Bionic robot skin layered structure

Abstract

The invention discloses a bionic robot skin layered structure, and relates to the field of bionic robots, and the bionic robot skin layered structure comprises a skull base layer as a support foundation of the whole structure; the muscle driving layer is arranged on the outer side of the skull base layer and comprises a plurality of driving units, and each driving unit corresponds to a specific muscle group of the human face; the fat buffer layer covers the outer side of the muscle driving layer and is made of an elastic platinum silica gel material; and the leather outer layer is positioned on the outermost side and is made of a platinum silica gel material with the texture similar to that of the human skin. According to the four-layer bionic skin structure based on the human face anatomical structure, comprehensive improvement of high bionic, fine driving, material characteristic optimization and high expandability of the robot skin structure is achieved.

Description

Technical Field

[0001] This invention relates to the field of biomimetic robots, and more particularly to a layered skin structure for a biomimetic robot. Background Technology

[0002] With the rapid development of robotics technology, bionic robots are increasingly being used in fields such as medicine, service, and entertainment. Among these applications, facial expression simulation is one of the key technologies for improving the anthropomorphism of robots and enhancing the human-computer interaction experience.

[0003] Currently, existing robotic facial skins suffer from the following technical defects: First, most use a solid silicone structure made of a single material, which is hard and has a color and feel that are very different from real skin, and is prone to cracking when the environment changes; second, they lack biomimetic design of the layered structure of skin, and cannot realistically simulate the mechanical properties and movement performance of human skin; third, the driving method mostly uses a single pulling point to drive the facial skin, which makes it difficult to achieve fine and natural facial expression changes, and lacks a biomimetic driving mechanism corresponding to human muscles and tendons.

[0004] In recent years, although some studies have attempted to use multi-layered composite structures to prepare simulated skin, such as three-layer composite simulated skin (surface, middle layer, and inner layer) for medical training models, as well as stretchable flexible skin with integrated sensors, these technologies mainly focus on tactile simulation or sensor integration. They have not yet addressed the need for facial expression driving by constructing a layered skin system and driving mechanism that corresponds to the human anatomy.

[0005] Therefore, how to provide a bionic robotic silicone skin that can realistically simulate the anatomical structure of the human face and achieve refined expression-driven facial expressions has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] This invention provides a biomimetic robot skin layer structure, comprising, from the inside out, the following layers: The skull base layer serves as the supporting foundation for the overall structure; The muscle-driven layer, located on the outer side of the skull base layer, includes multiple driving units, each corresponding to a specific muscle group of the human face; A fat buffer layer, covering the outer side of the muscle-driven layer, is made of elastic platinum silicone material; The outermost layer of dermis is made of platinum silicone material that has a texture similar to human skin.

[0007] The bionic robot skin layered structure described above, wherein the drive unit includes a motor, a lever arm, a connecting rod, and a front paddle; wherein the motor is fixed to the skull base layer, the output shaft of the motor is connected to one end of the lever arm, and the other end of the lever arm is connected to the front paddle via the connecting rod; the front paddle is attached to the inner side of the fat buffer layer.

[0008] The biomimetic robot skin layered structure described above, wherein the shape and size of the front paddle plate are designed to mimic the corresponding human muscle attachment area.

[0009] The biomimetic robotic skin layer structure described above, wherein the thickness of the fat buffer layer varies between 3mm and 10mm depending on the facial region.

[0010] The biomimetic robot skin layered structure described above, wherein the surface of the outer dermis has a fine skin texture structure and a thickness of 0.5mm-3mm.

[0011] The present invention also provides a driving system for the skin layering structure of the bionic robot, comprising: Multiple drive units, each corresponding to the main facial expression muscles; The controller is electrically connected to the motors of each drive unit; The facial expression data interface is used to receive external facial expression control signals; The controller drives the corresponding motor to rotate according to the received expression control signal, causing the front paddle to push the upper structure to deform, thereby presenting the corresponding expression on the outer surface of the dermis.

[0012] As described above, in the driving system, the data received by the expression data interface contains 52 blendshape coefficients from ARKit.

