A multifunctional humanoid joint motor driving system and a control method thereof

Abstract

A multifunctional humanoid joint motor drive and control system and its control method belong to the field of robotics technology. To address the problems of low integration, insufficient torque control bandwidth leading to poor compliant interaction, and lack of multimodal perception fusion in existing humanoid joint motor drive and control systems, the system includes: a frameless torque motor module for generating joint driving torque; a transmission module, located between the output end of the frameless torque motor module and the joint load, including a precision reducer and a controllable damper; a multimodal sensing fusion module, including: an encoder system for position feedback; a force / torque sensor for directly measuring the joint output torque; a temperature sensor for temperature monitoring; a power drive module for driving the frameless torque motor module; and an embedded intelligent control module for performing hierarchical control of the current inner loop, torque middle loop, velocity outer loop, and position outer loop, and running collision detection and thermal management strategies. It is intended for use in humanoid robots.

Description

Technical Field

[0001] This invention relates to a multifunctional humanoid joint motor drive and control system and its control method, belonging to the field of robotics technology. Background Technology

[0002] Humanoid robots, as important carriers of embodied intelligence, require their joint actuators to possess multiple performance characteristics, including high torque density, high bandwidth response, low torque ripple, and safe and compliant interaction. Currently, mainstream humanoid robot joint solutions include Tesla's Optimus' combination of rotary and linear actuators, UBTECH's Walker's all-servo joint solution, and Boston Dynamics' Atlas' hydraulic drive solution. Among these, motor-driven solutions have become the industry mainstream due to their high control precision, fast response speed, and ease of intelligentization.

[0003] In existing highly integrated joint drive and control solutions, the rigidity and flexibility of the transmission link are fixed and cannot be continuously adjusted online. There is an inherent contradiction between achieving precise positioning and compliant interaction: the bandwidth of the elastic element solution is limited, and the direct drive solution lacks compliance. It is difficult for the two to achieve seamless switching on the same joint.

[0004] However, existing humanoid robot joint motor drive systems still have the following significant shortcomings:

[0005] 1. Separation of drive and control, low integration: The motor, reducer, encoder, driver and controller are independent modules, the wiring harness is complicated, the assembly process is more complicated, which increases the weight of the whole machine and the failure rate.

[0006] 2. Low torque control bandwidth and poor compliant interaction capability: Most systems only achieve position or velocity closed loops and lack a high-bandwidth torque inner loop, making it difficult to respond quickly to unexpected contact and posing safety hazards.

[0007] 3. Insufficient multimodal perception fusion: It only provides position feedback and lacks comprehensive perception of temperature, current harmonics, transmission error, collision force, etc., making it difficult to achieve fault prediction and performance optimization.

[0008] 4. Lack of joint-level edge intelligence: It relies on a central controller to complete all calculations. The joints themselves do not have local decision-making capabilities and cannot maintain a safe state when communication is delayed or interrupted.

[0009] 5. Weak thermal management and efficiency optimization: The lack of a current adaptive limiting strategy based on thermal models makes it easy to trigger over-temperature protection or shorten the motor life.

[0010] Therefore, there is an urgent need for a new type of humanoid joint motor drive and control system with high integration, strong force control performance, comprehensive perception, edge intelligence and thermal management capabilities. Summary of the Invention

[0011] The purpose of this invention is to address the problems of low integration, insufficient torque control bandwidth leading to poor compliant interaction capabilities, and lack of multimodal perception fusion in existing humanoid joint motor drive and control systems, and to provide a multifunctional humanoid joint motor drive and control system and its control method.

[0012] The present invention discloses a multifunctional humanoid joint motor drive control system, which includes:

[0013] The frameless torque motor module adopts an external rotor and internal stator topology. The stator has a fractional slot concentrated winding structure to generate joint driving torque.

[0014] The transmission module, located between the output end of the frameless torque motor module and the joint load, includes a precision reducer and a controllable damper. The controllable damper is a magnetorheological fluid damper, used to dynamically adjust the equivalent damping coefficient of the transmission link.

[0015] The multimodal sensing fusion module includes at least: an encoder system for position feedback; a force / torque sensor for directly measuring joint output torque; and a temperature sensor for temperature monitoring.

[0016] A power drive module is used to drive the frameless torque motor module;

[0017] The embedded intelligent control module is used to perform hierarchical control of the inner current loop, middle torque loop, outer speed loop, and outer position loop, and to run collision detection and thermal management strategies.

[0018] Preferably, the controllable damper of the transmission module consists of a magnetorheological fluid cavity and an electromagnetic coil. The magnetorheological fluid cavity is disposed between the output shaft of the reducer and the joint housing, and has an annular thin-layer structure with a cavity gap of 0.5 to 1.0 mm. The apparent viscosity of the magnetorheological fluid is continuously adjusted within 0.5 milliseconds by adjusting the current of the electromagnetic coil, thereby dynamically changing the equivalent damping coefficient of the transmission link. The adjustable range of the equivalent damping coefficient is 0.01 to 50 Nm / s per radian.

[0019] Preferably, the encoder system of the multimodal sensing fusion module is a dual encoder system, including a 19-bit magnetic encoder on the motor side and a 17-bit absolute magnetic encoder on the reducer output side, which is used to calculate the actual transmission ratio deviation and backlash angle of the reducer in real time, compensate for transmission errors, and detect the wear degree of the reducer tooth surface.

[0020] The multimodal sensing fusion module also includes a current harmonic analysis module. The embedded intelligent control module performs a fast Fourier transform on the three-phase current collected by the power drive module and extracts the amplitude values ​​of the 5th, 7th, 11th and 13th current harmonics in real time. When the deviation of the current harmonic amplitude exceeds a preset threshold, an early warning signal of demagnetization fault or winding inter-turn short circuit is issued.

[0021] Preferably, the multimodal sensing fusion module further includes an inertial measurement unit, which is mounted on the joint housing and integrates a three-axis accelerometer and a three-axis gyroscope for detecting the joint's vibration acceleration and angular velocity.

