Expression controllable face robot based on magnetic control 4D printing technology

Abstract

The application discloses an expression-controllable face robot based on a magnetic control 4D printing technology, which is composed of five parts, including a bionic silica gel skin, a magnetic control 4D printing intelligent structure layer, a bionic rigid skull, a magnetic field generating module and a control system. The bionic silica gel skin is arranged on the surface of the bionic rigid skull in a shape-following mode, the magnetic control 4D printing intelligent structure layer is arranged on the inner side of the bionic silica gel skin and is composed of a plurality of magnetic control 4D printing intelligent structures arranged at face action coding system (FACS) control points; the magnetic field generating module is composed of a plurality of magnetic field generating units embedded in the bionic rigid skull. The control system independently adjusts each magnetic field generating unit to reconstruct a local dynamic magnetic field, non-contact triggering of the corresponding magnetic control 4D printing intelligent structure causes multimodal deformation, thereby driving the bionic silica gel skin to generate a target expression. The application can reduce the noise and driving lag caused by traditional mechanical transmission, and realize high-fidelity and fast-response natural expression simulation.

Description

Technical Field

[0001] This invention relates to the field of intelligent bionic robot technology, and more specifically, to a facial robot with controllable facial expressions based on magnetic control 4D printing technology. Background Technology

[0002] With the development of artificial intelligence and companion robots, achieving natural and empathetic human-computer interaction has become a core requirement. Facial expressions are the most direct medium for conveying emotions. The human face contains approximately 44 muscles, which work together to create various human facial expressions. The Facial Behavior Coding System (FACS) was first developed in 1978 by psychologists Paul Ekman and Wallace V. Friesen. Based on facial anatomy, this system breaks down facial expressions into action units for quantitative analysis, containing 46 basic action units that can recognize seven basic emotional expressions, including sadness, anger, and happiness.

[0003] Traditional facial robot actuation solutions can be mainly divided into three categories, but all have significant technical drawbacks. Mechanical transmission actuation (such as patent document CN121403464A) uses servo motors and linkage mechanisms. While this type of solution offers high control precision, the mechanical components are bulky, friction noise is high, and facial expressions are stiff. Fluid flexible actuation (patent document CN106393126B) uses micro-pumps and hydraulic channels. Although this type of solution produces smooth movements, it heavily relies on external pumps and valves, resulting in a large system, and exhibits significant fluid hysteresis, making it impossible to achieve high-frequency switching of facial expressions. Conventional intelligent material actuation (such as patent document CN109015598B) uses shape memory alloys. While this type of solution is lightweight and noiseless, the thermally actuated materials have extremely slow response and suffer from severe metal fatigue, resulting in a very short service life.

[0004] Magnetron 4D printing, building upon 3D printing, introduces "time" as a fourth dimension, enabling printed structures to undergo controlled changes in shape, performance, or function under the trigger of an external magnetic field. Magnetic field actuation, with its unique advantages such as non-contact remote control, deep tissue penetration, high safety, and millisecond-level response speed, shows enormous application potential in biomedicine, flexible robotics, and smart wearable devices. However, how to combine magnetron 4D printing technology with facial robots capable of controlling facial expressions to design a remotely driven, millisecond-level-response, and low-mechanical-noise facial robot remains a gap in the field. Therefore, there is an urgent need in this area for a novel facial robot solution that can overcome the aforementioned shortcomings. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a facial robot with controllable facial expressions based on magnetic control 4D printing technology. By using magnetic field generating modules distributed on different FACS control points to drive the corresponding magnetic control 4D printed smart structures to deform, millisecond-level response and high-precision expression control are achieved, thereby producing realistic facial expressions.

[0006] The technical solution adopted by this invention to solve its technical problem is: A facial robot with controllable facial expressions based on magnetron sputtering 4D printing technology includes: a biomimetic silicone skin 1, a magnetron sputtering 4D printed intelligent structure layer 2, a biomimetic rigid skull 3, a magnetic field generating module 4, and a control system 5. The biomimetic silicone skin 1 is conformally fitted onto the outer surface of the biomimetic rigid skull 3 under pre-tension stress and is fixedly connected to the biomimetic rigid skull 3. The magnetron sputtering 4D printed intelligent structure layer 2 is composed of multiple magnetron sputtering 4D printed intelligent structures 21, which are connected and fixed to the biomimetic silicone skin 1. The magnetic field generating module 4 is composed of multiple magnetic field generating units 41, and the biomimetic rigid skull 3 has pre-drilled mounting holes for each magnetic field generating unit 41. The magnetic field generating units 41 are respectively set and fixed in the pre-drilled mounting holes of the biomimetic rigid skull 3. The control system 5 is fixed in the biomimetic rigid skull 3, and two leads from each magnetic field generating unit 41 in the magnetic field generating module 4 are respectively connected to the control system 5, forming an independent electrical circuit for electrical connection.

