Modularized humanoid robot head intelligent control management system

Abstract

The invention relates to the technical field of robot technology, and discloses a modular humanoid robot head intelligent control management system, which comprises a sensing module, a fluid control module, an operation scheduling module and a mechanical support module, and is characterized in that the sensing module integrates visual and inertial components to acquire data; the fluid control module is located in the mechanical supporting module to establish a controlled fluid field; the operation scheduling module calculates a comprehensive motion intensity index based on the inertial data, determines a motion masking window, and drives the fluid control module to perform heat dissipation during the window period or during the non-exposure period of the visual component, and the mechanical support module defines a fluid channel and provides a physical interface. According to the invention, a rigid-flexible combined board architecture and a vector diversion technology are adopted, and mechanical vibration is covered by physical motion, so that the cooperation of dynamic heat dissipation and static imaging is realized, and the conflict between high-power-consumption heat dissipation and precise imaging and image stabilization in a compact space is solved.

Description

Technical Field

[0001] This invention relates to the field of robotics, and in particular to a modular humanoid robot head intelligent control and management system. Background Technology

[0002] With the evolution of humanoid robot technology, the head unit, as the core carrier for environmental perception and interaction, integrates high-resolution visual sensors, depth detection modules, and high-performance edge computing chips. This high-density hardware integration leads to a significant increase in system heat dissipation. Furthermore, the compact arrangement of electronic components within the limited physical space of the head makes it difficult for heat to dissipate quickly. Relying solely on passive heat conduction mechanisms such as heat sinks or heat pipes is increasingly insufficient to meet the thermal management requirements of the core computing unit operating under high loads for extended periods. Therefore, it is necessary to introduce active fluid cooling mechanisms to maintain the system's thermal stability.

[0003] However, when introducing active cooling components such as centrifugal fans into a compact mechanical structure, the mechanical vibrations and airflow pulsations generated by their high-speed rotation can easily couple to the precision sensing components in the same compartment through physical connections or fluid media. Since cameras, lidar, and MEMS inertial measurement units are highly sensitive to minute vibrations, the high-frequency vibrations generated by the cooling actuators often lead to motion blur, rolling shutter effects, or the introduction of high-noise substrates into the inertial measurement data. Existing thermal management solutions are typically independent of the robot's motion control and perception logic, relying solely on temperature sensor feedback to start / stop the fan or adjust its speed. This ignores the dynamic coupling between cooling actions and sensor exposure timing or the robot's motion state, making it difficult to ensure both efficient cooling and high-quality signal acquisition.

[0004] Furthermore, existing internal airflow designs for humanoid robot heads typically employ a uniform fluid distribution strategy, lacking differentiated flow path construction tailored to the thermal and force-sensitive characteristics of different components. For example, current technologies struggle to simultaneously provide low-disturbance laminar cooling for vibration-sensitive optical components and high-heat-generating turbulent cooling for robust computing chips driven by a single air source. This traditional architecture, lacking fluid topology optimization and mechanical decoupling design, fails to effectively isolate vibration interference when facing complex conditions of high computational load and concurrent high-dynamic motion, limiting the overall performance of the humanoid robot head system under extreme conditions. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a modular intelligent control and management system for the head of a humanoid robot, which solves the problem of mechanical vibration interference caused by high-power active heat dissipation within the compact space of the humanoid robot's head, affecting the imaging quality and measurement accuracy of precision sensing components.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a modular humanoid robot head intelligent control and management system, comprising a sensing module, a fluid control module, a computation and scheduling module, and a mechanical support module;

[0007] The perception module integrates a visual acquisition component and an inertial measurement component, and is configured to acquire environmental information and motion state information and output image data sequences and kinematic data.

[0008] The fluid control module is located in the internal cavity of the mechanical support module and is configured to generate heat dissipation airflow and establish a controlled heat dissipation fluid field by changing the airflow direction.

[0009] The computation scheduling module is connected to the sensing module and the fluid control module respectively, and is configured to calculate a comprehensive motion intensity index based on the kinematic data, determine a motion masking window based on the comprehensive motion intensity index, and send a drive command to the fluid control module during the motion masking window or during the non-exposure period of the visual acquisition component.

[0010] The mechanical support module is configured to provide a physical installation interface and internally defines a fluid channel that connects to the fluid control module;

[0011] The fluid control module responds to the drive command by performing a heat dissipation action within the fluid channel, and performs thermal management on the sensing module or the computation scheduling module.

[0012] Preferably, the sensing module adopts a rigid-flexible composite board architecture;

[0013] The inertial measurement unit is rigidly connected to the geometric center of the mechanical support module via a surface mount process, and is used to characterize the motion state of the fuselage coordinate system.

[0014] The visual acquisition component establishes a mechanical decoupling zone between the inertial measurement component and the optical imaging domain through a flexible plate. It decouples only from the high-frequency vibrations generated by the fan, while exhibiting rigidity in the low-frequency range of human movement to block high-frequency micro-vibrations from the fluid control module.

[0015] Preferably, the computation scheduling module includes a real-time coprocessing logic unit and an application main processing logic unit;

[0016] The real-time coprocessing logic unit is configured to read the angular velocity vector and acceleration vector output by the inertial measurement component via direct memory access, and calculate the comprehensive motion intensity index based on the weighted Euclidean norm equation.

[0017] The real-time coprocessing logic unit is also configured to maintain thermal credit scores and, when the motion masking window is detected and the thermal credit score is positive, output a high duty cycle pulse width modulation signal to drive the fluid control module.

[0018] Preferably, the calculation of the comprehensive motion intensity index involves angular velocity weighting coefficient and acceleration weighting coefficient;

[0019] The real-time coprocessing logic unit is configured to compare the comprehensive motion intensity index with the masking threshold to generate a masking window signal;

[0020] The masking threshold is configured to be inversely proportional to the current exposure time parameter of the visual acquisition component.

[0021] Preferably, the fluid control module includes a centrifugal booster assembly and a vector flow guide assembly;

[0022] The centrifugal booster assembly is suspended and installed inside the mechanical support module via an elastic damping element, and is configured to generate a high-pressure pulsed airflow.

[0023] The vector flow guiding component is located on the air outlet path of the centrifugal booster component, and includes deflectable guide vanes. It is configured to receive angle position commands and drive the deflectable guide vanes to deflect, thereby directionally guiding the airflow to the target heat dissipation area.

[0024] Preferably, the vector flow guide assembly is configured to dynamically distribute the volumetric flow rate through different heat dissipation branches by adjusting the deflection angle of the deflectable flow guide blades;

[0025] The operation scheduling module has a pre-stored mapping table of local resistance coefficient and deflection angle, and calculates the angle position command based on the difference between the target temperature and the current temperature to adjust the flow resistance ratio of each heat dissipation branch.

[0026] Preferably, the mechanical support module has a main air intake channel, a diversion chamber, a first heat dissipation channel pointing to the sensing module, and a second heat dissipation channel pointing to the computing and scheduling module.

