Multimodal adaptive electrochemical synapse e-skin and its closed-loop anti-slip control system
By using a multimodal adaptive electrochemical synaptic electronic skin, combined with shear electrode sheets, ion gels, and photosensitive OECT channels, multimodal signal fusion is achieved, solving the problem of insufficient sensing capabilities of existing electronic skins in complex scenarios, improving the system's judgment accuracy and efficiency, and reducing power consumption.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN HOTCHIP TECH
- Filing Date
- 2026-04-10
- Publication Date
- 2026-06-26
AI Technical Summary
Existing electronic skin and its closed-loop anti-slip control system mainly rely on single-modal input, which has limited perception capabilities, especially in low-light grasping and concealed scenarios, leading to misjudgment and delays in complex scenarios.
A multimodal adaptive electrochemical synaptic electronic skin is adopted, which combines shear electrode sheets, ion gel, triboelectric layer and photosensitive OECT channel. Through multimodal sensing of shear force, optical and normal pressure, signal processing and control are performed by shared gate OECT synaptic array and spatiotemporal coding and weight memory to achieve multimodal signal fusion and closed-loop anti-slip control.
It improves the accuracy and efficiency of the system's judgment in complex scenarios, reduces power consumption, and has the ability to control release and perform long-term tasks.
Smart Images
Figure CN122018413B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of flexible electronic skin technology, and in particular to a multimodal adaptive electrochemical synaptic electronic skin and its closed-loop anti-slip control system. Background Technology
[0002] Electronic skin mimics the mechanoreceptors and neural synapses in human skin to enable it to perceive touch, store and process these sensations. The core technology utilizes organic electrochemical synaptic transistors to integrate perception and processing, combining tactile receptors with organic electrochemical synaptic transistors to form an "artificial afferent nerve." This enables slip recognition at low operating voltages and combines it with timing algorithms to improve recognition accuracy, demonstrating a closed-loop grasping and anti-slip process.
[0003] Existing electronic skin and its closed-loop anti-slip control system are usually based on normal pressure, emphasizing dendritic integration, spatiotemporal pulse coding, and closed-loop action. The feasibility of tactile pattern differentiation and anti-slip control at the hardware level has been verified. However, existing solutions mainly focus on normal pressure, and the direct measurement of shear / slip vector components and parallel fusion of low-light optical information are still limited. Moreover, the sensing, storage, and recognition processes mostly rely on single-modal input. When in low-light grasping and concealed scenarios, the sensing input capability of a single modality is extremely limited, and the data reliability of a single modality is low, which can easily lead to misjudgment and delay in complex scenarios. Summary of the Invention
[0004] The purpose of this invention is to provide a multimodal adaptive electrochemical synaptic electronic skin and its closed-loop anti-slip control system to solve the problems mentioned in the background art.
[0005] To achieve the above objectives, the present invention provides the following technical solution: a multimodal adaptive electrochemical synaptic electronic skin, comprising a multimodal electrochemical electronic skin, wherein the multimodal electrochemical electronic skin is composed of a sensing unit, a signal processing unit, and a basal layer unit, wherein the sensing unit is used for the multimodal electrochemical electronic skin to acquire stimulation signals, the signal processing unit is used for converting and transmitting the acquired signals, and the basal layer unit is used for integrating the sensing unit and the signal processing unit to form an electronic skin;
[0006] The sensing unit includes a shear electrode sheet, an ion gel, a triboelectric layer, and a photosensitive OECT channel. The shear electrode sheet, ion gel, triboelectric layer, and photosensitive OECT channel are all disposed on the substrate unit. The shear electrode sheet, triboelectric layer, and photosensitive OECT channel are stacked in the ion gel. The triboelectric layer is disposed below the shear electrode sheet, and the photosensitive OECT channel is disposed below the triboelectric layer.
[0007] Preferably, the signal processing unit includes a microdome PDMS array and an electrochemical synaptic transistor. The microdome PDMS array is disposed above the shear electrode sheet. The microdome PDMS array forms multiple regularly arranged micro dome-shaped protrusions on the ion gel using microfabrication technology. The electrochemical synaptic transistor is disposed within the photosensitive OECT channel.
[0008] Preferably, the base layer unit includes an insulating film and a flexible substrate. The insulating film is disposed below the photosensitive OECT channel, and the flexible substrate is disposed below the insulating film. The flexible substrate is used to support the sensing unit and the signal processing unit.
[0009] Preferably, the upper and lower electrodes of the shear electrode sheet are arranged in an interdigital configuration, the ion gel is used as the electrolyte for the electrochemical synaptic transistor, the ion gel is used to sense the pressure and strain of the multimodal electrochemical electronic skin and convert it into an electrical signal, the triboelectric layer uses a nanofiber membrane to detect dynamic stimulation based on the coupling effect of contact electrification and electrostatic induction, and the photosensitive OECT channel is used to fuse photochemically coupled photosynapses and tactile shear-triggered electrical synapses in the same channel.
[0010] A closed-loop anti-slip control system for multimodal adaptive electrochemical synaptic electronic skin includes:
[0011] A multimodal sensing layer, which captures stimulus signals by contacting the external environment and converts them into electrical signals;
[0012] A shared gate OECT synaptic array is electrically connected to a multimodal sensing layer. The shared gate OECT synaptic array is used to share the same common gate electrode and electrolyte layer among multiple OECT channels. The shared gate OECT synaptic array controls the synaptic weight distribution of the array by inputting the signal generated by the multimodal sensing layer as a gate voltage to the shared gate.