[0007] Furthermore, the biomimetic silicone skin 1 is made of at least one of silicone, polyurethane elastomer, thermoplastic elastomer or other flexible polymer materials, with a thickness of 0.5 mm to 3 mm and an elastic modulus lower than that of the magnetron 4D printed smart structure 21.

[0008] Furthermore, the magnetron sputtering 4D printed smart structure layer 2 is composed of pre-set magnetron sputtering 4D printed smart structures 21 that can deform in response to a magnetic field. The magnetron sputtering 4D printed smart structure 21 is made of a smart material containing an elastomer and magnetic particles, wherein the elastomer is the matrix and the magnetic particles are uniformly dispersed in the matrix; the elastomer matrix is ​​at least one of thermoplastic polyurethane, polycaprolactone, silicone rubber, polydimethylsiloxane, elastic photosensitive resin or elastic epoxy resin; the magnetic particles are at least one of neodymium iron boron particles, samarium iron nitrogen particles or hard magnetic ferrite particles.

[0009] In one embodiment, the magnetron sputtering 4D printed smart structure 21 is a two-finger flexible hinge structure, which is physically divided into a left magnetic domain region, a right magnetic domain region, and a middle hinge region located between the two. The magnetic particles in the left and right magnetic domain regions have preset magnetization directions, and the magnetization directions of the left and right magnetic domain regions are the same, opposite, or set at a preset angle. The middle hinge region is in an unmagnetized state to form a purely elastic flexible hinge. Under the action of a local dynamic magnetic field, the left and right magnetic domain regions are deflected in concert by the magnetic torque, and open or fold with the middle unmagnetized hinge region as the bending center.

[0010] Furthermore, the number of the magnetically controlled 4D printed smart structure 21 and the magnetic field generating unit 41 are equal and correspond one-to-one. The magnetically controlled 4D printed smart structure 21 is located in front of its corresponding magnetic field generating unit 41, and the centroid of the magnetically controlled 4D printed smart structure 21 is located on the axis of the magnetic field generating unit 41.

[0011] Furthermore, the biomimetic rigid skull 3 is manufactured using additive manufacturing, injection molding, or composite material compression molding processes.

[0012] In one embodiment, the magnetic field generating unit 41 is an electromagnetic coil, which is cylindrical or a regular polygonal prism, and is composed of two parts: an excitation coil 411 and an iron core 412. The iron core 412 is a single-segment shaft or a double-segment stepped shaft. The excitation coil 411 is wound on the thin section of the single-segment shaft or double-segment stepped shaft of the iron core 412. The number of turns of the excitation coil 411 is 750 to 1400, and the wire diameter is 0.45 mm to 0.8 mm.

[0013] Furthermore, the control system 5 includes a host computer module and a slave controller built based on a microcontroller unit. The host computer module and the slave controller are communicatively connected. The slave controller is connected to a digitally adjustable constant current source drive module with current closed-loop detection and temperature compensation functions via a digital communication bus. The output terminal of the digitally adjustable constant current source drive module is connected to each of the magnetic field generating units 41. The host computer module defines different node addresses and issues control commands containing the absolute value of the target current. The digitally adjustable constant current source drive module independently adjusts the current magnitude, direction, and energizing time of the magnetic field generating unit 41 at the corresponding FACS control point to reconstruct the magnetic field strength and polarity generated by the magnetic field generating unit 41, so that the magnetically controlled 4D printed smart structure undergoes multimodal deformation, causing the biomimetic silicone skin 1 to produce corresponding expressions.

[0014] In one embodiment, the array layout of the magnetronically controlled 4D printed smart structure layer 2 maps the anatomical topology of human facial expression muscles; the plurality of magnetronically controlled 4D printed smart structures 21 serve as biomimetic artificial muscle units, with their centroid positions corresponding to the origin, insertion, or muscle belly positions of the facial expression muscles; the deformation response mode and force direction of the magnetronically controlled 4D printed smart structure 21 under the triggering of a local dynamic magnetic field are consistent with the biological contraction direction of the mapped facial expression muscles.