[0027] The first heat dissipation channel is configured as a flat laminar flow channel for surface-grabbing cooling of the vision acquisition component.

[0028] The second heat dissipation channel is configured as a turbulent channel with turbulence columns on the inner wall, which is used to perform impact cooling on the computing and scheduling module.

[0029] Preferably, the sensing module further includes a depth detection component and a temperature monitoring component;

[0030] The depth detection component is installed next to the vision acquisition component, and the parallelism deviation of their optical axes is within a preset tolerance range.

[0031] The temperature monitoring component consists of a distributed array of negative temperature coefficient thermistors, which are respectively attached to the back plate of the vision acquisition component sensor, the surface of the depth detection component driver, and the surface of the computing scheduling module chip.

[0032] Preferably, the visual acquisition component is configured as a complementary metal-oxide-semiconductor sensor component based on a global shutter, and has independent hardware trigger input pins and exposure synchronization output pins;

[0033] The computation scheduling module is configured to send a synchronous trigger signal to the fluid control module, and before sending the exposure trigger signal of the vision acquisition component, calculate the rotational speed decay time of the fluid control module or output a reverse braking voltage, so that the fluid control module is in a silent state at the moment the exposure window opens.

[0034] Preferably, the mechanical support module includes an outer frame made of a high specific strength material and an open-cell sound-absorbing foam coated on the inner wall;

[0035] The mechanical support module also includes a suspended mounting base for supporting the sensing module. The suspended mounting base and the outer frame are connected by a viscoelastic damper to form a mass damping system, and the natural frequency of the suspended mounting base is configured to be lower than the minimum operating fundamental frequency of the fluid control module.

[0036] This invention provides a modular intelligent control and management system for the head of a humanoid robot. It offers the following advantages:

[0037] 1. This invention constructs a comprehensive motion intensity index through a computational scheduling module and determines the motion masking window accordingly. Only during periods of intense system motion or when the visual acquisition component is in a non-exposed state, the fluid control module is driven to perform high-power heat dissipation. This control strategy utilizes the external dynamic acceleration generated by physical motion to mask the mechanical vibrations of the fluid control module during operation. Combined with the global shutter characteristics of the visual acquisition component and the millisecond-level start-stop response of the centrifugal pressurization component, it achieves heat dissipation during motion and imaging during stillness within the limited physical space of the head. This effectively resolves the timing conflict between the heat dissipation requirements of the high-computing-power unit and the image stabilization requirements of the precision optical components, ensuring that the image acquisition quality is not affected by the vibrations from active heat dissipation.

[0038] 2. This invention achieves on-demand distribution of the internal heat flow field by combining the vector flow guiding component in the fluid control module with the differentiated flow channel topology design within the mechanical support module. By dynamically changing the flow resistance ratio through adjusting the angle of the deflectable guide vanes, the first heat dissipation channel of the pointing vision acquisition component maintains a low Reynolds number laminar flow state for surface-grabbing cooling, preventing sensor displacement caused by airflow pulsation. Simultaneously, the second heat dissipation channel of the pointing computation scheduling module forms a high Reynolds number turbulent flow for impact cooling, improving the convective heat transfer coefficient. This design, driven by a single air source, balances the low-vibration environment of the sensitive components with the efficient heat dissipation requirements of the high-power computing unit, improving the overall energy efficiency ratio of the system.

[0039] 3. This invention optimizes the measurement accuracy of multimodal sensors through a rigid-flexible plate structure and a multi-level mechanical vibration isolation physical layout. By rigidly fixing the inertial measurement unit to the geometric center of the mechanical support module, the centripetal and tangential acceleration interference caused by non-ideal installation positions is eliminated at the physical level, ensuring the authenticity of motion data. Simultaneously, by utilizing the natural bending characteristics of the flexible plate and the damping system of the suspended mounting base, a mechanical decoupling zone is established between the visual acquisition component and the fluid control module. This effectively blocks the transmission of high-frequency mechanical vibrations generated by the fluid control module to the optical imaging domain, reducing the noise level during multi-sensor fusion. Attached Figure Description

[0040] Figure 1 This is a diagram of the module architecture of the present invention;

[0041] Figure 2 This is a schematic diagram of the internal electrical connections and physical layout of the sensing module of the present invention;

[0042] Figure 3 This is an exploded view of the structure and a schematic diagram of the fluid dynamics principle of the fluid control module of the present invention;

[0043] Figure 4 This is a schematic diagram of the logical architecture and data flow of the computation scheduling module of the present invention;

[0044] Figure 5 This is a three-dimensional cross-sectional view of the mechanical support module of the present invention and a schematic diagram of the internal air duct topology.

[0045] Among them, 100 is the sensing module; 200 is the fluid control module; 300 is the calculation and scheduling module; and 400 is the mechanical support module. Detailed Implementation

[0046] The technical solutions in 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.

[0047] See attached document Figure 1 , Figure 1 This is a module architecture diagram according to an embodiment of the present invention. The present invention provides a head intelligent control and management system based on inertial fluid vector control and vibration masking. The system adopts a tightly coupled electromechanical-thermal integrated architecture and may include: a sensing module 100, a fluid control module 200, a computation and scheduling module 300, and a mechanical support module 400.

[0048] The sensing module 100 is configured to construct the system's multi-dimensional state space, integrating visual acquisition, depth detection, inertial measurement, and temperature monitoring functions. It outputs multimodal data sequences to the computation and scheduling module 300 via an electrical interface. The sensing module 100 is responsible for data acquisition and preliminary signal conversion. Internally, it integrates a high-resolution CMOS sensor assembly, a lidar or time-of-flight camera assembly, a six-axis or nine-axis inertial measurement assembly, and distributed thermal detection components. To balance motion capture accuracy and imaging stability, the inertial measurement assembly is rigidly connected to the physical anchor points of the mechanical support module 400 to ensure that the measured angular velocity and acceleration vectors accurately represent the overall motion state of the system, i.e., the true motion of the fuselage coordinate system. The vibration-sensitive visual assembly is installed using a flexible decoupling method. The sensing module 100 connects to the computation and scheduling module 300 via a high-speed data bus to transmit image and point cloud data, while simultaneously transmitting low-latency kinematic data through a high-priority real-time interrupt interface.

[0049] The fluid control module 200 responds to drive commands from the computation and scheduling module 300 to establish a controlled heat dissipation fluid field within the system. Located within the internal cavity of the mechanical support module 400, the fluid control module 200 includes a high static pressure centrifugal booster assembly with rapid acceleration characteristics and an adjustable-angle vector guide assembly. The centrifugal booster assembly is suspended by elastic damping elements and configured to rapidly increase its rotational speed upon receiving a high duty cycle signal to generate a high-pressure pulsed airflow. The vector guide assembly is located in the outlet path of the centrifugal booster assembly and receives angular position commands to drive the guide vanes to deflect, directing the airflow to the target heat dissipation area specified by the computation and scheduling module 300.