[0022] Preferably, the embedded intelligent control module is equipped with a dual-core microcontroller, wherein the main core runs a current loop and a position, speed, and torque servo loop, and the slave core runs a communication protocol stack and intelligent algorithms;

[0023] The current inner loop bandwidth is not less than 5kHz, and it is based on magnetic field orientation control and space vector modulation, using a PI controller with back EMF feedforward compensation and dead zone compensation.

[0024] The torque loop bandwidth is not less than 1kHz, and the signal of the force / torque sensor is used as feedback. A PD controller with friction torque feedforward compensation and reducer hysteresis compensation is adopted.

[0025] The bandwidth of the outer velocity loop and the outer position loop is not less than 200Hz. The position loop adopts an S-curve trajectory generator plus a PD controller; the velocity loop adopts a PI controller with inertia feedforward.

[0026] The embedded intelligent control module runs a collision detection algorithm based on a generalized momentum observer. It estimates the expected torque using motor current, joint position and reducer model, and compares it with the measured torque of the force / torque sensor. When the residual exceeds the adaptive threshold, it determines that a collision has occurred and switches to zero torque mode or impedance buffer mode in less than 2 milliseconds.

[0027] In collision detection, residuals are calculated. When the residual exceeds the adaptive threshold When a collision is detected, it is determined that a collision has occurred; among them, The torque measured by the force / torque sensor. The expected torque is estimated based on motor current, joint position, and reducer model. The mean of the residuals, The standard deviation of the residuals. These are preset coefficients.

[0028] Preferably, the embedded intelligent control module runs an online winding temperature estimation algorithm based on a thermal equivalent circuit model, uses the current square integral and ambient temperature as inputs to estimate the temperature of the hottest spot in the winding in real time, and dynamically adjusts the current limiting curve according to the estimated temperature to achieve flexible derating protection.

[0029] The online winding temperature estimation algorithm adopts a two-node RC thermal equivalent circuit model, and its current limiting satisfies ,in, For limiting current, Rated current, It is a continuously decreasing function. To estimate the winding temperature, This is the rated temperature.

[0030] Preferably, the system supports seamless switching between three operating modes: high stiffness position mode, high bandwidth torque mode, and impedance control mode;

[0031] Wherein, the joint output torque under the impedance control mode satisfies:

[0032] ;

[0033] in, This represents the total torque actually output by the joint. For virtual stiffness, For virtual damping, For feedforward torque, For the target joint position, This represents the current actual position of the joint.

[0034] Preferably, the power drive module is a GaN-based wide bandgap power drive module, adopting a three-phase full-bridge inverter topology, with a bus voltage of 48V, a continuous phase current output capability of not less than 30 amps, a peak current of not less than 60 amps, a switching frequency of 40kHz to 100kHz, and integrated hardware overcurrent protection circuit, overvoltage protection circuit, undervoltage protection circuit, and short-circuit protection circuit with a response time of less than 1 microsecond.

[0035] Preferably, the system is connected to the whole-body motion controller via an EtherCAT industrial Ethernet bus with a communication cycle of 1 millisecond and supports distributed clock synchronization; it also supports CAN-FD bus as a backup communication channel, automatically switching to the CAN-FD channel to maintain basic safety control when the EtherCAT link is interrupted.

[0036] The present invention discloses a control method for a multifunctional humanoid joint motor drive system, comprising:

[0037] Step 1: Determine the working mode and identify the target working mode, which includes high stiffness position mode, high bandwidth torque mode and impedance control mode.

[0038] Step 2: Damping coefficient setting. The equivalent damping coefficient of the controllable damper is set according to the target operating mode: the maximum value is taken in the high-stiffness position mode. In impedance control mode, according to virtual damping Configure the mapping, and take the minimum value in high bandwidth torque mode. ;

[0039] Step 3: Damping switching. The damping coefficient is switched within 0.5 milliseconds by adjusting the electromagnetic coil current of the magnetorheological fluid damper.

[0040] Step 4: Four-loop control. The embedded controller of the joint module synchronously runs the inner current loop, the middle torque loop, the outer speed loop, and the outer position loop to perform nested control of position-speed-torque-current four-loop closed loops.

[0041] Step 5: Collision safety. The host controller sends position, velocity and feedforward torque commands through the EtherCAT bus at a 1kHz cycle; it collects multimodal sensor data in real time, runs a collision detection algorithm based on a generalized momentum observer, calculates the residual, and switches to zero torque mode or impedance buffer mode within 2 milliseconds when the residual exceeds the adaptive threshold.

[0042] Step 6: Thermal protection. Real-time estimation of winding temperature. When the winding temperature exceeds 0.8 times the rated temperature value, flexible derating protection is implemented, and the current limiting curve is dynamically adjusted.

[0043] Advantages of the present invention: The multifunctional humanoid joint motor drive and control system and its control method proposed in this invention have the following advantages:

[0044] 1. High integration: This invention adopts a "five-in-one" architecture, integrating the frameless torque motor, reducer, sensor, driver, and controller into a single module. Compared with the traditional split solution, the module size is reduced by about 40%, the number of wiring harnesses is reduced by about 70% (each joint only requires one power cable and one EtherCAT communication cable), and the overall assembly time is reduced by about 50%.

[0045] 2. Wideband Torque Control and Compliant Interaction: This invention features a torque control bandwidth of no less than 1 kHz, which, combined with a magnetorheological fluid controllable damper, enables continuous online adjustment of rigidity and flexibility characteristics with a time constant of less than 1 millisecond. It combines the force transparency of a series elastic actuator with the high bandwidth of a direct-drive actuator, achieving seamless switching between precise positioning and safe, compliant interaction on the same joint.

[0046] 3. Joint-level collision safety: The local collision detection response time based on the generalized momentum observer is less than 2 milliseconds, without relying on the upper controller, meeting the collision response time requirements in the ISO / TS 15066 collaborative robot safety standard.