[0015] The present invention also provides a method for manufacturing an expression-controlled facial robot based on magnetic control 4D printing technology, comprising the following steps: S1 uses 3D printing technology or carbon fiber molding process to manufacture a biomimetic rigid skull 3 with a honeycomb sandwich structure, and simultaneously forms or processes the reserved installation holes at the corresponding facial FACS control points of the skull. S2 fixes the magnetic field generating unit 41 in the reserved mounting hole on the bionic rigid skull 3 by interference fit or special structural adhesive to form an arrayed magnetic field generating module 4. Then the lead wires of each magnetic field generating unit 41 are gathered and connected to the control system 5. S3 mixes an elastomer matrix with micron-sized magnetic particles at a set weight percentage to prepare a magnetron 4D printing ink, and uses an ink direct-write 4D printer to manufacture a magnetron 4D printed smart structure 21 and perform magnetic domain orientation programming. S4 uses a flexible adhesive to bond one side of the magnetically controlled 4D printed smart structure 21 programmed with magnetic domains to the inner surface of the biomimetic silicone skin 1, based on the topological layout of the facial FACS control points, to form the magnetically controlled 4D printed smart structure layer 2. S5 positions and tensions the biomimetic silicone skin 1, which is bonded with the magnetized 4D printed smart structure layer 2, and attaches it to the outer surface of the biomimetic rigid skull 3.

[0016] The present invention also provides a method for controlling an expression-controlled facial robot based on magnetic 4D printing technology, comprising the following steps: S1 Discretization and Trajectory Planning of Target Expression: Obtain the target expression command from external input, and based on the preset Facial Behavior Coding System (FACS) database, discretize the target expression command into a set of smooth motion trajectory curves of specific facial control points changing over time; S2 Establish a magnetic-solid coupling simulation mapping model: Based on the finite element analysis method, construct a nonlinear multiphysics simulation model of the biomimetic silicone skin and the magnetically controlled 4D printed smart structure, obtain the deformation response dataset of the control point under different magnetic field generating unit driving parameters, and establish a basic driving dictionary based on it; S3 Dynamic control parameter prediction and compensation based on machine learning: The smooth motion trajectory curve described in step S1 is input into the pre-trained machine learning algorithm model. Combined with the basic driving dictionary described in step S2, a series of nonlinear and dynamically changing electrical signal control parameters are predicted and output. The machine learning algorithm model is used to compensate for the viscoelastic hysteresis of the biomimetic silicone skin and the nonlinear deformation error of the magnetically controlled 4D printed smart structure, so that the deformation process is smooth. S4 Multi-channel Dynamic Magnetic Field Real-time Reconstruction: The drive control module analyzes the electrical signal control parameters and outputs dynamically modulated current signals to magnetic field generating units distributed at different anatomical locations to reconstruct the local spatial magnetic field in real time. S5 Smart Structure Non-Contact Actuation and Reset: The magnetically controlled 4D printed smart structure senses the local magnetic field to generate multimodal deformation, driving the bionic silicone skin to generate the target expression according to the smooth motion trajectory curve; after the target expression command ends, the control system 5 cuts off or gradually reduces the current, so that the bionic silicone skin rebounds to the initial state by relying on the elastic potential energy of the material itself.

[0017] S6 Expression State Perception and Closed-Loop Correction: The actual expression state generated by the bionic silicone skin is obtained through the image acquisition module, depth vision module, or facial deformation sensing module. Facial key point displacement, expression action unit intensity, control point trajectory deviation, or expression similarity evaluation parameters are extracted and compared with the target expression parameters to generate control correction quantities. The control system 5 updates the electrical signal control parameters in step S3 or the basic driving dictionary in step S2 according to the control correction quantities to reduce the deviation between the actual expression state and the target expression.