[0050] The computation scheduling module 300, serving as the system's logic control center, executes hierarchical control logic and is physically integrated into the back or bottom of the mechanical support module 400. The computation scheduling module 300 internally includes real-time co-processing logic and application main processing logic. The real-time co-processing logic is responsible for reading inertial and temperature data via direct memory access, determining whether the current state is suitable for activating a powerful cooling motion masking window, and maintaining the system's thermal credit score. Based on this calculation result, the computation scheduling module 300 outputs motor speed control signals and flow guidance control signals to the fluid control module 200, and outputs a synchronization trigger signal to the sensing module 100, ensuring that high-pressure cooling actions are only performed during periods when the visual components are not exposed or are under motion blur masking.

[0051] The mechanical support module 400 provides a physical mounting interface for the aforementioned modules and defines the topology of the internal fluid channels. The outer surface of the mechanical support module 400 has an asymmetrically distributed array of air inlets and outlets. The air inlet array is connected to the suction end of the fluid control module 200 via an internal sealed flow channel. The mechanical support module 400 internally constructs a directional airflow network, which constrains airflow within a preset path and structurally ensures that the vibration energy of the fluid control module 200 is absorbed by elastic damping elements, reducing the impact on the precision optical components in the sensing module 100.

[0052] See attached document Figure 2 , Figure 2 This is a schematic diagram of the internal electrical connections and physical layout of a sensing module according to an embodiment of the present invention. The sensing module 100, as the multi-dimensional information capture front end of the system, preferably employs a highly integrated rigid-flexible board design in this embodiment. This design not only adapts to the compact, irregularly shaped space constraints of the head unit, but more importantly, utilizes the natural bending characteristics of the flexible board to establish a mechanical decoupling zone between the inertial measurement domain and the optical imaging domain. This prevents the high-frequency micro-vibrations generated by the fluid control module 200 from being directly coupled to the sensor chip through a rigid connector, thereby ensuring the physical stability of the signal link.

[0053] The sensing module 100 includes a visual imaging unit 101, which is configured as a complementary metal-oxide-semiconductor sensor assembly based on a global shutter. Considering the high-speed airflow disturbance generated by the centrifugal pressurization unit 201 within the system, unlike a rolling shutter sensor, the global shutter sensor assembly can expose all pixels at once, fundamentally eliminating the rolling shutter effect and intra-frame distortion that may occur when shooting in high-speed motion or high-frequency vibration environments caused by internal fans. The visual imaging unit 101 establishes a high-speed data channel with the main processing unit 302 in the computation scheduling module 300 via a low-voltage differential signal interface or a mobile industrial processor interface for transmitting raw image data. Simultaneously, the visual imaging unit 101 has independent hardware trigger input pins and exposure synchronization output pins, with the hardware trigger input pins physically connected to the general-purpose input / output interface of the co-processing unit 301 in the computation scheduling module 300. In this embodiment, the analog gain amplifier circuit and the analog-to-digital converter circuit of the visual imaging unit 101 are integrated on the sensor chip. The bias voltage of the analog gain amplifier circuit is controlled by the analog voltage signal output by the coprocessing unit 301, which is used to quickly adjust the dynamic range of the sensor when a sudden change in ambient light is detected.

[0054] The perception module 100 also includes a depth detection unit 102, which is configured as a solid-state lidar module or a time-of-flight camera module. The depth detection unit 102 is mounted beside the visual imaging unit 101, and the parallelism deviation of their optical axes is calibrated within a preset tolerance range using a precision optical fixture to ensure pixel-level alignment of the depth map data and visible light image data in the spatial coordinate system, avoiding ghosting or artifacts in subsequent fusion algorithms. The depth detection unit 102 internally includes an infrared emitter array and a single-photon avalanche diode receiver array. The driving circuit for the infrared emitter array and the readout circuit for the single-photon avalanche diode receiver array can be implemented using conventional driving and readout chips by those skilled in the art, and will not be described in detail here. The depth detection unit 102 is connected to the main processing unit 302 via a serial peripheral interface or a universal serial bus interface, outputting a point cloud data stream containing distance information.

[0055] The core component of the sensing module 100 is the inertial measurement unit 103, which is configured as a six-axis or nine-axis microelectromechanical system (MEMS) sensor, integrating a three-axis gyroscope and a three-axis accelerometer. The inertial measurement unit 103 is surface-mounted to the geometric center of the system motherboard. From a rigid body dynamics perspective, placing the inertial measurement unit 103 at the geometric center is to decouple the interference of rotational motion on linear acceleration measurement at a physical level. When the system undergoes angular motion, if the sensor deviates from the center of rotation, the accelerometer will inevitably sense additional tangential and centripetal accelerations, which will act as noise superimposed on the actual linear motion data.

[0056] To accurately characterize the system's motion state, the raw data output by the inertial measurement unit 103 needs to undergo coordinate transformation and error compensation. Considering that the fluid control module 200 will generate instantaneous high G-value vibrations when operating at full speed, in this embodiment, the internal registers of the inertial measurement unit 103 are configured in maximum range mode, for example, the accelerometer range is set to ±16g or higher, and the gyroscope range is set to ±2000dps. This is to prevent the instantaneous impact caused by the high burst airflow generated by the fluid control module 200 from causing sensor reading clipping saturation, and to ensure the continuity and linearity of motion data under severe vibration conditions. The data output pin of the inertial measurement unit 103 is connected to the direct memory access (DMA) controller channel of the coprocessor unit 301, enabling the coprocessor unit 301 to read angular velocity and acceleration data at kilohertz-level frequencies, such as 1kHz to 8kHz sampling rates, without occupying CPU core cycles.

[0057] To eliminate the impact of minor installation deviations caused by manufacturing tolerances on measurement accuracy, the system performs error compensation based on a rigid body kinematics model. This compensation algorithm runs in the hardware logic of the coprocessor unit 301 and is used to map measurement values ​​from non-ideal installation positions back to the ideal geometric center. Assuming the origin of the system coordinate system is located at the geometric center, if the inertial measurement unit 103 has an installation deviation, the relationship between its measured acceleration vector and the true acceleration vector follows the following dynamic equation:

[0058] ;

[0059] In the formula: This represents the measured acceleration vector output by the inertial measurement unit 103, which includes three-axis components; The vector representing the true translational acceleration at the geometric center, i.e., the actual motion state that the system is trying to solve; This represents the angular velocity vector measured by the gyroscope; This represents the angular acceleration vector, which, in a discrete-time system, is expressed through the angular velocity vector. Obtained by performing first-order difference or Kalman filter differentiation operations; The vector representing the position offset of the inertial measurement unit 103 relative to the geometric center is obtained through the static calibration process before leaving the factory and stored in non-volatile memory. This represents the gravitational acceleration vector, measured during the static initialization phase; This represents the zero bias error vector of the accelerometer, which drifts with temperature changes; Represents the measurement noise vector; This represents the cross product operation of vectors.