[0013] The drive system described above, wherein the human muscles corresponding to the drive unit include at least 30 of the following: frontalis muscle, corrugator supercilii muscle, orbicularis oculi muscle, zygomaticus major muscle, zygomaticus minor muscle, levator labii superioris muscle, orbicularis oris muscle, depressor anguli oris muscle, and platysma muscle.

[0014] The beneficial effects achieved by this invention are as follows: 1. Highly biomimetic: Based on the anatomical structure of the human face, the four-layer biomimetic skin structure fully simulates the layered structure and mechanical transmission mechanism of human skin, from skull support, muscle drive, fat buffer to the outer dermis. 2. Precision drive: It adopts a drive structure of motor-lever-link-front paddle, with the front paddle directly corresponding to the muscle attachment area, which can accurately simulate the pulling effect of muscle contraction on the skin, and achieve delicate and natural facial expression changes; 3. Optimized material properties: The fat buffer layer and the outer dermis use platinum silicone materials with different ratios, which not only ensures the skin's flexibility, elasticity and touch, but also avoids the problems of traditional solid silicone being easy to crack and having poor texture. 4. High scalability: The number and layout of the driving units can be adjusted according to the bionic accuracy requirements, and it supports data interface with existing facial capture standards (such as ARKit blendshapes), which facilitates data-driven automated facial expression control. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a schematic diagram of a layered structure of a bionic robot skin provided in Embodiment 1 of this application; Figure 2 This is a flowchart of a method for fabricating a layered structure for the skin of a biomimetic robot, as provided in Embodiment 3 of this application. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0018] Example 1 like Figure 1 As shown, Embodiment 1 of this application provides a bionic robot skin layer structure, including a skull base layer 100, a muscle drive layer 200, a fat buffer layer 300 and an outer dermis layer 400 arranged sequentially from the inside to the outside.

[0019] The skull base layer 100 is made of nylon material and is 3D printed. Its shape is reconstructed based on CT scan data of the human skull. The surface of the skull base layer 100 has multiple threaded mounting holes for fixing the motor of the drive unit.

[0020] The muscle-driven layer 200 includes multiple driving units, each including a motor, a lever arm, a connecting rod, and a front paddle plate. The motor is fixed to the corresponding mounting point on the skull base layer. One end of the lever arm is fixedly connected to the output shaft of the motor, and the other end is connected to one end of the connecting rod via a first ball joint. The other end of the connecting rod is connected to the connecting seat on the back of the front paddle plate via a second ball joint. The front paddle plate has a flat structure, and its shape is designed to mimic the attachment area of ​​the corresponding muscle. For example, the front paddle plate corresponding to the zygomaticus major muscle region is elongated and measures 20mm × 8mm × 2mm; the front paddle plate corresponding to the lateral region of the orbicularis oculi muscle is arc-shaped.

[0021] The fat buffer layer 300 is cast from E600AB type platinum silicone material with a thickness of 5mm. It covers the outside of the muscle driving layer 200 and encapsulates the front paddle inside. When the motor drives the front paddle to move, the fat buffer layer transmits and buffers the displacement, simulating the function of subcutaneous fat.

[0022] The outer dermal layer 400 is made of E610 type platinum silicone mixed with semi-transparent silicone, with a thickness of 0.5~2.2mm. Its outer surface is formed with fine leather texture by mold transfer. The outer dermal layer is closely attached to the outer surface of the fat buffer layer 300.

[0023] Example 2 Embodiment 2 of this application provides a driving system for a bionic robot skin layer structure, including: a controller, an expression data interface, and multiple driving units; The controller uses bus-based servos and communicates via a TTL serial bus (UART). It supports multi-servo series control and provides real-time feedback on position, speed, temperature, voltage, and load. All servos are connected in parallel on the same bus pair (TX / RX for TTL, or A / B for RS485), and are distinguished by ID numbers. The last servo typically does not require a terminating resistor (this is not necessary for short-range TTL). The facial expression data interface uses a CAN bus, USB, or Bluetooth module to receive facial expression control data from the host computer or facial capture device. When the facial expression data interface receives a data frame containing 52 blendshape coefficients from ARKit, the controller converts each blendshape coefficient into the target displacement of the corresponding drive unit according to a preset shape key-displacement mapping table. For example, when the "browDownLeft" coefficient is received as 0.5, the controller calculates that the drive unit corresponding to the left corrugator supercilii muscle region should produce a 2mm inward and downward displacement, and then outputs the corresponding bus signal to drive the motor to rotate. Through the lever arm and linkage transmission, the front paddle plate produces a corresponding displacement, pushing the fat buffer layer and the outer dermis to form a frowning action.