[0018] The beneficial effects of this invention are as follows: By independently adjusting the current amplitude, direction, and energizing duration of the magnetic field generating unit 41 through the control system 5, non-contact, refined local magnetic field driving of the magnetically controlled 4D-printed intelligent structure 21 can be achieved, realizing millisecond-level response and reducing the nonlinear hysteresis effect present in traditional pneumatic or partially intelligent material driving to a certain extent; the traditional motor linkage and valve mechanical structure is eliminated, reducing mechanical noise, while the magnetically controlled 4D-printed intelligent structure layer 2 directly acts on the FACS control points of the biomimetic silicone skin 1, achieving smooth, soft, and realistic micro-expression simulation; embedding the magnetic field generating module 4 into the biomimetic rigid skull 3 and using lightweight 4D-printed intelligent materials as actuators significantly reduces the overall weight and volume of the facial robot. Simultaneously, through the expression state perception and closed-loop correction mechanism, the electrical signal control parameters or basic driving dictionary can be dynamically updated according to the deviation between the actual expression state and the target parameters, improving the naturalness, anthropomorphism, and repeatability of expression synthesis. Attached Figure Description

[0019] Figure 1This is an exploded view of an expression-controlled facial robot structure based on magnetron sputtering 4D printing technology. The reference numerals are as follows: 1-Bionic silicone skin; 2-Magnetron sputtering 4D printing smart structural layer; 3-Bionic rigid skull; 4-Magnetic field generating module; 5-Control system.

[0020] Figure 2 The figure shows a front view of the FACS control point distribution. The labels are explained as follows: 101 - Left eyebrow head control point; 102 - Right eyebrow head control point; 103 - Left eyebrow tail control point; 104 - Right eyebrow tail control point; 105 - Left cheek control point; 106 - Right cheek control point; 107 - Left corner of mouth control point; 108 - Right corner of mouth control point; 109 - Lower lip center control point.

[0021] Figure 3 The attached diagram is a schematic diagram of the magnetic field generating module and a single magnetic field generating unit of the present invention. The reference numerals are explained as follows: 4-Magnetic field generating module; 41-Magnetic field generating unit; 411-Excitation coil; 412-Iron core.

[0022] Figure 4 The figure shows the deformation principle of the magnetically controlled 4D printed smart structure layer and a single structure under the action of a local magnetic field. The reference numerals are explained as follows: 2-Magnetically controlled 4D printed smart structure layer; 21-Magnetically controlled 4D printed smart structure.

[0023] Figure 5 This is a control flowchart of the facial expression controllable robot based on magnetic control 4D printing technology of the present invention. It is used to illustrate the process of generating target expressions and updating control parameters after the target expression command is processed through FACS trajectory planning, magnetic-solid coupling simulation mapping modeling, dynamic control parameter prediction, multi-channel dynamic magnetic field reconstruction, intelligent structure non-contact actuation, expression state perception and closed-loop correction. Detailed Implementation

[0024] The technical solution of the present invention will be further described in detail below through specific embodiments.

[0025] In this invention, unless otherwise specified, all raw materials and equipment used can be purchased from the market or are commonly used in the field. Unless otherwise specified, the methods in the following embodiments are conventional methods in the field.

[0026] Example: This embodiment provides an expression-controlled facial robot based on magnetic control 4D printing technology, which is applied to the head of a humanoid robot to generate biomimetic and realistic expressions.

[0027] The biomimetic silicone skin 1 is made of a high-strength polyurethane elastomer film with a thickness of 2 mm. It is coated and connected to the outside of the magnetron 4D printed smart structure layer 2 by coating, hot pressing or bonding. The elastic modulus of the biomimetic silicone skin 1 is lower than that of the magnetron 4D printed smart structure 21, so that the deformation of the magnetron 4D printed smart structure 21 can be effectively transferred to the surface of the biomimetic silicone skin 1.

[0028] The material of the magnetron sputtering 4D printed smart structure layer 2 is made of silicone rubber as the magnetically responsive polymer matrix, uniformly mixed with 65% by weight of micron-sized neodymium iron boron magnetic particles. Using an ink-writing magnetron sputtering 4D printer, this material is printed into a 2mm thick, double-finger flexible hinge-shaped magnetron sputtering 4D printed smart structure 21. Physically, this structure is divided into a left magnetic domain region, a right magnetic domain region, and a central hinge region connecting the two sides. During printing, a magnetic domain orientation device integrated at the nozzle programs the left and right magnetic domain regions of the magnetron sputtering 4D printed smart structure 21 with preset orientations, ensuring that the magnetic particles in the left and right magnetic domain regions have preset magnetization directions, and that these directions are the same, opposite, or at a preset angle. The central hinge region remains unmagnetized to form a purely elastic flexible hinge. The magnetron sputtering 4D printed smart structure 21 is then bonded to the biomimetic silicone skin 1 according to the FACS control point positions, forming the magnetron sputtering 4D printed smart structure layer 2.