[0047] 4. Full-state perception and diagnosis: Real-time compensation for transmission errors is achieved through differential measurement with dual encoders, improving accuracy by about one order of magnitude, and wear of the reducer can be monitored; early warning of demagnetization and inter-turn short circuit is achieved through current harmonic analysis; flexible derating protection is achieved through winding thermal model to avoid hard protection cut-off interruption of movement.

[0048] 5. High efficiency and compact GaN drive: By using GaN power devices, the area of ​​the drive board is reduced by about 50%, the switching loss is reduced by about 30%, and the audible noise frequency is pushed to the ultrasonic range (above 40kHz), improving the comfort of human-machine coexistence environment.

[0049] 6. Communication Redundancy and Edge Intelligence: The system adopts a dual-bus redundancy architecture of EtherCAT and CAN-FD to ensure communication reliability; the joint-level embedded intelligent controller can independently execute collision safety response and thermal management decisions without relying on the central controller, thus improving system robustness. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the structure of the multifunctional humanoid joint motor drive control system described in this invention;

[0051] Figure 2 This is a schematic diagram of the cross-sectional structure of a frameless torque motor with an external rotor and an internal stator.

[0052] Figure 3 This is a schematic diagram of a transmission module with switchable rigidity and flexibility.

[0053] Figure 4 This is a configuration diagram of the multimodal sensing fusion module;

[0054] Figure 5 This is a circuit topology diagram of a GaN-based power drive module;

[0055] Figure 6 It is a hierarchical control architecture block diagram;

[0056] Figure 7 This is a flowchart of the collision detection and safety response process;

[0057] Figure 8 This is a schematic diagram of the winding thermal model and flexible derating protection strategy;

[0058] Figure 9 This is a diagram of a multi-joint coordinated communication architecture. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0061] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0062] Example 1:

[0063] The following is combined Figures 1-9 This embodiment describes a multifunctional humanoid joint motor drive control system, which includes:

[0064] The frameless torque motor module adopts an external rotor and internal stator topology. The stator has a fractional slot concentrated winding structure to generate joint driving torque.

[0065] The transmission module, located between the output end of the frameless torque motor module and the joint load, includes a precision reducer and a controllable damper. The controllable damper is a magnetorheological fluid damper, used to dynamically adjust the equivalent damping coefficient of the transmission link.

[0066] The multimodal sensing fusion module includes at least: an encoder system for position feedback; a force / torque sensor for directly measuring joint output torque; and a temperature sensor for temperature monitoring.

[0067] A power drive module is used to drive the frameless torque motor module;

[0068] The embedded intelligent control module is used to perform hierarchical control of the inner current loop, middle torque loop, outer speed loop, and outer position loop, and to run collision detection and thermal management strategies.

[0069] Furthermore, the controllable damper of the transmission module consists of a magnetorheological fluid cavity and an electromagnetic coil. The magnetorheological fluid cavity is located between the output shaft of the reducer and the joint housing, and has an annular thin-layer structure with a cavity gap of 0.5 to 1.0 mm. By adjusting the current of the electromagnetic coil, the apparent viscosity of the magnetorheological fluid can be continuously adjusted within 0.5 milliseconds, thereby dynamically changing the equivalent damping coefficient of the transmission link. The adjustable range of the equivalent damping coefficient is 0.01 to 50 Nm / s per radian.

[0070] Furthermore, the encoder system of the multimodal sensing fusion module is a dual encoder system, including a 19-bit magnetic encoder on the motor side and a 17-bit absolute magnetic encoder on the reducer output side. It is used to calculate the actual transmission ratio deviation and backlash angle of the reducer in real time, compensate for transmission errors, and detect the wear degree of the reducer tooth surface.

[0071] The multimodal sensing fusion module also includes a current harmonic analysis module. The embedded intelligent control module performs a fast Fourier transform on the three-phase current collected by the power drive module and extracts the amplitude values ​​of the 5th, 7th, 11th and 13th current harmonics in real time. When the deviation of the current harmonic amplitude exceeds a preset threshold, an early warning signal of demagnetization fault or winding inter-turn short circuit is issued.

[0072] Furthermore, the multimodal sensing fusion module also includes an inertial measurement unit, which is mounted on the joint housing and integrates a three-axis accelerometer and a three-axis gyroscope for detecting the joint's vibration acceleration and angular velocity.

[0073] Furthermore, the embedded intelligent control module is equipped with a dual-core microcontroller, wherein the main core runs a current loop and position, speed, and torque servo loops, while the slave core runs a communication protocol stack and intelligent algorithms;

[0074] The current inner loop bandwidth is not less than 5kHz, and it is based on magnetic field orientation control and space vector modulation, using a PI controller with back EMF feedforward compensation and dead zone compensation.

[0075] The torque loop bandwidth is not less than 1kHz, and the signal of the force / torque sensor is used as feedback. A PD controller with friction torque feedforward compensation and reducer hysteresis compensation is adopted.

[0076] The bandwidth of the outer velocity loop and the outer position loop is not less than 200Hz. The position loop adopts an S-curve trajectory generator plus a PD controller; the velocity loop adopts a PI controller with inertia feedforward.

[0077] The embedded intelligent control module runs a collision detection algorithm based on a generalized momentum observer. It estimates the expected torque using motor current, joint position, and reducer model, and compares it with the measured torque of the force / torque sensor. When the residual exceeds the adaptive threshold, a collision is determined, and the module switches to zero torque mode or impedance buffer mode within less than 2 milliseconds.

[0078] In collision detection, residuals are calculated. When the residual exceeds the adaptive threshold When a collision is detected, it is determined that a collision has occurred; among them, The torque measured by the force / torque sensor. The expected torque is estimated based on motor current, joint position, and reducer model. The mean of the residuals, The standard deviation of the residuals. These are preset coefficients.

[0079] Furthermore, the embedded intelligent control module runs an online winding temperature estimation algorithm based on a thermal equivalent circuit model. It uses the current square integral and ambient temperature as inputs to estimate the temperature of the hottest spot in the winding in real time, and dynamically adjusts the current limiting curve according to the estimated temperature to achieve flexible derating protection.