[0024] The construction process of the morphological key-displacement mapping table is specifically divided into the following sub-steps: Step S21: Using electromyography signal simulation and muscle activation inversion model, the given blendshape coefficients are converted into muscle activation vectors; Muscle activation level is introduced as an intermediate metric unit for expression-driven calculations, transforming the physiologically meaningless Blendshape coefficient space into a muscle activation space conforming to human anatomy, thus providing a physiological and physical basis for subsequent biomechanical transmission. The construction and data processing flow of the electromyography signal simulation and muscle activation level inversion model are as follows: Step S211: Establish the mapping matrix between muscle functional units and blendshape coefficients; First, we define the set of facial muscle functional units as follows: Where K=24, each corresponding to one of the 24 core facial muscles of the human face, including the zygomaticus major, orbicularis oculi, frontalis, corrugator supercilii, and depressor anguli oris; the mapping matrix between muscle functional units and blendshape coefficients is defined as follows: Its elements This represents the contribution weight of the j-th blendshape coefficient to the activation of the i-th muscle functional unit.

[0025] The mapping matrix F is constructed using principal component analysis under anatomical constraints. The specific implementation process is as follows: 1. Sample Data Acquisition: Ten healthy adult subjects (aged 20-35 years, 5 males and 5 females) were recruited. Surface electromyography (sEMG) signals of 24 facial muscles were acquired using an 8-channel surface electromyography (SEMG) analyzer at a sampling frequency of 1000 Hz. Simultaneously, a structured light 3D scanner was used to acquire facial 3D point cloud data at a sampling frequency of 30 Hz. Each acquisition corresponded to the complete activation process of a standard ARKitBlendshape expression, and each expression was acquired 5 times to complete the full sample data acquisition. Based on the acquired facial 3D point cloud data, the Blendshape linear blending model was used to extract 52 ARKit blendshape coefficients corresponding to each frame of point cloud, generating a 52-dimensional blendshape coefficient time series that corresponds one-to-one with the acquisition timestamp. Timestamp realignment was performed on the sEMG signals and the blendshape coefficient time series. 2. Electromyography (EMG) signal decoupling: After preprocessing the sEMG signal using bandpass filtering and power frequency notch filtering, the FastICA independent component analysis algorithm was used to remove ECG and motion noise interference, and the EMG signal time series of 24 facial muscles were extracted for independent activation. For each independent source EMG signal, the root mean square (RMS) feature was extracted using a sliding time window with a sliding window length of 50ms and a step size of 1ms, resulting in an EMG RMS feature time series that matched the sampling frequency. For the complete activation process of each expression, the RMS feature time series of the corresponding activated muscles was subjected to min-max normalization: the minimum RMS value in the neutral state was 0 (corresponding to complete muscle relaxation), and the maximum RMS value in the maximum activation state of the expression was 1 (corresponding to maximum muscle contraction), finally obtaining a 24-dimensional muscle activation time series with a fixed value range of [0,1]. This series was completely synchronized with the blendshape coefficient time series and had the same length. 3. Initial mapping relationship establishment: Canonical correlation analysis (CCA) was performed on the collected 52-dimensional Blendshape coefficient time series and 24-dimensional muscle activation time series to maximize the linear correlation coefficient between the two sets of variables and obtain the initial mapping matrix F; 4. Anatomical Constraint Correction: The initial mapping matrix is ​​regularized and corrected, with constraints including: ① Spatial Smoothness Constraint: The L2 norm difference between the row vectors of the mapping matrix corresponding to muscles at adjacent anatomical locations does not exceed 0.2, ensuring smooth and continuous activation of adjacent muscles; ② Antagonistic Muscle Constraint: For antagonistic muscle pairs (such as frontalis and corrugator supercilii, zygomaticus major and depressor anguli oris), their contribution weights satisfy... , where i and k are the muscle functional unit indices of antagonistic muscle pairs, and j is the Blendshape index, ensuring that the same Blendshape will not simultaneously activate antagonistic muscle groups beyond the physiological limit.