[0029] The magnetically controlled 4D printed smart structure 21 can change its shape under the action of a magnetic field, causing the bionic silicone skin 1 bonded to the magnetically controlled 4D printed smart structure 21 to produce a bulge or depression with a maximum deformation of 0.1 mm to 5 mm in the vertical direction of the tangential plane compared to the same area of ​​the bionic silicone skin 1 bonded to the same magnetically controlled 4D printed smart structure 21 when there is no deformation. Stretching or contraction occurs at the corresponding FACS control points of the bionic silicone skin 1, producing corresponding facial expressions.

[0030] The biomimetic rigid skull 3 is made of carbon fiber reinforced material and formed into a lightweight honeycomb sandwich structure through additive manufacturing, providing rigid support for the magnetic field generating module 4.

[0031] The magnetic field generating module 4 specifically employs an electromagnetic actuation scheme. It consists of multiple cylindrical electromagnetic coils (i.e., magnetic field generating units 41) with a diameter of 30 mm and a height of 20 mm, embedded in the biomimetic rigid skull 3 at the FACS control point. Specifically, the cylindrical electromagnetic coil is composed of two parts: an excitation coil 411 and an iron core 412. The iron core 412 is a single-segment shaft shape, and the excitation coil 411 is wound around the shaft segment of the iron core 412. Each excitation coil 411 uses enameled wire with 750 turns and a wire diameter of 0.45 mm.

[0032] The control system 5 consists of a host computer terminal running signal processing programs and a lower-level microcontroller unit communicating via a data bus. During operation, the host computer software interface receives the target expression command and maps it into a byte stream command containing specific pin numbers, direction control bits, and target current digital values, which is then sent to the lower-level microcontroller unit. After parsing the command, the lower-level microcontroller unit sends a control message to the digitally adjustable constant current driver chip via a digital bus such as SPI or I2C. This driver chip automatically performs dynamic current stabilization based on an internal closed-loop feedback mechanism, outputting a precise millisecond-level constant current to the magnetic field generating units 41 distributed at each FACS control point on the face, thereby generating a local magnetic field that triggers the corresponding magnetically controlled 4D printed smart structure 21 to deform, completing the dynamic control of the expression.

[0033] To achieve a highly biomimetic mapping between the underlying structure and biological anatomy, this embodiment selected control points corresponding to 9 key action units (AUs) based on FACS, and mapped the topological position and deformation response mode of the magnetically controlled 4D printed smart structure 21 to the real human facial expression muscles. The specific distribution and mapping scheme is as follows: Eyebrow area: Four control points were selected for the inner eyebrow elevation / depression related to AU1 and the outer eyebrow elevation / depression related to AU2, namely, control point 101 at the head of the left eyebrow, control point 103 at the tail of the left eyebrow, control point 102 at the head of the right eyebrow, and control point 104 at the tail of the right eyebrow. The corresponding magnetically controlled 4D printed smart structure 21 spatially maps the human frontalis muscle and corrugator supercilii muscle. Its deformation response mode is: under the triggering of a positive or negative magnetic field, it produces a linear stretching or contraction vertically upward or downward along the frontal plane, replicating the eyebrow raising or frowning action caused by the contraction of the frontalis muscle.

[0034] Around the eyes and cheeks: Two control points related to the rise / fall of the cheeks were selected, namely control point 105 on the left cheek and control point 106 on the right cheek. The corresponding magnetically controlled 4D printed smart structure 21 maps the human zygomaticus major and zygomaticus minor muscles. Its deformation response mode is: to generate a three-dimensional spatial bending and bulging deformation that is obliquely outward and upward, in order to overcome the prestress of the bionic silicone skin 1 and simulate the "real smile" cheek bulge produced by the compression of the orbicularis oculi muscle when the zygomaticus major muscle contracts.

[0035] Perioral area: Two control points related to the lifting / lowering of the corners of the mouth (AU12 and AU15) were selected: control point 107 on the left corner of the mouth and control point 108 on the right corner of the mouth; and one control point in the middle of the lower lip related to the lower lip pull-down (AU16). The magnetically controlled 4D-printed smart structure 21 at both corners of the mouth maps to the levator anguli oris and depressor anguli oris muscles, generating diagonal lifting or lowering through magnetic field polarity reversal; the middle structure 21 of the lower lip maps to the mentalis and depressor labii inferioris muscles, with a response mode of vertical downward stretching.