[0080] The online winding temperature estimation algorithm adopts a two-node RC thermal equivalent circuit model, and its current limiting satisfies ,in, For limiting current, Rated current, It is a continuously decreasing function. To estimate the winding temperature, This is the rated temperature.

[0081] Furthermore, the system supports seamless switching between three operating modes: high stiffness position mode, high bandwidth torque mode, and impedance control mode.

[0082] Wherein, the joint output torque under the impedance control mode satisfies:

[0083] ;

[0084] in, This represents the total torque actually output by the joint. For virtual stiffness, For virtual damping, For feedforward torque, For the target joint position, This represents the current actual position of the joint.

[0085] Furthermore, the power drive module is a GaN-based wide bandgap power drive module, adopting a three-phase full-bridge inverter topology. Its bus voltage is 48V, the continuous output capability of phase current is not less than 30 amps, the peak current is not less than 60 amps, the switching frequency is 40kHz to 100kHz, and it integrates hardware overcurrent protection circuit, overvoltage protection circuit, undervoltage protection circuit and short circuit protection circuit with a response time of less than 1 microsecond.

[0086] Furthermore, the system is connected to the full-body motion controller via the EtherCAT industrial Ethernet bus, with a communication cycle of 1 millisecond and support for distributed clock synchronization; it also supports the CAN-FD bus as a backup communication channel, automatically switching to the CAN-FD channel to maintain basic safety control when the EtherCAT link is interrupted.

[0087] Example 2:

[0088] The multifunctional humanoid joint motor drive control method described in this embodiment includes:

[0089] Step 1: Determine the working mode and identify the target working mode, which includes high stiffness position mode, high bandwidth torque mode and impedance control mode.

[0090] Step 2: Damping coefficient setting. The equivalent damping coefficient of the controllable damper is set according to the target operating mode: the maximum value is taken in the high-stiffness position mode. In impedance control mode, according to virtual damping Configure the mapping, and take the minimum value in high bandwidth torque mode. ;

[0091] Step 3: Damping switching. The damping coefficient is switched within 0.5 milliseconds by adjusting the electromagnetic coil current of the magnetorheological fluid damper.

[0092] Step 4: Four-loop control. The embedded controller of the joint module synchronously runs the inner current loop, the middle torque loop, the outer speed loop, and the outer position loop to perform nested control of position-speed-torque-current four-loop closed loops.

[0093] Step 5: Collision safety. The host controller sends position, velocity and feedforward torque commands through the EtherCAT bus at a 1kHz cycle; it collects multimodal sensor data in real time, runs a collision detection algorithm based on a generalized momentum observer, calculates the residual, and switches to zero torque mode or impedance buffer mode within 2 milliseconds when the residual exceeds the adaptive threshold.

[0094] Step 6: Thermal protection. Real-time estimation of winding temperature. When the winding temperature exceeds 0.8 times the rated temperature value, flexible derating protection is implemented, and the current limiting curve is dynamically adjusted.

[0095] This invention proposes a multifunctional humanoid joint motor drive and control system, which adopts a "five-in-one" highly integrated architecture, integrating the following five functional sub-modules into a single joint module: a high torque density frameless torque motor module; a transmission module with switchable rigidity and flexibility characteristics; a multimodal sensing fusion module; a GaN-based wide bandgap power drive module; and an embedded control module with joint-level edge intelligence.

[0096] like Figure 1 The diagram shown is an exploded view of the overall structure of the multi-functional humanoid joint motor drive control system, demonstrating the assembly relationship of the various sub-modules of the "five-in-one" module.

[0097] The system supports seamless switching between three working modes: high-rigidity position mode (for precise positioning), high-bandwidth torque mode (for compliant interaction and collision safety), and impedance control mode (for force-position hybrid control). It achieves a four-closed-loop nested servo system with position, speed, torque, and current through a hierarchical control architecture.

[0098] High torque density frameless torque motor module:

[0099] The motor module (1) adopts a frameless permanent magnet synchronous torque motor with an external rotor and internal stator topology. The stator (1a) has a fractional slot concentrated winding structure with slot-pole matching of 12 slots and 14 poles (suitable for medium-speed joints such as shoulder / elbow) or 18 slots and 20 poles (suitable for low-speed, high-torque joints such as hip / knee) to achieve low cogging torque and high winding coefficient. The stator core is made of 0.2 mm thick ultra-thin silicon steel sheets (grade 20SW1200 or amorphous alloy) stamped and stacked to reduce high-frequency iron loss. The winding adopts flat copper wire (cross-sectional width-to-height ratio not less than 4:1) vertical winding process to increase the slot fill factor to more than 75% and enhance the heat dissipation path.

[0100] The rotor (1b) has an outer annular structure with a V-shaped built-in permanent magnet array mounted on its inner wall. The permanent magnet material is N52UH sintered neodymium iron boron (remanence Br not less than 1.42T, coercivity Hcj not less than 1990kA / m, maximum operating temperature 180 degrees Celsius). The opening angle of the V-shaped arrangement is 100 to 130 degrees (typically 120 degrees), and each pole consists of two trapezoidal magnets (6 to 10 mm wide, 3 to 5 mm thick). Compared with surface-mounted magnets, this increases the reluctance torque component by approximately 15% to 20%, improving the overall torque density without increasing the amount of permanent magnets used. The outer wall of the rotor is equipped with a thin-walled carbon fiber sheath (0.3 to 0.5 mm thick) to prevent the permanent magnets from scattering during high-speed operation.

[0101] like Figure 2 The diagram shown is a cross-sectional structure diagram of a frameless torque motor with an external rotor and an internal stator, demonstrating the V-shaped built-in permanent magnet arrangement, fractional slot concentrated winding, and flat copper wire vertical winding structure.

[0102] Motor design specifications: Continuous output torque not less than 15 Nm (hip / knee type) or 5 Nm (shoulder / elbow type); peak torque 3 to 4 times the continuous torque; torque constant Kt not less than 0.8 Nm per ampere; torque fluctuation (cogging torque + electromagnetic torque pulsation) not exceeding 1.5% of rated torque. Motor outer diameter not exceeding 90 mm (hip / knee type) or 60 mm (shoulder / elbow type), axial length not exceeding 30 mm.