[0026] Step S212: Calculate the muscle activation vector based on the mapping matrix between muscle functional units and blendshape coefficients; For any input 52-dimensional Blendshape coefficient vector By solving a quadratic programming problem with physiological constraints, a 24-dimensional muscle activation vector is obtained through inversion. The expression for the quadratic programming problem is: ,in, The sparsity regularization coefficient is set to 0.05 in this embodiment to ensure that the muscle activation driven by facial expressions conforms to physiological sparsity (i.e., a single facial expression is mainly activated by only a few core muscles). Let be the i-th component in vector A, representing the activation level of the i-th muscle functional unit, with a value range of [0,1], where 0 represents complete relaxation and 1 represents maximum contraction; For antagonistic muscle pair aggregation, For set The p-th and q-th components in the equation represent the activation levels of two muscle functional units within the same antagonistic muscle pair, constraining... To ensure that both muscles are not overactivated at the same time, it is necessary to comply with the physiological limitations of human muscle movement. In natural facial expressions, antagonistic muscles have a "push-pull" relationship, where one muscle contracts while the other relaxes. The sum of their activation levels has a physiological upper limit.

[0027] This embodiment uses the OSQP open-source QP solver, which is compatible with embedded systems, to solve the aforementioned quadratic programming problem. The solution time is less than 1ms, meeting the requirements for real-time driving. Through the above inversion model, an analytical mapping relationship can be established between the robot's blendshape driving commands and human muscle physiological parameters, completely eliminating the meaningless numerical mapping of traditional black-box fitting.

[0028] Step S22: Based on the skin layer mechanical analysis model, the muscle activation vector is analyzed into the actual execution displacement vector of the driving unit; The construction and data processing of the skin layered biomechanical analysis model are specifically divided into the following sub-steps: Step S221: Construct the transfer function from the displacement of the driving element to the surface normal deformation; The biomimetic robotic skin layered structure described in this application includes a fat buffer layer and an outer dermis layer. The two layers are made of platinum-silicone blends with different ratios, resulting in different thicknesses and elastic moduli. The front-end paddle of the drive unit is encased within the fat buffer layer. When the motor drives the front-end paddle to generate a displacement u, this displacement is transmitted through the fat buffer layer to the outer dermis layer, ultimately forming a normal deformation on the skin surface. .

[0029] Displacement u of the driving unit and surface normal deformation Transfer function between Represented as: ,in, The thickness of the fat buffer layer; The thickness of the outer layer of dermis; The elastic modulus of the fat buffer layer material; The elastic modulus of the outer layer material of dermis; Material property parameters were obtained through prototype calibration experiments. The specific calibration method was as follows: a series of known displacements u were applied to a single drive unit on the prototype, and a laser displacement sensor was used to measure the surface deformation of the corresponding area. Substituting the measured data into the above function and performing nonlinear least squares fitting, we obtain... .

[0030] The transfer function has the following characteristics: First, it exhibits nonlinear saturation characteristics; when the displacement u is small... It exhibits approximately linearity; when the displacement approaches the skin's elastic limit, the gain automatically decreases to prevent excessive stretching and skin damage; secondly, the layered structure compensation term... The explicit inclusion of the thickness ratio and modulus ratio of the fat layer to the dermis ensures that the same displacement produces different surface deformations in different facial regions (such as the cheeks and forehead, where skin thickness and elasticity differ), which is more consistent with human physiology. Thirdly, the function's domain... It is strictly monotonic and has an analytic inverse function. This provides a mathematical basis for the subsequent inverse calculation of the driving displacement from the target deformation.

[0031] Step S222: Map the muscle activation vector to the target deformation vector on the skin surface; For the nth driving unit, the set of indices of its corresponding muscle functional units (i.e., facial muscles that can act) is denoted as That is, the drive unit is responsible for simulating the contraction effect of one or more synergistic muscles.