[0060] From the above formula, it can be seen that when the position offset vector When the magnitude approaches zero, the centripetal acceleration term With tangential acceleration term The influence of [the previous method] will be eliminated, and the measured value will directly approximate the true value. This invention uses physical layout design to make the position offset vector [the new value]. This minimizes the complexity and latency of subsequent motion calculations at the hardware level, providing a low-latency data foundation for achieving microsecond-level motion masking determination.

[0061] The sensing module 100 also integrates a temperature monitoring unit 104, which consists of a distributed array of negative temperature coefficient thermistors. The thermistors are attached to the sensor backplate of the vision imaging unit 101, the laser emission driver surface of the depth detection unit 102, and the main logic chip package surface of the computation scheduling module 300 using high thermal conductivity double-sided adhesive or thermal grease, respectively, to ensure minimal thermal resistance and accurate reflection of junction temperature changes. Each thermistor is connected to a high-precision analog-to-digital conversion interface, such as a 12-bit or higher precision ADC, of ​​the coprocessing unit 301 via a multiplexer. The temperature monitoring unit 104 not only serves as overheat protection but also as an input source for a thermal credit scoring algorithm; its real-time collected resistance values ​​are converted into digital temperature values ​​to construct a thermal distribution field model within the system.

[0062] In the circuit board design, the rigid area of ​​the sensing module 100 is used to mount the aforementioned sensor components and supporting passive components such as power management chips and filter capacitors, providing stable mechanical support and electrical connection. In particular, for high-speed signal lines transmitting image data and inertial data, impedance matching control, such as a 100-ohm differential impedance, is adopted in the PCB design to ensure signal integrity. The lens module of the visual imaging unit 101 is fixed to the lens mount on the rigid circuit board by UV-curable adhesive. A buffer rubber gasket is provided between the lens mount and the circuit board to form a suspended optical structure. This structure serves as a secondary vibration damping system, which, together with the primary vibration damping of the rigid-flexible composite board, further isolates the aerodynamic vibration from the fluid control module 200.

[0063] See attached document Figure 3 , Figure 3 This is an exploded view of the structure and a schematic diagram of the fluid dynamics principle of a fluid control module according to an embodiment of the present invention. The fluid control module 200, as the actuator of the system, is physically housed in a fluid chamber defined inside the mechanical support module 400. Its core function is to convert electrical energy into fluid kinetic energy and to perform vector reconstruction of the internal flow field according to the spatiotemporal distribution requirements of the heat load.

[0064] The fluid control module 200 includes a centrifugal booster unit 201, which is configured as a high static pressure blower with axial intake at the inlet and radial exhaust at the outlet. To achieve vibration masking coordination with the sensing module 100, i.e., to complete the burst and decay of the cooling airflow within an extremely short non-exposure time window, the rotor component of the centrifugal booster unit 201 is preferably made of a low moment of inertia material in this embodiment, such as carbon fiber reinforced polymer or high-strength polycarbonate. By reducing the moment of inertia, the rotor can respond to changes in the duty cycle of a pulse width modulation signal within milliseconds, for example, 50ms to 200ms, achieving a step response from maintaining a base positive pressure in a quiet, low-speed state to a high-burst operating speed state.

[0065] In terms of mechanical mounting architecture, the stator base of the centrifugal booster unit 201 is not directly and rigidly locked to the outer shell of the mechanical support module 400, but is suspended via an elastic damping connector. This elastic damping connector is made of a viscoelastic material with nonlinear stiffness characteristics, such as ethylene-vinyl acetate copolymer or silicone rubber with a Shore hardness of 30A-50A, creating a second-order mechanical low-pass filter along the physical connection path. This suspension structure is configured to have a natural frequency significantly lower than the lowest operating fundamental frequency of the centrifugal booster unit 201, for example, a designed natural frequency below 10Hz. This effectively isolates and attenuates high-frequency mechanical vibrations caused by motor commutation torque pulsations and impeller aerodynamic imbalances, preventing vibration energy from being transmitted to the sensing module 100 and causing micro-vibrations in the precision optical components, thus preventing image blurring.

[0066] The centrifugal booster unit 201 is equipped with a three-phase brushless DC motor, featuring a built-in high-precision Hall sensor or back EMF detection circuit for closed-loop speed control. The drive motor receives control voltage from the computation and scheduling module 300, establishing an anisotropic pressure field. Based on the principles of motor dynamics and fluid mechanics, the centrifugal booster unit 201 generates instantaneous airflow pressure... The relationship between the input control signal and the rotor dynamic characteristics can be described by the following differential equation:

[0067] ;

[0068] In the formula: This represents the instantaneous static pressure at the outlet of centrifugal booster unit 201, in Pascals (Pa). The mechanical time constant of the electromechanical system is determined by the rotor moment of inertia and the motor torque constant. This parameter characterizes the system's response speed to input commands. In this embodiment, this value is minimized by selecting high flux density magnets and lightweight impellers. This represents the pressure coefficient, which depends on the impeller's geometric parameters such as the blade exit angle and fluid density, and is obtained through experimental calibration or computational fluid dynamics simulation. This represents the duty cycle of the input pulse width modulation signal, and its value range is normalized. interval; This represents the pure time delay caused by the processing delay of the electronic speed controller and the inductance effect of the windings, which needs to be compensated for in the control algorithm.

[0069] This formula shows that by minimizing the mechanical time constant... The system can shorten the delay of airflow establishment and ensure that the high-pressure airflow falls accurately within the motion masking window predicted by the operation and scheduling module 300, realizing the timing control of heat dissipation in motion and imaging in stillness.

[0070] The fluid control module 200 also includes a vector flow guiding unit 202, which is located at the downstream bifurcation node of the outlet airflow channel of the centrifugal booster unit 201. The function of the vector flow guiding unit 202 is to distribute the high-speed airflow from a single source to different heat dissipation branches as needed, thereby achieving regionalized control of thermal management. In this embodiment, the vector flow guiding unit 202 includes a fluid distribution cavity, deflectable guide vanes, and a micro servo driver.

[0071] The fluid distribution cavity has one inlet and at least two outlets. The inlet is connected to the centrifugal booster unit 201, and the outlets are connected to the sensing and cooling branch where the sensing module 100 is located and the computing and cooling branch where the computing and scheduling module 300 is located, respectively. Deflectable guide vanes are located inside the fluid distribution cavity, and their cross-sectional shape preferably adopts a NACA symmetrical airfoil design to suppress flow separation when airflow passes over the vane surface at high angles of attack, reducing frictional drag and flow-induced noise. The output shaft of the micro servo driver is mechanically connected to the rotation shaft of the deflectable guide vane, driving the vane to rotate within a preset angle range, for example, -45 degrees to +45 degrees.

[0072] The vector flow guiding unit 202 dynamically adjusts the flow resistance ratio of each outlet branch by changing the deflection angle of the deflectable guide vanes. When the computation scheduling module 300 determines that the temperature of the image sensor in the sensing module 100 is too high and it is in a non-exposure period, it sends a command to deflect the guide vanes, reducing the flow resistance to the sensing heat dissipation branch and increasing the flow resistance to the computation heat dissipation branch, thereby achieving directional strong cooling.