[0103] Transmission modules with switchable rigidity and flexibility:

[0104] like Figure 3 The diagram shows a transmission module with switchable rigidity and flexibility, illustrating the assembly relationship and working principle of the reducer (2a) and the magnetorheological fluid controllable damper (2b).

[0105] The transmission module (2) is located between the motor output and the joint load, and includes a precision reducer (2a) and a controllable damper (2b).

[0106] The reducer (2a) is a harmonic reducer (suitable for high reduction ratio scenarios, reduction ratio 80:1 to 120:1, for hip / knee / ankle joints) or a precision planetary reducer (suitable for medium reduction ratio scenarios, reduction ratio 10:1 to 50:1, for shoulder / elbow / wrist joints), or can be replaced with a cycloidal pinwheel reducer or an RV reducer depending on the application requirements. The reducer output end is equipped with a mounting flange for connecting a six-dimensional force / torque sensor (4) to realize direct measurement of joint output torque.

[0107] The controllable damper (2b) is one of the key innovations of this invention, consisting of a magnetorheological fluid cavity and an electromagnetic coil. The magnetorheological fluid cavity is located between the output shaft of the reducer and the joint housing, and has an annular thin-layer structure. By changing the current of the electromagnetic coil, the apparent viscosity of the magnetorheological fluid can be continuously adjusted within 0.5 milliseconds, thereby dynamically changing the equivalent damping coefficient of the transmission link. In position control mode, a large current is applied to the electromagnetic coil to solidify the magnetorheological fluid, and the transmission chain is in a high-stiffness state; in compliant interaction mode, the current is reduced or turned off to restore the magnetorheological fluid to a flow state, and the transmission chain is in a low-damping state, achieving force transmission transparency similar to a series elastic actuator, without sacrificing bandwidth and stiffness.

[0108] The magnetorheological fluid chamber has a gap of 0.5 to 1.0 mm, and the magnetorheological fluid is a carbon-based microparticle suspension type (such as Lord MRF-140CG), with a zero-field viscosity of 0.2 to 0.4 Pa·s and a yield stress adjustment range of 0 to 80 kPa. The electromagnetic coil has 200 to 400 turns, and the excitation current is 0 to 2 amperes (corresponding to a magnetic field strength of 0 to 250 kA / m). The equivalent damping coefficient is adjustable from 0.01 to 50 Nm / s / radian.

[0109] The core advantage of this controllable damper is that its stiffness and flexibility characteristics can be continuously adjusted online instead of being switched in a binary manner, with an adjustment time constant of less than 1 millisecond (far faster than the tens of milliseconds of mechanical switching mechanisms), and it does not introduce additional elastic energy storage elements (avoiding the inherent oscillation frequency limitation of the SEA), enabling the system to maintain high performance throughout the entire force bandwidth.

[0110] Multimodal sensing fusion module:

[0111] like Figure 4 The diagram shows the configuration of the multimodal sensing fusion module, illustrating the installation positions of the dual encoder (3a), the six-dimensional force / torque sensor (4), the NTC thermistor, and the IMU (5).

[0112] The sensing module (3) integrates the following sensors to form a joint-level full-state sensing system:

[0113] (3a) Dual encoder differential position measurement: A 19-bit (524,288 lines) magnetic encoder is installed on the motor side, and a 17-bit (131,072 lines) absolute magnetic encoder is installed on the reducer output side. The differential signal from the dual encoders can calculate the actual transmission ratio deviation and backlash angle of the reducer in real time, compensate for transmission errors, and also detect the wear degree of the reducer tooth surface.

[0114] (3b) Six-dimensional force / torque sensor (4): It adopts a strain gauge structure and is installed between the reducer output flange and the joint housing. It can measure the three-dimensional force and three-dimensional torque at the joint output end. The torque measurement resolution is not less than 0.01 Nm and the bandwidth is not less than 1 kHz. This sensor provides direct physical quantity feedback for torque closed-loop control and collision detection.

[0115] (3c) Online estimation of winding temperature: An NTC thermistor (sampling period 10 milliseconds) is embedded at the end of the stator winding, and an online estimation algorithm for winding temperature based on a thermal equivalent circuit model is run in the embedded controller, using the square of the current as the integral ( Using ambient temperature as input, the system estimates the hottest temperature of the winding in real time with an accuracy better than ±3 degrees Celsius.

[0116] (3d) Current Harmonic Analysis Module: The driver ADC acquires three-phase current at a sampling rate of 200kHz, and the embedded controller extracts the amplitudes of the 5th, 7th, 11th, and 13th current harmonics in real time through Fast Fourier Transform (FFT). Abrupt changes in harmonic characteristics can serve as early warning signals for demagnetization faults or inter-turn short circuits in the windings.

[0117] (3e) Inertial Measurement Unit (IMU) (5): Integrates a triaxial accelerometer and a triaxial gyroscope, mounted on the joint housing, to detect the joint's vibration acceleration and angular velocity. Combined with dual encoder signals, it enables high-precision online identification of joint kinematic parameters.

[0118] GaN-based wide bandgap power drive module:

[0119] like Figure 5 The diagram shown is a GaN-based power drive module circuit topology, illustrating a three-phase full-bridge inverter, a synchronous sampling ADC, and hardware protection circuitry.

[0120] The power drive module (6) adopts a three-phase full-bridge inverter topology based on gallium nitride (GaN) field-effect transistors. Compared with traditional silicon MOSFETs, GaN devices have lower on-resistance, faster switching speed and smaller package size, which reduces the area of ​​the drive board by about 50%, making it suitable for embedding in the compact space inside the joint module.

[0121] Key parameters of the drive module: Bus voltage 48V (compatible with general low-voltage power supply standards for robots); continuous phase current output capability of not less than 30 amps (peak 60 amps); switching frequency configurable from 40kHz to 100kHz (high-frequency operation reduces current ripple and audible noise); current sampling uses a synchronous sampling ADC with a sampling rate of not less than 200kHz to ensure current loop control accuracy. The drive board integrates hardware overcurrent protection (response time less than 1 microsecond), overvoltage / undervoltage protection, and short-circuit protection circuits.