[0036] To further demonstrate that the facial robot described in this invention can achieve high-fidelity and natural, smooth expression synthesis, this embodiment introduces multiphysics simulation and machine learning algorithms for collaborative control. The specific workflow is as follows: First, a multiphysics simulation model of the entire machine is established in the host computer of the control system, setting the constitutive parameters of the silicone skin hyperelastic and the magnetostriction parameters of the magnetostrictive 4D printed smart structure. By applying various ideal constant current excitation signals with amplitudes varying between 0 A and 5 A to the model, the three-dimensional spatial displacement response of each FACS control point under the action of a local magnetic field ranging from 0 mT to 300 mT is recorded, thereby generating an offline simulation dataset containing hundreds of thousands of sets of electrical signal-magnetic field-displacement responses.

[0037] Secondly, using this simulation dataset, a machine learning model for processing time series data is trained on a host computer; in this embodiment, a Long Short-Term Memory (LSTM) network is preferred. When actually running the "natural smile" expression, the control system first plans a non-linear, smooth, upward trajectory that conforms to the kinematic characteristics of human facial muscles. This trajectory sequence is then input into the trained LSTM model, which can predict and output a dynamically compensated target constant current time series sequence.

[0038] Subsequently, the lower-level microcontroller receives the current timing sequence and converts it into a bus control message, synchronously supplying a continuously dynamically changing current with an amplitude between 0 A and 3 A to the magnetic field generating units corresponding to the left and right cheek control points (105, 106) and the left and right corners of the mouth control points (107, 108). The magnetic field generating unit generates a modulated local dynamic magnetic field with a peak value between 150 mT and 200 mT in the external space of the skull. This magnetic field acts on the corresponding magnetron-controlled 4D printed smart structure 21. With the dynamic compensation algorithm of the LSTM model, the neodymium iron boron particles inside the magnetron-controlled 4D printed structure are subjected to force, causing the matrix to undergo dynamic bending deformation. This effectively compensates for the nonlinear elastic resistance and hysteresis effect of the 2 mm thick high-strength polyurethane film, allowing the cheeks and corners of the mouth to smoothly rise to a maximum convex deformation of 3 mm according to the planned S-shaped velocity curve. The entire process is free of mechanical noise and abrupt jitter, achieving high-fidelity and natural and smooth expression synthesis. After the command ends, the current decays smoothly according to the model plan, and the system naturally resets by relying on elastic potential energy.

[0039] Furthermore, to improve the anthropomorphism and repeatability of the synthesized expressions, this embodiment also includes an expression state perception and closed-loop correction process. During or after the generation of the target expression, the control system 5 acquires the actual expression state of the bionic silicone skin 1 through an image acquisition module, a depth vision module, or a facial deformation sensing module located in front of or inside the robot's head. The image acquisition module can be an RGB camera or a structured light camera, and the facial deformation sensing module can be a flexible strain sensor, a pressure sensor, or a displacement sensor.

[0040] The control system 5 extracts facial key point displacement, expression action unit intensity, control point trajectory deviation, or expression similarity evaluation parameters based on the acquired actual expression state, and compares them with the target expression parameters to obtain the deviation between the actual expression state and the target expression. Subsequently, the control system 5 generates control correction quantities based on the deviation, correcting the current amplitude, direction, energizing duration, or current change curve of each magnetic field generating unit 41; or the deviation is written into the basic driving dictionary for updating the electrical signal control parameters during subsequent target expression control. Through the above closed-loop correction process, the system can reduce the deviation between the actual deformation of the bionic silicone skin 1 and the target expression trajectory, thereby improving the naturalness, coordination, and anthropomorphism of the expression synthesis.

Claims

1. A facial robot with controllable facial expressions based on magnetic control 4D printing technology, characterized in that... Composed of five parts, including: a biomimetic silicone skin (1), a magnetically controlled 4D printed smart structure layer (2), a biomimetic rigid skull (3), a magnetic field generating module (4), and a control system (5); the biomimetic silicone skin (1) is conformally applied to the outer surface of the biomimetic rigid skull (3) under pre-tension stress and is fixedly connected to the biomimetic rigid skull (3); the magnetically controlled 4D printed smart structure layer (2) is composed of multiple magnetically controlled 4D printed smart structures (21), and the magnetically controlled 4D printed smart structures (21) are combined with the biomimetic silicone skin. The skin (1) is connected and fixed; the magnetic field generating module (4) is composed of multiple magnetic field generating units (41), and the bionic rigid skull (3) has reserved mounting holes for each magnetic field generating unit (41). The magnetic field generating units (41) are respectively set and fixed in the reserved mounting holes of the bionic rigid skull (3); the control system (5) is fixed in the bionic rigid skull (3), and the two leads of each magnetic field generating unit (41) in the magnetic field generating module (4) are respectively connected to the control system (5) to form an independent electrical circuit to realize electrical connection.