[0032] First, based on the muscle activation vector A obtained from step S21, calculate the overall activation level corresponding to the driving unit. ,Right now ,in It is a muscle functional unit index. , The first muscle activation vector A is the first muscle activation vector. The component represents the first component. Muscle activation level of individual muscle functional units For set Total number of indexes within; Furthermore, the overall activation level Converted into the target surface deformation of the driving unit The conversion formula is expressed as: ,in The maximum surface deformation that the nth driving unit can produce is determined through a tensile test (e.g., the cheek region driving unit). Forehead area ); The coefficient is a nonlinear coefficient used to simulate the "initiation-acceleration-saturation" characteristics of muscle contraction. In this embodiment, it is obtained through calibration experiments. ; The target surface deformation of all driving units is transformed using the above method, and the transformed vectors are then concatenated to form the target deformation vector. .

[0033] Step S223: Based on the inverse operation of the transfer function, the target deformation vector is inversely calculated into the initial displacement vector of the driving unit; Based on the inverse operation of the transfer function, the target deformation vector is... Each component in The initial displacement of the driving unit is calculated in reverse. ,Right now ,in It is the inverse function of the transfer function; all the preliminary displacements obtained after the conversion are concatenated into a preliminary displacement vector.

[0034] Step S224: Introduce the feedforward decoupling algorithm based on the principle of linear superposition of deformation fields to remove displacement redundancy from the initial displacement vector and obtain the final actual execution displacement vector; When multiple drive units work together, since the skin is a continuous elastic body, the surface deformation fields generated by each drive unit overlap spatially. If the preliminary displacement calculated independently by each unit is directly used... When driving the deformation, the actual surface deformation will deviate from the target deformation, resulting in displacement redundancy or conflict. To address this issue, this step introduces a feedforward decoupling algorithm based on the principle of linear superposition of deformation fields to remove displacement redundancy or conflict. The specific data processing flow is as follows: Define the deformation influence coefficient matrix Where N is the total number of driving units, matrix elements in This indicates the displacement of the nth drive unit relative to the nth drive unit. The additional deformation generated by the target action area of ​​each driving unit is represented by this matrix, obtained through prototype calibration experiments. The specific calibration method involves applying a unit displacement individually to each driving unit on the prototype, extracting the deformation response of the action areas of other driving units, and normalizing the result. matrix; Based on the deformation influence coefficient matrix, the actual deformation of the effective region of each driving unit The actual displacement of each drive unit The relationship is: To ensure the actual deformation of the operating area of ​​each driving unit Equal to the target deformation variable calculated Demand solution regarding The nonlinear equation system: The equations are solved using the Newton-Raphson iterative method, with each initial displacement... As the initial value for iteration, the final execution displacement of each driving unit is obtained after the iteration converges. Due to the matrix and transfer function All solutions are of known analytical form, and the computational load of the iterative process is controllable. The solution time on the STM32 microcontroller is about 2ms, which meets the real-time requirements.

[0035] The final execution displacement of each driving unit obtained by the solution The vectors are then concatenated to obtain the actual displacement vector.

[0036] Step S23: Store the one-to-one correspondence between the blendshape coefficients and the actual displacement vectors of the driving unit to obtain the shape key-displacement mapping table; The core data structure of the morphological key-displacement mapping table is defined as a triplet containing an index identifier, a 52-dimensional Blendshape coefficient vector, and an N-dimensional actual displacement vector of the driving unit, where N is the total number of driving units; the index identifier serves as the unique addressing primary key. The range of values ​​[0,1] for each ARKit standard Blendshape coefficient is discretized into 21 sampling points (0, 0.05, 0.1, ..., 1.0) with a step size of 0.05. Based on the anatomical constraints of human facial expressions, the full combination space of the 52-dimensional Blendshape coefficients is filtered: invalid samples with the sum of mutually antagonistic Blendshape combination coefficients greater than 1 are removed, as are non-physiological samples that simultaneously activate more than 8 Blendshapes. The filtered valid Blendshape coefficient combinations are sorted in ascending order of the number of activated Blendshapes in a single frame to generate a Blendshape coefficient sequence. For each group of 52-dimensional Blendshape coefficients in the Blendshape coefficient sequence, the following calculations are performed sequentially: the Blendshape coefficients are inverted into a 24-dimensional muscle activation vector through electromyography signal simulation and muscle activation inversion model; the muscle activation vector is converted into a target deformation vector based on the skin layer mechanical analytical model; the target deformation vector is inverted into a preliminary displacement vector through the inverse operation of the transfer function; and a feedforward decoupling algorithm based on the linear superposition principle of deformation field is introduced to remove displacement redundancy, thus obtaining the final actual execution displacement vector. Each Blendshape coefficient is assigned a unique index identifier to its corresponding actual displacement vector and stored in a predefined morphological key-displacement mapping table structure.