[0122] Joint-level embedded intelligent control module:

[0123] The embedded control module (7) is equipped with a dual-core microcontroller (such as an ARM Cortex-M7+M4 or RISC-V dual-core). The main core runs a current loop and a position / speed / torque servo loop, while the slave core runs a communication protocol stack and advanced intelligent algorithms. The control module implements the following hierarchical control architecture:

[0124] like Figure 6 The diagram shown is a hierarchical control architecture block diagram, illustrating the four-loop nested structure of current loop, torque loop, speed loop, and position loop, as well as the impedance control mode switching logic.

[0125] (1) Inner current loop (bandwidth not less than 5kHz): Field-oriented control (FOC) based on space vector modulation (SVPWM) to achieve Axis current decoupling. The current loop employs a PI controller with back EMF feedforward compensation and dead-zone compensation, or a finite-set current control strategy based on model predictive control (MPC) can be selected, which can select the optimal voltage vector in each switching cycle and reduce current harmonic distortion.

[0126] (2) Torque loop (bandwidth not less than 1kHz): High-precision closed-loop control of joint output torque is achieved using the six-dimensional force / torque sensor signal as feedback. The torque loop controller adopts a PD controller with friction torque feedforward compensation and reducer hysteresis compensation. The friction model adopts the LuGre model, and the parameters are identified offline and stored in the controller Flash during joint factory calibration. The first step is to identify Coulomb friction using a stepped velocity curve. Static friction and Stribeck speed The second step involves using micro-amplitude vibration excitation to identify the stiffness coefficient from the hysteresis curve. (1e4 to 1e6 N·m / rad); Step 3, residual minimization identification Accuracy verification: Root mean square error is less than 0.05 N·m.

[0127] (3) Velocity / position outer loop (bandwidth not less than 200Hz): The position loop uses an S-curve trajectory generator plus a PD controller; the velocity loop uses a PI controller with inertia feedforward. The position and velocity references are sent by the host whole-body motion controller via EtherCAT bus at a 1kHz cycle.

[0128] (4) Impedance control mode: A spring-damping virtual model is superimposed on the torque loop, and the joint output torque is... ,in For virtual stiffness, For virtual damping, This is the feedforward torque. and It can be controlled in real time by the upper controller or adjusted online by the joint-level adaptive algorithm to achieve a seamless transition from rigid servo to compliant interaction.

[0129] (5) Collision detection and safety response: The collision detection algorithm based on the generalized momentum observer runs locally at the joint level. The algorithm estimates the expected torque using the motor current, joint position and reducer model, compares it with the measured torque of the force sensor, and determines that a collision has occurred when the residual exceeds the adaptive threshold. It then switches to zero torque mode or impedance buffer mode within less than 2 milliseconds without waiting for instructions from the upper controller, thus ensuring human-machine safety.

[0130] like Figure 7 The diagram shown is a flowchart of collision detection and safety response, illustrating the generalized momentum observer algorithm and multi-level safety strategies.

[0131] (6) Thermal Management and Life Protection: The embedded controller runs the winding thermal model in real time and dynamically adjusts the current limiting curve based on the estimated winding temperature. When the temperature approaches the warning value, the system automatically reduces the peak current allowable value (instead of directly cutting off the power), achieving "flexible derating" rather than "hard protection cut-off", ensuring that the robot can still maintain degraded motion capability under overheating trends. At the same time, the historical temperature curve is recorded and reported through EtherCAT for whole-body thermal management scheduling and predictive maintenance.

[0132] like Figure 8 The diagram shown is a schematic of the winding thermal model and the flexible derating protection strategy.

[0133] Communication and multi-joint coordination:

[0134] like Figure 9 The diagram shown is a multi-joint coordinated communication architecture diagram, illustrating the topology connection between the EtherCAT master station and multiple slave stations, as well as the CAN-FD redundant channel.

[0135] Each joint module is connected to the full-body motion controller (master station) via an EtherCAT industrial Ethernet bus, with a communication cycle of 1 millisecond. It supports distributed clock synchronization (DC mode), and the time synchronization accuracy between joints is better than 1 microsecond. Each joint module acts as an EtherCAT slave station, exchanging position commands, torque commands, sensor feedback, and status information in real time via Process Data Objects (PDOs). The full-body motion controller can simultaneously coordinate and control 20 to 40 joint modules, meeting the full-body degrees of freedom requirements of the humanoid robot.

[0136] The joint module also supports the CAN-FD bus as a backup communication channel. When the EtherCAT link is interrupted, the system automatically switches to the CAN-FD channel to maintain basic safety control (zero torque or position holding), thus ensuring communication redundancy.

[0137] In this invention, it is applied to the hip joint of a humanoid robot (high torque, high stiffness).

[0138] In the hip joint of a humanoid robot, the joint needs to bear the entire body weight and provide the large torque output required for walking. In this embodiment, the motor is a frameless torque motor with 18 slots, 20 poles, and an outer diameter of 90 mm, with a continuous output torque of 2.5 Nm and a peak torque of 8 Nm. The reducer is a harmonic reducer with a reduction ratio of 50:1 (transmission efficiency of approximately 80%), with a continuous joint output torque of approximately 100 Nm and a peak torque of approximately 320 Nm, which matches the walking requirements (80 to 120 Nm) and running peak requirements (200 to 300 Nm) of the humanoid robot's hip joint.

[0139] During walking, the whole-body motion controller sends position and feedforward torque commands to each joint at a frequency of 1kHz via EtherCAT. The outer loop of the hip joint module tracks the gait trajectory, while the middle loop compensates for gravity and dynamic loads. During the support phase (single-leg strike), the magnetorheological fluid controllable damper is supplied with maximum current, and the transmission chain is in a high-stiffness state to ensure accurate position tracking. During the swing phase (leg suspension), the damper current is reduced, and impedance control mode is used to achieve smooth leg swing and reduce energy consumption.

