current value collected at any given time. The sampling period is the current signal period; the joint acceleration, when calculated using the inverse dynamics model, satisfies the following: ,in The joint angular acceleration vector. For the mass matrix The inverse matrix, This is the joint torque vector. This is the matrix of Coriolis force and centrifugal force coefficients. The joint angular velocity vector. The gravity term vector, This is the vector of the friction force term.

4. The robot collision detection and compliant control method based on joint current feedback according to claim 1, characterized in that, The collision level classification satisfies: ,in This is a quantification value for the collision level. This is the overshoot weighting coefficient for the rate of change of current. For the first The first joint The rate of change of current at time t. For dynamic threshold vectors, The weighting coefficient is the duration of the anomaly. For the duration of the abnormality, As the baseline time threshold, For the weighting coefficients involving the number of joints, The number of joints exhibiting abnormality; the acceleration influence coefficient in the dynamic threshold satisfies: ,in For acceleration influence coefficient, This is the proportionality coefficient. For the mass matrix The traces.

5. The robot collision detection and compliant control method based on joint current feedback according to claim 1, characterized in that, The multi-joint information fusion mechanism satisfies: ,in This is the fusion determination value. The total number of joints. For the first The weighting coefficient of each joint, For the first The first joint The rate of change of current at time t. For the first Dynamic values ​​of each joint. The sign function is used; the duration window determination satisfies: ,in The value is determined within the time window. The moment when the alarm is first triggered. The number of control cycles included in the time window. For the first The first joint The rate of change of current at time t. For the first Dynamic threshold of each joint It is a symbolic function.

6. The robot collision detection and compliant control method based on joint current feedback according to claim 1, characterized in that, The servo driver uses a multi-channel synchronous acquisition module, which supports parallel acquisition of three-phase current signals. The sampling rate is adjustable from 1kHz to 10kHz, the resolution of the acquisition channel is not less than 16 bits, and it has a shielded structure design to resist electromagnetic interference. The numerical differentiation operation of the current signal adopts the center difference algorithm, and the operation step size is consistent with the sampling period. The control cycle number of the duration window ranges from 2 to 5. When fusing multi-joint information, the weight coefficient of each joint is allocated according to the force sensitivity of the joint in the robot's motion, and the value ranges from 0.1 to 1.

0. In the dynamic threshold, the static reference current change rate threshold is determined according to the current fluctuation characteristics of the robot when it is running without load. The adjustable empirical bias term is adjusted online according to the collision detection sensitivity requirements of different operation scenarios.

7. The robot collision detection and compliant control method based on joint current feedback according to claim 1, characterized in that, S3 includes: receiving the collected three-phase current feedback signals of each joint motor; classifying and organizing the current signals according to the joint number to establish a time sequence of the current signal for each joint; determining the interval between adjacent sampling points for numerical differentiation operation; extracting the current sampling values ​​of two consecutive adjacent moments for each joint based on the time sequence of the current signals; performing calculations on the extracted adjacent moment current sampling values ​​according to preset differentiation operation rules to obtain the current change amount at the corresponding moment for each joint; and storing the calculated current change amount in association with the joint number according to the time sequence to form the time sequence data of the current change rate of each joint, providing observations for collision detection.

8. The robot collision detection and compliant control method based on joint current feedback according to claim 1, characterized in that, S4 includes: If the inverse dynamics model is used to calculate the joint acceleration, the real-time position signals of each joint of the robot are first obtained, and the position signals are continuously sampled at preset time intervals to form position time series data. The first-order differential operation is performed on the collected position timing data to obtain the angular velocity timing signal of each joint. Then, the second-order differential operation is performed on the angular velocity timing signal to obtain the initial value of the joint angular acceleration. If the desired acceleration command is read directly, a data communication link is established with the controller planning module to receive the desired acceleration command data of each joint output by the module in real time. The acceleration data obtained by the two methods are time-aligned to remove the delay deviation in the data transmission process, forming a set of real-time acceleration information for each joint.

9. The robot collision detection and compliant control method based on joint current feedback according to claim 1, characterized in that, S5 includes: Obtain the mass matrix correlation parameters in the robot dynamics model, determine the initial value of the acceleration influence coefficient in combination with the joint motion characteristics, and set the static reference current change rate threshold based on the current noise level when the system has no external contact; collect the acceleration information of each joint in real time, calculate the absolute value of acceleration and the acceleration influence coefficient to obtain the dynamic adjustment component associated with the motion; Based on the collision detection requirements of the work scenario, the specific value of the adjustable empirical bias term is set. The static reference threshold, dynamic adjustment component and empirical bias term are superimposed to obtain the real-time dynamic threshold. The absolute value of the real-time current change rate of each joint is extracted and compared with the corresponding real-time dynamic threshold one by one. When the absolute value of the current change rate exceeds the dynamic threshold, a primary contact or collision alarm signal is generated and output.

10. A robot collision detection and compliant control method based on joint current feedback according to any one of claims 1-9, characterized in that, This method is implemented through a robot collision detection and compliance control system based on joint current feedback, including: The joint dynamics model building unit is used to establish the joint space dynamics model of the controlled robot, determine the correlation between various mechanical parameters, and provide a theoretical basis for analysis. The multi-channel current signal acquisition unit connects to the robot servo driver to acquire the three-phase current feedback signals of each joint motor in real time with a high sampling rate, and then classifies and stores them. The current change rate calculation unit receives data output from the current signal acquisition unit and obtains the current change rate timing data of each joint through numerical differentiation calculation. The joint acceleration acquisition unit collects and processes real-time acceleration information of each joint through two methods: inverse dynamics model calculation or direct reading of controller commands. The dynamic threshold calculation and collision primary judgment unit receives acceleration information and current change rate data, calculates the real-time dynamic threshold and compares it, and triggers a primary alarm. The collision level confirmation and control response unit receives alarm signals from the primary judgment unit, confirms valid collision events and classifies them into levels through duration window judgment and multi-joint information fusion, and outputs corresponding compliant control commands according to the level. Each unit establishes a bidirectional communication connection through the data bus to transmit and process signals in real time.

Images