2. The facial expression-controlled robot based on magnetically controlled 4D printing technology according to claim 1, characterized in that, The biomimetic silicone skin (1) is made of at least one of silicone, polyurethane elastomer, thermoplastic elastomer or other flexible polymer materials, with a thickness of 0.5 mm to 3 mm and an elastic modulus lower than that of the magnetron 4D printed smart structure (21).

3. The facial expression-controlled robot based on magnetically controlled 4D printing technology according to claim 1, characterized in that, The magnetron 4D printed smart structure layer (2) is composed of a pre-set magnetron 4D printed smart structure (21) that can deform in response to a magnetic field. The magnetron 4D printed smart structure (21) is made of a smart material containing an elastomer and magnetic particles, wherein the elastomer is the matrix and the magnetic particles are uniformly dispersed in the matrix. The elastomer matrix is ​​at least one of thermoplastic polyurethane, polycaprolactone, silicone rubber, polydimethylsiloxane, elastic photosensitive resin or elastic epoxy resin. The magnetic particles are at least one of neodymium iron boron particles, samarium iron nitrogen particles or hard magnetic ferrite particles.

4. The facial expression-controlled robot based on magnetically controlled 4D printing technology according to claim 1, characterized in that, The magnetically controlled 4D printed smart structure (21) is a double-finger flexible hinge structure, which is physically divided into a left magnetic domain area, a right magnetic domain area and a middle hinge area between the two; the magnetic particles in the left and right magnetic domain areas have preset magnetization directions, and the magnetization directions of the left and right magnetic domain areas are the same, opposite or set at a preset angle; the middle hinge area is in an unmagnetized state to form a purely elastic flexible hinge; under the action of a local dynamic magnetic field, the left and right magnetic domain areas are affected by magnetic torque and deflected in concert, and open or fold with the middle unmagnetized hinge area as the bending center.

5. The facial expression-controlled robot based on magnetically controlled 4D printing technology according to claim 1, characterized in that, The number of the magnetic control 4D printing smart structure (21) and the magnetic field generating unit (41) are equal and correspond one-to-one. The magnetic control 4D printing smart structure (21) is located in front of its corresponding magnetic field generating unit (41), and the centroid of the magnetic control 4D printing smart structure (21) is located on the axis of the magnetic field generating unit (41).

6. The facial expression-controlled robot based on magnetically controlled 4D printing technology according to claim 1, characterized in that, The magnetic field generating unit (41) is an electromagnetic coil. The electromagnetic coil is cylindrical or a regular polygonal prism and is composed of two parts: an excitation coil (411) and an iron core (412). The iron core (412) is a single-segment shaft or a double-segment stepped shaft. The excitation coil (411) is wound on the thin shaft segment of the single-segment shaft or double-segment stepped shaft of the iron core (412). The number of turns of the excitation coil (411) is 750 to 1400, and the wire diameter is 0.45 mm to 0.8 mm.

7. The facial expression-controlled robot based on magnetically controlled 4D printing technology according to claim 1, characterized in that, The control system (5) includes a host computer module and a lower computer controller built on a microcontroller unit. The host computer module and the lower computer controller are connected in communication. The lower computer controller is connected to a digitally adjustable constant current source drive module with current closed-loop detection and temperature compensation functions through a digital communication bus. The output terminal of the digitally adjustable constant current source drive module is connected to each of the magnetic field generating units (41). The host computer module defines different node addresses and issues control commands containing the absolute value of the target current. The digitally adjustable constant current source drive module independently adjusts the current magnitude, direction and energizing time of the magnetic field generating unit (41) at the corresponding FACS control point to reconstruct the magnetic field strength and polarity generated by the magnetic field generating unit (41), so that the magnetic control 4D printed smart structure undergoes multimodal deformation and drives the bionic silicone skin (1) to produce corresponding expressions.