[0140] The online winding temperature estimation model based on thermal equivalent circuit estimates the stator winding temperature in real time. When continuous high torque output causes the temperature rise to approach the warning value (150 degrees Celsius), the controller automatically reduces the peak current limit to 70% of the rated value and notifies the whole body controller to adjust the gait parameters (such as reducing the step frequency or reducing the step size) to achieve system-level thermal coordination management.

[0141] In this invention, it is applied to the shoulder joint of a humanoid robot (fast response, compliant operation).

[0142] In the shoulder joint of a humanoid robot, the joint needs to enable rapid arm movements and precise force control operations (such as grasping objects and delivering tools). In this embodiment, the motor is a 12-slot, 14-pole, 60 mm outer diameter frameless torque motor, with a continuous output torque of 6 Nm and a peak torque of 20 Nm. The reducer is a precision planetary reducer with a reduction ratio of 30:1.

[0143] When the robot performs a grasping task, the position control mode is used during the rapid movement phase of the arm, and the magnetorheological fluid damper is in a high-stiffness state. When the end effector approaches the target object, the system seamlessly switches to impedance control mode, reducing the damper current and virtual stiffness. Set to 5 Nm per radian, virtual damping The force is set to 0.5 Nm / s per radian to achieve smooth contact. If the six-dimensional force / torque sensor detects an abnormally large contact force during the grasping process (such as hitting a person's arm), the collision detection algorithm will switch the joint to zero torque mode within 1.5 milliseconds to ensure safety.

[0144] In this invention, it is applied to the ankle joint of a humanoid robot (impact resistant, high dynamic).

[0145] In the ankle joint of a humanoid robot, the joint must withstand an impact load several times the body weight at the moment of landing, while requiring high dynamic response to maintain balance. In this embodiment, the motor is a frameless torque motor with 12 slots, 14 poles, and an outer diameter of 75 mm, coupled with a harmonic reducer with a reduction ratio of 80:1.

[0146] Upon impact, the magnetorheological fluid damper, based on the impact acceleration signal detected by the IMU, increases the damping coefficient to its maximum value within 0.5 milliseconds to absorb impact energy and protect the gear teeth. Normal damping settings are immediately restored after the impact. The differential signal from the dual encoders monitors the deformation of the flexspline of the harmonic reducer in real time. When the accumulated deformation exceeds a preset threshold, the system issues a reducer life warning.

[0147] The high switching frequency (80kHz) of the GaN driver ensures that the ankle motor has a current tracking accuracy of better than 2% during rapid torque switching (such as the transition from push-off to leg swing, where the torque direction reverses within 10 milliseconds), meeting the requirements of gait dynamic balance.

[0148] Compared with existing humanoid robot joint control technology, this invention has the following significant advantages and effects:

[0149] (1) Ultra-high integration: The "five-in-one" architecture integrates the motor, reducer, sensor, driver and controller into a single module. Compared with the traditional split solution, the module volume is reduced by about 40%, the number of wiring harnesses is reduced by about 70% (each joint only needs one power line and one EtherCAT communication line), and the assembly time of the whole machine is reduced by about 50%.

[0150] (2) Wideband torque control and compliant interaction: The torque control bandwidth is not less than 1kHz. Combined with the magnetorheological fluid controllable damper, the rigidity and flexibility characteristics can be continuously adjusted online (time constant is less than 1 millisecond). It combines the force transparency of the series elastic actuator and the high bandwidth of the direct drive actuator, and achieves seamless switching between precise positioning and safe compliant interaction on the same joint.

[0151] (3) Joint-level collision safety: The local collision detection response time based on the generalized momentum observer is less than 2 milliseconds, without relying on the upper controller, which meets the collision response time requirements in the ISO / TS 15066 collaborative robot safety standard.

[0152] (4) Full-state perception and diagnosis: Dual encoder differential measurement realizes real-time compensation of transmission error (improving accuracy by about one order of magnitude) and reducer wear monitoring; current harmonic analysis realizes early warning of demagnetization and inter-turn short circuit; winding thermal model realizes flexible derating protection to avoid hard protection cut-off interruption of movement.

[0153] (5) GaN drive is efficient and compact: GaN power devices reduce the area of ​​the drive board by about 50%, reduce switching losses by about 30%, and push the audible noise frequency to the ultrasonic range (above 40kHz), improving the comfort of human-machine coexistence environment.

[0154] (6) Communication redundancy and edge intelligence: The EtherCAT plus CAN-FD dual-bus redundancy architecture ensures communication reliability; the joint-level embedded intelligent controller can independently execute collision safety response and thermal management decisions without relying on the central controller, thus improving system robustness.

[0155] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A multifunctional humanoid joint motor drive control system, characterized in that, It includes: The frameless torque motor module adopts an external rotor and internal stator topology. The stator has a fractional slot concentrated winding structure to generate joint driving torque. The transmission module, located between the output end of the frameless torque motor module and the joint load, includes a precision reducer and a controllable damper. The controllable damper is a magnetorheological fluid damper, used to dynamically adjust the equivalent damping coefficient of the transmission link. The multimodal sensing fusion module includes at least: an encoder system for position feedback; a force / torque sensor for directly measuring joint output torque; and a temperature sensor for temperature monitoring. A power drive module is used to drive the frameless torque motor module; The embedded intelligent control module is used to perform hierarchical control of the inner current loop, middle torque loop, outer speed loop, and outer position loop, and to run collision detection and thermal management strategies.

2. The multifunctional humanoid joint motor drive control system according to claim 1, characterized in that, The controllable damper of the transmission module consists of a magnetorheological fluid cavity and an electromagnetic coil. The magnetorheological fluid cavity is located between the output shaft of the reducer and the joint housing, and has an annular thin-layer structure with a cavity gap of 0.5 to 1.0 mm. By adjusting the current of the electromagnetic coil, the apparent viscosity of the magnetorheological fluid can be continuously adjusted within 0.5 milliseconds, thereby dynamically changing the equivalent damping coefficient of the transmission link. The adjustable range of the equivalent damping coefficient is 0.01 to 50 Nm / s per radian.