[0073] Based on fluid network theory, the mapping relationship between the flow distribution ratio of each branch and the deflection angle of the guide vanes follows the following nonlinear equation:

[0074] ;

[0075] ;

[0076] In the formula: Indicates the first Each heat dissipation branch has a guide vane deflection angle of [value missing]. Volumetric flow rate at time; This indicates the total flow rate output by the centrifugal booster unit 201, which is determined by the current rotational speed. This represents the equivalent current resistance of all parallel branches, satisfying the reciprocal sum of parallel resistances. Indicates the first The dynamic flow resistance of each branch is ensured by the physical design. This avoids the singularity of a denominator of zero; Indicates the first The flow resistance of the basic pipeline of each branch is determined by the duct length, cross-sectional shape and wall roughness; This represents the local drag coefficient determined by the angle of the guide vanes. This coefficient is pre-determined through wind tunnel experiments and stored in the controller in the form of a look-up table. Indicates fluid density; This indicates the fluid flow velocity. The local drag coefficient is pre-calibrated. With angle Based on the functional relationship, the operation and scheduling module 300 can accurately calculate the angle and position commands required to maintain the target temperature, realize continuous adjustment of the heat flow vector rather than simple on / off control, and thus optimize the energy efficiency ratio of the whole machine.

[0077] In terms of the physical construction of the fluid channel, the fluid control module 200 and the inner wall of the mechanical support module 400 cooperate to form a sealed flow channel. To optimize the flow field quality, an expanding rectifier section is provided between the outlet of the centrifugal booster unit 201 and the inlet of the vector guide unit 202. The cross-sectional area of ​​this rectifier section gradually increases along the flow direction, using Bernoulli's principle to convert the high-speed dynamic pressure part at the outlet of the centrifugal booster unit 201 into static pressure, while simultaneously smoothing out the turbulent airflow lines, transforming strong turbulence into quasi-laminar flow, improving the uniformity of the airflow entering the vector guide unit 202, and ensuring the control accuracy and stability of the flow distribution.

[0078] See attached document Figure 4 , Figure 4 This is a schematic diagram of the logical architecture and data flow of a computation scheduling module according to an embodiment of the present invention. The computation scheduling module 300, as the logical control center of the system, adopts a heterogeneous computing architecture and is physically integrated into the back or bottom heat dissipation area of ​​the mechanical support module 400. The initial design intention of this architecture is to resolve the timing conflict between the high-throughput computing power requirements of high-level perception algorithms and the hard real-time response requirements of low-level motion control by utilizing physical isolation at the hardware level.

[0079] The computation scheduling module 300 includes a coprocessing unit 301 and a main processing unit 302, which interact with each other via a high-speed board-level bus such as a PCIe x1 channel or a high-speed SPI bus. The main processing unit 302 is configured as a high-performance system-on-a-chip running a Linux or Android operating system, responsible for processing high-bandwidth visual data and depth point cloud data from the sensing module 100. The coprocessing unit 301 is configured as a microcontroller or field-programmable gate array that does not run an operating system or runs a real-time operating system, responsible for performing microsecond-level sensor data acquisition and actuator loop control.

[0080] The coprocessor unit 301 is internally equipped with a direct memory access controller. This controller directly reads the angular velocity and acceleration registers of the inertial measurement unit 103 and the analog-to-digital conversion results of the temperature monitoring unit 104 through an independent I2C or SPI data channel, ensuring that the data acquisition process does not consume CPU computing power and guaranteeing the deterministic latency of the control loop. The coprocessor unit 301 runs a core motion masking and thermal management engine, which includes motion intensity determination logic and thermal credit integration logic.

[0081] For the motion intensity determination logic, the coprocessor unit 301 is configured to calculate the comprehensive motion intensity index of the system in real time. This index is used to quantify whether the external motion state of the system at the current moment is sufficient to mask the mechanical vibration generated by the operation of the fluid control module 200. The calculation of the comprehensive motion intensity index follows the following weighted Euclidean norm equation:

[0082] ;

[0083] In the formula: express The comprehensive motion intensity index at any given moment is a dimensionless scalar used to quantify the current external motion state of the system. This indicates that the angular velocity vector measured by the inertial measurement unit 103 contains three-axis components; The zero bias vector of the gyroscope is obtained through static initialization calibration; This indicates that the measured acceleration vector contains three-axis components; This represents the gravitational acceleration vector, obtained by separating the gravitational components through low-pass filtering. This represents the zero bias vector of the accelerometer; This represents the Euclidean norm (L2 norm) operation, used to extract the magnitude of a vector; The angular velocity weighting coefficient is proportional to the focal length of the visual imaging unit 101. That is, the longer the focal length, the more sensitive it is to blurring caused by angular velocity. This coefficient needs to be determined through visual calibration experiments. This represents the acceleration weighting coefficient, the value of which depends on the structural stiffness mode of the mechanical support module 400. The lower the stiffness, the larger the value of this coefficient.

[0084] Based on the calculated comprehensive motion intensity index, the coprocessing unit 301 further performs masking window determination. When the comprehensive motion intensity index... Exceeding the preset masking threshold When the visual imaging unit 101 receives a non-exposure state signal, i.e., during shutter closure, the coprocessing unit 301 generates a valid masking window signal. In this embodiment, the masking threshold... It is not a fixed value, but rather a parameter related to the current exposure time of the visual imaging unit 101. Inversely proportional K is a constant, which means that in short exposure mode, the system's tolerance to vibration is increased, thus allowing the cooling function to be activated more frequently.

[0085] For the thermal credit integral logic, the coprocessor unit 301 introduces the concept of thermal credit to manage the system's thermal inertia. Unlike traditional threshold-triggered control, the thermal credit integral logic allows component temperatures to exceed the rated operating point for short periods in exchange for quiet operation during the unmasked window. In the digital control system, the rate of change of the thermal credit integral is defined by the following discretized difference equation:

[0086] ;

[0087] In the formula: Indicates the first The cumulative thermal credit score for each control cycle; This represents the integral value from the previous period; Indicates the time step of the control cycle; Indicates the number of key heat-generating components; Indicates the first The heat capacity weighting coefficient of each component reflects the size and importance of the component's heat capacity. This indicates the temperature data collected by temperature monitoring unit 104. The current sampled temperature of each component; Indicates the first The target steady-state temperature of each component; The heat dissipation efficiency coefficient represents the ability of the fluid control module 200 to remove heat, and this coefficient is adjusted according to changes in ambient temperature. This indicates the current cooling power of the fluid control module 200.