8. The facial expression-controlled robot based on magnetically controlled 4D printing technology according to claim 1, characterized in that, The array layout of the magnetically controlled 4D printed smart structure layer (2) maps the anatomical topology of human facial expression muscles; the multiple magnetically controlled 4D printed smart structures (21) serve as biomimetic artificial muscle units, with their centroid positions corresponding to the origin, insertion, or muscle belly positions of facial expression muscles; the deformation response mode and force direction of the magnetically controlled 4D printed smart structure (21) under the triggering of a local dynamic magnetic field are consistent with the biological contraction direction of the mapped facial expression muscles.

9. A method for manufacturing an expression-controlled facial robot based on magnetron sputtering 4D printing technology as described in any one of claims 1 to 8, characterized in that, Including the following steps: S1 uses 3D printing technology or carbon fiber molding process to manufacture a biomimetic rigid skull with a honeycomb sandwich structure (3), and simultaneously forms or processes the reserved installation holes at the facial FACS control point position corresponding to the skull. S2 fixes the magnetic field generating unit (41) in the reserved mounting hole on the bionic rigid skull (3) by interference fit or special structural adhesive to form an arrayed magnetic field generating module (4), and then gathers the lead wires of each magnetic field generating unit (41) and connects them to the control system (5). S3 mixes the elastomer matrix with micron-sized magnetic particles at a set weight percentage to prepare a magnetron 4D printing ink, and uses an ink direct writing 4D printer to manufacture a magnetron 4D printed smart structure (21) and perform magnetic domain orientation programming. S4 uses a flexible adhesive to bond one side of the magnetically controlled 4D printed smart structure (21) programmed with magnetic domains to the inner surface of the bionic silicone skin (1) according to the topological layout of the facial FACS control points, forming a magnetically controlled 4D printed smart structure layer (2). S5 positions and tensions the biomimetic silicone skin (1) with the magnetically controlled 4D printed smart structure layer (2) attached, and attaches it to the outer surface of the biomimetic rigid skull (3).

10. A method for controlling an expression-controlled facial robot based on magnetically controlled 4D printing technology as described in any one of claims 1 to 8, characterized in that, Including the following steps: S1 Discretization and Trajectory Planning of Target Expression: Obtain the target expression command from external input, and based on the preset Facial Behavior Coding System (FACS) database, discretize the target expression command into a set of smooth motion trajectory curves of specific facial control points changing over time; S2 Establish a magnetic-solid coupling simulation mapping model: Based on the finite element analysis method, construct a nonlinear multiphysics simulation model of the biomimetic silicone skin and the magnetically controlled 4D printed smart structure, obtain the deformation response dataset of the control point under different magnetic field generating unit driving parameters, and establish a basic driving dictionary based on it; S3 Dynamic control parameter prediction and compensation based on machine learning: The smooth motion trajectory curve described in step S1 is input into the pre-trained machine learning algorithm model. Combined with the basic driving dictionary described in step S2, a series of nonlinear and dynamically changing electrical signal control parameters are predicted and output. The machine learning algorithm model is used to compensate for the viscoelastic hysteresis of the biomimetic silicone skin and the nonlinear deformation error of the magnetically controlled 4D printed smart structure, so that the deformation process is smooth. S4 Multi-channel Dynamic Magnetic Field Real-time Reconstruction: The drive control module analyzes the electrical signal control parameters and outputs dynamically modulated current signals to magnetic field generating units distributed at different anatomical locations to reconstruct the local spatial magnetic field in real time. S5 Smart Structure Non-contact Actuation and Reset: The magnetically controlled 4D printed smart structure senses the local magnetic field to generate multimodal deformation, driving the bionic silicone skin to generate the target expression according to the smooth motion trajectory curve; after the target expression command ends, the control system (5) cuts off or gradually reduces the current, so that the bionic silicone skin rebounds to the initial state by relying on the elastic potential energy of the material itself. S6 Expression State Perception and Closed-Loop Correction: The actual expression state generated by the bionic silicone skin is obtained through the image acquisition module, depth vision module or facial deformation sensing module. The facial key point displacement, expression action unit intensity, control point trajectory deviation or expression similarity evaluation parameters are extracted and compared with the target expression parameters to generate control correction amount; the control system (5) updates the electrical signal control parameters in step S3 or the basic driving dictionary in step S2 according to the control correction amount to reduce the deviation between the actual expression state and the target expression.

Images