[0037] Example 3 like Figure 2 As shown, Embodiment 3 of this application provides a method for fabricating a layered structure for the skin of a bionic robot, comprising: Step S31: Prepare the skull base layer: Based on human skull data, use 3D printing or plaster casting to make a skull base layer model, and reserve motor mounting holes at predetermined positions; Step S32: Create a head model: Build a 3D model based on real facial data, make a data mold, print it out, pour wax material into it, and form a wax figure model. Step S33: Sculpt skin texture and eye details on the wax figure; Step S34: Make an epoxy resin mold: Apply a mixture of epoxy resin and curing agent to the surface of the sculpted wax figure, and after curing, form a second head mold. Step S35: Prepare the drive unit mounting structure: Install the motors of each drive unit on the skull base layer, and adjust the angles of the lever arm and connecting rod to make the front paddle plate in the initial position; Step S36, Layered casting: Place the assembled skull base layer and drive unit into the epoxy resin mold. First, pour the fat buffer layer material into the corresponding area. After it partially cures, pour the outer skin material. Finally, cure the whole thing into shape. Step S37, Demolding and Finishing: Remove the mold and perform necessary finishing and coloring treatment on the skin surface.

[0038] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.

Claims

1. A biomimetic robotic skin layered structure, characterized in that, Including the following, arranged sequentially from the inside out: The skull base layer serves as the supporting foundation for the overall structure; The muscle-driven layer, located on the outer side of the skull base layer, includes multiple driving units, each corresponding to a specific muscle group of the human face; A fat buffer layer, covering the outer side of the muscle-driven layer, is made of elastic platinum silicone material; The outermost layer of dermis is made of platinum silicone material that has a texture similar to human skin.

2. The bionic robotic skin layered structure according to claim 1, characterized in that, The drive unit includes a motor, a lever arm, a connecting rod, and a front paddle; wherein, the motor is fixed to the skull base layer, the output shaft of the motor is connected to one end of the lever arm, and the other end of the lever arm is connected to the front paddle through the connecting rod; the front paddle is attached to the inner side of the fat buffer layer.

3. The bionic robotic skin layered structure according to claim 2, characterized in that, The shape and size of the front-end paddle are designed to mimic the corresponding muscle attachment area in the human body.

4. The bionic robotic skin layered structure according to claim 1, characterized in that, The thickness of the fat buffer layer varies between 3mm and 10mm depending on the facial area.

5. The bionic robotic skin layered structure according to claim 1, characterized in that, The outer layer of the dermis has a fine textured surface and a thickness of 0.5mm-3mm.

6. A driving system for the bionic robot skin layering structure according to any one of claims 1-5, characterized in that, include: Multiple drive units, each corresponding to the main facial expression muscles; The controller is electrically connected to the motors of each drive unit; The facial expression data interface is used to receive external facial expression control signals; The controller drives the corresponding motor to rotate according to the received expression control signal, causing the front paddle to push the upper structure to deform, thereby presenting the corresponding expression on the outer surface of the dermis.

7. The drive system according to claim 6, characterized in that, The data received by the facial expression data interface contains 52 blendshape coefficients from ARKit.

8. The drive system according to claim 6, characterized in that, The human muscles corresponding to the driving unit include at least 30 of the following: frontalis, corrugator supercilii, orbicularis oculi, zygomaticus major, zygomaticus minor, levator labii superioris, orbicularis oris, depressor anguli oris, and platysma.

9. A biomimetic robot face, characterized in that, It includes the biomimetic robotic skin layering structure as described in any one of claims 1-5.

10. A biomimetic robot, characterized in that, It includes the bionic robot face as described in claim 9.

Images