[0088] The coprocessor unit 301 is configured to output control commands based on the status of the thermal credit score and the masking window signal. The specific execution logic is as follows:

[0089] When the system is stationary and the camera is in an unmasked window (i.e., when the system is stationary and the camera is in exposure), the forced fluid control module 200 is in a silent state with low speed or stops. At this time, due to… The item approaches 0, indicating a high level of credit score. It accumulates positively with increasing temperature. Once a valid masking window signal is detected (i.e., the system moves violently or the shutter closes), and the thermal credit is positive, the coprocessing unit 301 immediately outputs a high duty cycle pulse width modulation signal to the centrifugal booster unit 201, and simultaneously sends an angle command to the vector flow guiding unit 202, pointing it towards the component with the highest temperature. The largest item In the corresponding area, the accumulated thermal credits are rapidly consumed. If the thermal credits exceed the system's safety threshold, it indicates that thermal inertia has been exhausted. Regardless of whether there is a masking window, the coprocessor unit 301 triggers a forced cooling mode, sacrificing image quality to ensure hardware physical safety.

[0090] The main processing unit 302 runs environmental perception and macroscopic path planning algorithms. The main processing unit 302 receives visual and depth data, constructs a local environment map, and plans future motion trajectories. The main processing unit 302 is configured to send macroscopic motion commands containing expected motion events to the co-processing unit 301.

[0091] The macroscopic motion command includes the timestamp and duration of the anticipated large maneuver, such as a rapid turn or acceleration sprint. Upon receiving this command, the coprocessor unit 301 incorporates it into the feedforward control of the thermal management strategy. For example, if it is known that a significant turn will occur within the next 500 milliseconds, the coprocessor unit 301 will postpone its current cooling action, waiting for the turn to occur before simultaneously activating the centrifugal booster unit 201, thereby achieving an optimal balance between heat dissipation efficiency and image quality.

[0092] To ensure the coordinated operation of multiple actuators, a timing synchronization state machine also runs inside the coprocessing unit 301. This state machine aligns the trigger signal of the vision imaging unit 101 with the drive signal of the fluid control module 200 with microsecond-level precision. Specifically, before sending the camera exposure trigger signal, the state machine pre-calculates the speed decay time of the centrifugal booster unit 201 and outputs a reverse braking voltage to implement electronic braking when necessary, ensuring that the fan speed has dropped below the threshold that generates vibration and noise at the moment the exposure window opens, thus achieving strict temporal isolation.

[0093] Through the aforementioned hierarchical control architecture, the computation scheduling module 300 achieves high-power heat dissipation by utilizing physical motion as cover without sacrificing the overall computing performance of the system, thus resolving the inherent contradiction between heat dissipation and image stabilization in compact head devices.

[0094] See attached document Figure 5 , Figure 5This is a three-dimensional cross-sectional view and a schematic diagram of the internal air duct topology of a mechanical support module according to an embodiment of the present invention. The mechanical support module 400, as the physical skeleton of the system, not only provides mechanical mounting interfaces for the sensing module 100, the fluid control module 200, and the computing and scheduling module 300, but also defines closed airflow channels with specific fluid dynamic characteristics internally, forming an integrated heat dissipation architecture with interconnected structure.

[0095] The main structure of the mechanical support module 400 is composed of an outer frame 401 made of a high-strength material. Considering the lightweight requirements of the head-mounted device, in this embodiment, the outer frame 401 is preferably made of magnesium-aluminum alloy die casting or carbon fiber reinforced polyetheretherketone composite material to minimize the overall weight while ensuring structural rigidity. To suppress aerodynamic noise, the inner wall of the outer frame 401 is coated with open-cell sound-absorbing foam or micro-perforated plate. The porosity of this sound-absorbing material is designed to be 80% to 95%, and its thickness matches a quarter wavelength of the main noise frequency of the fluid control module 200, thereby effectively absorbing mid-to-high frequency noise generated by airflow turbulence, such as the 1kHz to 5kHz frequency band.

[0096] In terms of fluid topology design, the mechanical support module 400 internally comprises a main air intake channel 402, a flow distribution chamber 403, and multiple independent heat dissipation branches. A dustproof mesh is installed at the inlet of the main air intake channel 402, with its aperture set according to IP54 or higher protection standards. The cross-sectional shape of the main air intake channel 402 exhibits a contraction-expansion Venturi geometry along the airflow direction. Utilizing the principle of fluid continuity, it accelerates the airflow and stabilizes the static pressure at the throat position, rectifying the airflow field before entering the fluid control module 200 and reducing inlet vortex distortion.

[0097] The flow distribution chamber 403 is located downstream of the fluid control module 200 and physically encloses the vector flow guiding unit 202. The flow distribution chamber 403 extends into two main physical flow channels: a first heat dissipation flow channel 404 pointing to the sensing module 100 and a second heat dissipation flow channel 405 pointing to the computing and scheduling module 300.

[0098] Based on an on-demand thermal management strategy, and to accommodate the heat dissipation needs of different modules, the first heat dissipation channel 404 and the second heat dissipation channel 405 have differentiated hydraulic diameter designs. The first heat dissipation channel 404 is configured as a flat laminar flow channel, designed to perform surface-swept cooling of the visual imaging unit 101, avoiding sensor micro-displacement caused by airflow pulsation. The second heat dissipation channel 405 is configured as a turbulent flow channel with turbulence columns on its inner wall, designed to perform impact cooling of the heat dissipation fins of the computing scheduling module 300, thereby disrupting the boundary layer and improving the heat transfer coefficient. The geometric parameters of the channel cross-sections are designed according to the following hydraulic diameter calculation formula:

[0099] ;

[0100] In the formula: Indicates the equivalent hydraulic diameter of the flow channel, in meters; This refers to the cross-sectional area of ​​the flow channel perpendicular to the flow direction. This area must be greater than 50% of the outlet area of ​​the fluid control module 200 to avoid excessive back pressure. This represents the wetted perimeter of the flow channel cross-section, i.e., the circumference of the fluid in contact with the channel wall. The cross-sectional area is adjusted... With wet week The ratio, combined with a preset flow rate range, determines the Reynolds number (R) of the first heat dissipation channel 404 in this embodiment. The flow rate is controlled below the critical value to maintain laminar flow; at the same time, the Reynolds number of the second heat dissipation channel 405 is increased to above 4000 to form fully developed turbulence, thereby achieving a physical balance between noise reduction requirements and heat dissipation performance.

[0101] The mechanical support module 400 also includes an embedded heat sink 406, which is physically coupled to the heat source surface of the computing and scheduling module 300 and extends into the interior of the second heat dissipation channel 405. The heat sink 406 is made of copper or aluminum alloy and consists of a substrate and an array of microchannel fins. To maximize heat dissipation performance per unit volume, the thermal resistance design of the heat sink 406 satisfies the following one-dimensional steady-state heat conduction equation:

[0102] ;

[0103] In the formula: This represents the total thermal resistance from the junction temperature of the heat source to the air in the fluid passage, expressed in K / W. This indicates the thickness of the thermal interface material, which is controlled to be below 0.1 mm in the manufacturing process. Indicates the thermal conductivity of the thermal interface material; This indicates the effective physical contact area between the heat source and the heat sink substrate. Fin efficiency is the ratio of actual heat dissipation to ideal isothermal fin heat dissipation, typically ranging from 0.6 to 0.9. This represents the convective heat transfer coefficient, which is proportional to the nth power of the flow velocity (n is usually taken as 0.5 to 0.8). This represents the total heat transfer surface area of ​​the fin array.