3. The multifunctional humanoid joint motor drive control system according to claim 1, characterized in that, The encoder system of the multimodal sensing fusion module is a dual encoder system, including a 19-bit magnetic encoder on the motor side and a 17-bit absolute magnetic encoder on the reducer output side. It is used to calculate the actual transmission ratio deviation and backlash angle of the reducer in real time, compensate for transmission errors, and detect the wear degree of the reducer tooth surface. The multimodal sensing fusion module also includes a current harmonic analysis module. The embedded intelligent control module performs a fast Fourier transform on the three-phase current collected by the power drive module and extracts the amplitude values ​​of the 5th, 7th, 11th and 13th current harmonics in real time. When the deviation of the current harmonic amplitude exceeds a preset threshold, an early warning signal of demagnetization fault or winding inter-turn short circuit is issued.

4. The multifunctional humanoid joint motor drive control system according to claim 1, characterized in that, The multimodal sensing fusion module also includes an inertial measurement unit, which is mounted on the joint housing and integrates a three-axis accelerometer and a three-axis gyroscope for detecting the joint's vibration acceleration and angular velocity.

5. The multifunctional humanoid joint motor drive control system according to claim 1, characterized in that, The embedded intelligent control module is equipped with a dual-core microcontroller, in which the main core runs a current loop and position, speed and torque servo loops, and the slave core runs a communication protocol stack and intelligent algorithms. The current inner loop bandwidth is not less than 5kHz, and it is based on magnetic field orientation control and space vector modulation, using a PI controller with back EMF feedforward compensation and dead zone compensation. The torque loop bandwidth is not less than 1kHz, and the signal of the force / torque sensor is used as feedback. A PD controller with friction torque feedforward compensation and reducer hysteresis compensation is adopted. The bandwidth of the outer velocity loop and the outer position loop is not less than 200Hz. The position loop adopts an S-curve trajectory generator plus a PD controller; the velocity loop adopts a PI controller with inertia feedforward. The embedded intelligent control module runs a collision detection algorithm based on a generalized momentum observer. It estimates the expected torque using motor current, joint position and reducer model, and compares it with the measured torque of the force / torque sensor. When the residual exceeds the adaptive threshold, it determines that a collision has occurred and switches to zero torque mode or impedance buffer mode in less than 2 milliseconds. In collision detection, residuals are calculated. When the residual exceeds the adaptive threshold When a collision is detected, it is determined that a collision has occurred; among them, The torque measured by the force / torque sensor. The expected torque is estimated based on motor current, joint position, and reducer model. The mean of the residuals, The standard deviation of the residuals. These are preset coefficients.

6. The multifunctional humanoid joint motor drive control system according to claim 1, characterized in that, The thermal management strategy includes: the embedded intelligent control module runs an online winding temperature estimation algorithm based on a thermal equivalent circuit model, uses the current square integral and ambient temperature as inputs to estimate the temperature of the hottest spot in the winding in real time, and dynamically adjusts the current limiting curve according to the estimated temperature to achieve flexible derating protection. The online winding temperature estimation algorithm adopts a two-node RC thermal equivalent circuit model, and its current limiting satisfies ,in, For limiting current, Rated current, It is a continuously decreasing function. To estimate the winding temperature, This is the rated temperature.

7. The multifunctional humanoid joint motor drive control system according to claim 1, characterized in that, The system supports seamless switching between three operating modes: high stiffness position mode, high bandwidth torque mode, and impedance control mode. Wherein, the joint output torque under the impedance control mode satisfies: ； in, This represents the total torque actually output by the joint. For virtual stiffness, For virtual damping, For feedforward torque, For the target joint position, This represents the current actual position of the joint.

8. The multifunctional humanoid joint motor drive control system according to claim 1, characterized in that, The power drive module is a GaN-based wide bandgap power drive module, adopting a three-phase full-bridge inverter topology. Its bus voltage is 48V, the continuous output capability of phase current is not less than 30 amps, the peak current is not less than 60 amps, the switching frequency is 40kHz to 100kHz, and it integrates hardware overcurrent protection circuit, overvoltage protection circuit, undervoltage protection circuit and short circuit protection circuit with a response time of less than 1 microsecond.

9. A multifunctional humanoid joint motor drive control system according to claim 1, characterized in that, The system is connected to the whole-body motion controller via the EtherCAT industrial Ethernet bus, with a communication cycle of 1 millisecond and supports distributed clock synchronization. It also supports the CAN-FD bus as a backup communication channel, automatically switching to the CAN-FD channel to maintain basic safety control when the EtherCAT link is interrupted.

10. A control method for a multifunctional humanoid joint motor drive control system according to any one of claims 1-9, characterized in that, It includes: Step 1: Determine the working mode and identify the target working mode, which includes high stiffness position mode, high bandwidth torque mode and impedance control mode. Step 2: Damping coefficient setting. The equivalent damping coefficient of the controllable damper is set according to the target operating mode: the maximum value is taken in the high-stiffness position mode. In impedance control mode, according to virtual damping Configure the mapping, and take the minimum value in high bandwidth torque mode. ; Step 3: Damping switching. The damping coefficient is switched within 0.5 milliseconds by adjusting the electromagnetic coil current of the magnetorheological fluid damper. Step 4: Four-loop control. The embedded controller of the joint module synchronously runs the inner current loop, the middle torque loop, the outer speed loop, and the outer position loop to perform nested control of position-speed-torque-current four-loop closed loops. Step 5: Collision safety. The host controller sends position, velocity and feedforward torque commands through the EtherCAT bus at a 1kHz cycle; it collects multimodal sensor data in real time, runs a collision detection algorithm based on a generalized momentum observer, calculates the residual, and switches to zero torque mode or impedance buffer mode within 2 milliseconds when the residual exceeds the adaptive threshold. Step 6: Thermal protection. Real-time estimation of winding temperature. When the winding temperature exceeds 0.8 times the rated temperature value, flexible derating protection is implemented, and the current limiting curve is dynamically adjusted.

Images