[0104] This formula reveals that the mechanical support module 400 maximizes the contact area. Total heat exchange surface area Furthermore, high thermal conductivity thermal interface materials are selected to reduce total thermal resistance and ensure thermal stability under high computing loads.

[0105] For vibration isolation, the mechanical support module 400 is equipped with a suspended mounting base 407 to support the sensing module 100. The suspended mounting base 407 is connected to the outer frame 401 via a viscoelastic damper, forming a mass-spring-damping system. To quantify the vibration isolation effect, the dynamic characteristics of the suspended mounting base 407 are defined by the vibration transmissibility formula:

[0106] ;

[0107] In the formula: It represents the vibration transmissibility, which is the ratio of the output vibration amplitude (sensing module 100 end) to the input vibration amplitude (external frame 401 end); The damping ratio is indicated by the material loss factor of the viscoelastic damper. In this embodiment, a damping ratio of 0.1 to 0.3 is preferred to balance resonance peak suppression and high-frequency attenuation. This represents the frequency ratio, specifically the ratio of the rotational frequency of the interference frequency fluid control module 200 to the natural frequency of the suspension system. According to the vibration isolation principle, when the frequency ratio... At this time, the system enters the vibration isolation zone. In this embodiment, by adjusting the mass of the suspension mounting base 407 and the stiffness of the damper, the natural frequency is designed to be far below the minimum operating frequency of the fluid control module 200 (for example, the natural frequency is designed to be 20Hz, while the minimum operating frequency of the fan is 60Hz), so that the frequency ratio is... , thereby ensuring This achieves vibration attenuation at the physical level.

[0108] Furthermore, the mechanical support module 400 is covered with an electromagnetic shielding layer on the outside of the outer frame 401. This shielding layer is formed by spraying conductive paint or attaching metal foil and is electrically connected to the system's digital ground plane to provide a low-impedance return path. The shielding layer is used to block the influence of external electromagnetic interference on the internal high-frequency digital signals, and also to prevent the high-frequency harmonics generated by the motor drive of the fluid control module 200 from radiating outwards. At the connection interfaces of each module, the mechanical support module 400 is provided with sealing O-ring grooves, inlaid with fluororubber O-rings. This material is selected because of its excellent high-temperature resistance and anti-aging properties to ensure the airtightness of the fluid channels and prevent efficiency reduction or dust intrusion caused by cooling airflow leakage.

[0109] This embodiment constructs a humanoid robot head system for post-disaster search and rescue. The environment is characterized by high temperature and the robot's high-frequency running and jumping movements.

[0110] Sensing module 100: Visual imaging unit 101 uses a Sony IMX global shutter sensor with a resolution of 5 megapixels; Inertial measurement unit 103 uses a Bosch BMI088 six-axis industrial-grade IMU; Temperature monitoring unit 104 uses an NTC10K thermistor, which is attached to the SoC surface and the camera module backplate respectively.

[0111] Fluid control module 200: Centrifugal booster unit 201 uses a custom 30mm high static pressure turbine fan with a maximum speed of 15,000 RPM and a response time (0 to 10,000 RPM) of less than 150ms; Vector guide unit 202 uses deflection blades driven by micro servo motors.

[0112] The computation scheduling module 300: the main processing unit 302 adopts the NVIDIA Jetson Orin NX module; the coprocessing unit 301 adopts the STM32G4 series MCU with an operating frequency of 170MHz.

[0113] Mechanical support module 400: It adopts a magnesium-aluminum alloy frame, and the surface of the internal flow channel is coated with Teflon to reduce flow resistance.

[0114] During the search and rescue operation, the coprocessor unit 301 reads BMI088 data at a frequency of 1kHz via DMA. When the robot performs a rapid traversal of the rubble, the IMU measures that the angular velocity ωω of the robot body exceeds 3 rad / s, and the linear acceleration aa exceeds 2g. Based on the formula, the comprehensive motion intensity index MtMt is calculated. At this point, MtMt is much greater than the threshold MthMth, and the coprocessor unit 301 determines that the robot is currently in a motion masking window. Although the SoC core temperature is only 65℃ at this time, the system detects a positive thermal credit and immediately sends a 100% duty cycle PWM signal to drive the fan to run at full speed and adjusts the airflow vanes to point towards the SoC heat sink fins. The resulting mechanical vibration is masked by the large movement of the robot body and does not affect the navigation and positioning accuracy.

[0115] Static imaging coordination: When the robot reaches the observation point and brakes to a stop, the IMU reading drops. 50ms before the predicted stop of motion, the coprocessor unit 301 sends a reverse braking voltage to the fan, causing its speed to drop below 2000 RPM within 100ms.

[0116] After confirming that the fan is running at low speed and the camera body is stable, a trigger signal is sent to the visual imaging unit 101 for multi-frame exposure. Due to the absence of high-frequency micro-flicker, clear edge images and depth maps are obtained.

[0117] To verify the effectiveness of this invention, a simulation test platform was built. The platform includes a six-degree-of-freedom motion simulator, a temperature control chamber, and an ISO12233 resolution test chart.

[0118] Control group: A traditional PID temperature control strategy was used. The fan speed was linearly related only to the temperature, without considering the robot's motion state or camera exposure sequence.

[0119] Experimental group: A strategy based on inertial fluid vector control and vibration masking was adopted. The motion intensity threshold Mth=1.5 was set, and the single exposure time Texp=10ms.

[0120] Image clarity: During the robot's cyclical motion-stop-shoot-move action, the average MTF50 value of the images at the static shooting moment is statistically analyzed. The higher the value, the clearer the image.

[0121] Attitude calculation noise: Statistical angular velocity noise density of the IMU output; the lower the value, the better.

[0122] Core temperature stability: The highest junction temperature of the SoC after 30 minutes of continuous high load operation.

[0123] The graph consists of three vertically arranged subgraphs, with time as the horizontal axis.

[0124] The above sub-graph shows the overall motion intensity Mt represented by the solid blue line and the fan speed represented by the dashed red line. It can be clearly observed that the red curve strictly follows the peak of the blue curve, exhibiting a pulse-like characteristic of rotating when moving and stopping when stationary.

[0125] Neutron plot: The solid green line represents the fan speed of the control group. Due to the slow rise in core temperature, the fan speed increases in a stepwise manner and remains at a high level, which is not correlated with the motion state.

[0126] The black pulses in the image below represent the camera exposure times. In the experimental group, the exposure pulses all fell at the troughs of the fan speed; while in the control group, the exposure pulses often overlapped with high fan speeds.

[0127] Left column: The mean MTF50 of the experimental group was 0.85 lp / mm with a very small standard deviation; the mean of the control group was 0.52 lp / mm, and there was obvious data dispersion.

[0128] Intermediate column: The noise density of the experimental group was reduced by about 40% compared with the control group, indicating that eliminating fan vibration significantly improved the signal-to-noise ratio of the IMU in the low-frequency band.

[0129] Right column: The highest temperature in the experimental group was 78℃, and in the control group it was 75℃. Although the temperature in the experimental group was slightly higher, it was still within a safe range, proving that the pulsed heat dissipation using the moving window was sufficient to maintain thermal equilibrium.

[0130] Test metrics Measurement Dimensions control group experimental group Performance improvements / changes MTF50 Image center sharpness 0.52±0.15 0.85±0.02 Clarity improved by 63%, and stability significantly enhanced. GyroNoise Gyroscope noise density 0.012 0.007 Noise floor reduced by 41.6% ThermalLimit SoC maximum junction temperature 75.0 78.0 Temperature rose slightly (4%), but did not reach the throttling threshold.

[0131] As shown in the table and experimental results above, although this invention makes a slight strategic concession in absolute heat dissipation capacity, it successfully achieves a significant improvement of 63% in image clarity and a reduction of 41.6% in inertial measurement noise. This indicates that the dynamic heat dissipation and static imaging strategy proposed in this invention effectively solves the industry problem of active heat dissipation vibration interference perception in the head space of highly dynamic humanoid robots without sacrificing system thermal safety, and achieves globally optimal control of multiphysics fields.

[0132] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various 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 modular humanoid robot head intelligent control and management system, characterized in that, include: The module comprises a sensing module, a fluid control module, a computation and scheduling module, and a mechanical support module. The perception module integrates a visual acquisition component and an inertial measurement component, and is configured to acquire environmental information and motion state information and output image data sequences and kinematic data. The fluid control module is located in the internal cavity of the mechanical support module and is configured to generate heat dissipation airflow and establish a controlled heat dissipation fluid field by changing the airflow direction. The computation scheduling module is connected to the sensing module and the fluid control module respectively, and is configured to calculate a comprehensive motion intensity index based on the kinematic data, determine a motion masking window based on the comprehensive motion intensity index, and send a drive command to the fluid control module during the motion masking window or during the non-exposure period of the visual acquisition component. The mechanical support module is configured to provide a physical installation interface and internally defines a fluid channel that connects to the fluid control module; The fluid control module responds to the drive command by performing a heat dissipation action within the fluid channel, and performs thermal management on the sensing module or the computation scheduling module.

2. The modular humanoid robot head intelligent control management system according to claim 1, characterized in that, The sensing module adopts a rigid-flexible composite board architecture. The inertial measurement unit is rigidly connected to the geometric center of the mechanical support module via a surface mount process, and is used to characterize the motion state of the fuselage coordinate system. The visual acquisition component establishes a mechanical decoupling zone between the inertial measurement component and the optical imaging domain through a flexible plate. It decouples only from the high-frequency vibrations generated by the fan, while exhibiting rigidity in the low-frequency range of human movement to block high-frequency micro-vibrations from the fluid control module.

3. The modular humanoid robot head intelligent control management system according to claim 1, characterized in that, The computation scheduling module includes a real-time coprocessing logic unit and an application main processing logic unit; The real-time coprocessing logic unit is configured to read the angular velocity vector and acceleration vector output by the inertial measurement component via direct memory access, and calculate the comprehensive motion intensity index based on the weighted Euclidean norm equation. The real-time coprocessing logic unit is also configured to maintain thermal credit scores and, when the motion masking window is detected and the thermal credit score is positive, output a high duty cycle pulse width modulation signal to drive the fluid control module.

4. The modular humanoid robot head intelligent control management system according to claim 3, characterized in that, The calculation of the comprehensive exercise intensity index involves angular velocity weighting coefficient and acceleration weighting coefficient; The real-time coprocessing logic unit is configured to compare the comprehensive motion intensity index with the masking threshold to generate a masking window signal; The masking threshold is configured to be inversely proportional to the current exposure time parameter of the visual acquisition component.

5. The modular humanoid robot head intelligent control management system according to claim 1, characterized in that, The fluid control module includes a centrifugal booster assembly and a vector flow guide assembly; The centrifugal booster assembly is suspended and installed inside the mechanical support module via an elastic damping element, and is configured to generate a high-pressure pulsed airflow. The vector flow guiding component is located on the air outlet path of the centrifugal booster component, and includes deflectable guide vanes. It is configured to receive angle position commands and drive the deflectable guide vanes to deflect, thereby directionally guiding the airflow to the target heat dissipation area.

6. The modular humanoid robot head intelligent control management system according to claim 5, characterized in that, The vector flow guide assembly is configured to dynamically distribute the volumetric flow rate through different heat dissipation branches by adjusting the deflection angle of the deflectable flow guide vanes. The operation scheduling module has a pre-stored mapping table of local resistance coefficient and deflection angle, and calculates the angle position command based on the difference between the target temperature and the current temperature to adjust the flow resistance ratio of each heat dissipation branch.

7. The modular humanoid robot head intelligent control management system according to claim 1, characterized in that, The mechanical support module has a main air intake channel, a diversion chamber, a first heat dissipation channel pointing to the sensing module, and a second heat dissipation channel pointing to the computing and scheduling module. The first heat dissipation channel is configured as a flat laminar flow channel for surface-skimming cooling of the vision acquisition component. The second heat dissipation channel is configured as a turbulent channel with turbulence columns on the inner wall, which is used to perform impact cooling on the computing and scheduling module.

8. The modular humanoid robot head intelligent control management system according to claim 1, characterized in that, The sensing module also includes a depth detection component and a temperature monitoring component; The depth detection component is installed next to the vision acquisition component, and the parallelism deviation of their optical axes is within a preset tolerance range. The temperature monitoring component consists of a distributed array of negative temperature coefficient thermistors, which are respectively attached to the back plate of the vision acquisition component sensor, the surface of the depth detection component driver, and the surface of the computing scheduling module chip.

9. The modular humanoid robot head intelligent control management system according to claim 1, characterized in that, The visual acquisition component is configured as a complementary metal-oxide-semiconductor sensor component based on a global shutter, and has independent hardware trigger input pins and exposure synchronization output pins; The computation scheduling module is configured to send a synchronous trigger signal to the fluid control module, and before sending the exposure trigger signal of the vision acquisition component, calculate the rotational speed decay time of the fluid control module or output a reverse braking voltage, so that the fluid control module is in a silent state at the moment the exposure window opens.

10. The modular humanoid robot head intelligent control management system according to claim 1, characterized in that, The mechanical support module includes an outer frame made of high specific strength material and an open-cell sound-absorbing foam coated on the inner wall; The mechanical support module also includes a suspended mounting base for supporting the sensing module. The suspended mounting base and the outer frame are connected by a viscoelastic damper to form a mass damping system, and the natural frequency of the suspended mounting base is configured to be lower than the minimum operating fundamental frequency of the fluid control module.

Images