[0013] A spatiotemporal encoding and weighting memory is electrically connected to a shared gate OECT synaptic array. The spatiotemporal encoding and weighting memory generates a pulse sequence carrying time information by receiving signals from the shared gate OECT synaptic array and stores the weight information of synaptic connections.
[0014] The controller is electrically connected to a spatiotemporal coding and weight memory and a shared gate OECT synaptic array. The controller adopts a combination of threshold discrimination and lightweight timing network to realize closed-loop control of early warning of slip and adaptive re-grabbing.
[0015] An actuator, which is electrically connected to a controller, performs mechanical motion by receiving signals from the controller.
[0016] Preferably, the multimodal sensing layer achieves differential readout of shear force through shear electrode sheets, detects vibration signals using a triboelectric layer, utilizes a photosensitive OECT channel to achieve fusion of photosynapses and electrical synapses in the same channel and provides additional optical features under low illumination, and converts and transmits pressure and strain signals through an ion gel.
[0017] Preferably, the shared-gate OECT synaptic array adopts a cross-array structure of row shared gate and column source-drain. The synaptic unit in the shared-gate OECT synaptic array consists of a source, a drain, an organic semiconductor channel, an electrolyte layer, and a shared gate. Multiple synaptic units share a single gate electrode. The gate signal regulates the channel conductance through ion transport in the electrolyte layer. The source and drain are used to read the synaptic weight of the synaptic unit. Through the OECT structure and ion dynamics regulation, near-zero static power consumption is achieved in the unstimulated state.
[0018] Preferably, the spatiotemporal coding and the spatiotemporal coding in the weight memory simulate the information processing mode of brain neurons, encoding information simultaneously in the time and space dimensions. By applying continuous gate voltage pulses, the short-term plasticity of synapses is simulated to achieve dynamic caching and time-series processing of information. Furthermore, by activating different shared gate groups at different times, spatial distribution modulation of synaptic weights is achieved. Finally, the time pulse sequence is combined with the spatial activation mode to form an efficient spatiotemporal joint coding.
[0019] Preferably, the weight memory in the spatiotemporal encoding and weight memory changes the flow direction of ions in the OECT channel by applying positive and negative gate voltages, thereby adjusting the conductivity. After the gate voltage is removed, the doping state of the channel can be maintained, realizing non-volatile storage of weights. Furthermore, by controlling the amplitude and time of the gate voltage, continuous conductivity modulation is achieved, simulating the polymorphic weights of biological synapses.
[0020] Preferably, the controller performs lightweight feature extraction and feature encoding on the parameters of the multimodal perception layer, and fine-tunes the policy network using a lightweight PPO algorithm. It constructs a three-dimensional threshold space of object characteristics based on the parameters of the multimodal perception layer, and dynamically generates control strategies based on the object characteristics to achieve compliant grasping.
[0021] The technical effects and advantages of this invention are as follows:
[0022] This invention utilizes a multimodal sensing layer to perceive stimuli received by electronic skin. It combines photoelectrochemically coupled photosynapses with shear vectors and normal pressure to predict critical conditions in the early stages of slippage, providing diverse data for slippage detection. Based on the multimodal signals, a gate voltage is applied to the shared-gate OECT synaptic array, altering the ion flow direction in the OECT channel and regulating conductivity. This redistributes synaptic weights, while simultaneously feeding back object characteristics to the shared-gate OECT synaptic array for further synaptic weight adjustment. A control strategy is dynamically generated based on the object's characteristics. The system provides feedback adjustments to the actuators, altering their gripping force on objects to achieve closed-loop anti-slip control of the electrochemical synaptic electronic skin. Based on early critical conditions and synaptic weight updates in the multimodal phase, the actuators are adjusted in real time to improve the accuracy and efficiency of the system's judgment in complex scenarios. Furthermore, through an enhanced OECT structure and ion trapping and detrapping dynamics regulation, near-zero static power consumption is achieved without stimulation, along with controllable release to accelerate reset. Standby power consumption is near-zero, and the system maintains a long hold time and controllable reset after triggering, meeting the robot's long-duration task requirements. Attached Figure Description
[0023] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention, but do not constitute a limitation thereof. In the drawings:
[0024] Figure 1 This is a schematic diagram of the multimodal electrochemical electronic skin structure of the present invention;
[0025] Figure 2 This is a schematic diagram of the sensing unit structure of the present invention;
[0026] Figure 3 This is a schematic diagram of the signal processing unit structure of the present invention;
[0027] Figure 4 This is a schematic diagram of the base layer unit structure of the present invention;
[0028] Figure 5 This is a frontal cross-sectional view of the multimodal electrochemical electronic skin structure of the present invention;
[0029] Figure 6 This is a schematic diagram of the multimodal electrochemical electronic skin anti-slip control system of the present invention;
[0030] Figure 7 This is a schematic diagram of the electronic skin anti-slip control system of the present invention.
[0031] In the attached image:
[0032] 1. Microdome PDMS array; 2. Shear electrode sheet; 3. Ionic gel; 4. Triboelectric layer; 5. Photosensitive OECT channel; 6. Electrochemical synaptic transistor; 7. Insulating film; 8. Flexible substrate. Detailed Implementation
[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0034] This invention provides, for example Figures 1-5 The illustration shows a multimodal adaptive electrochemical synaptic electronic skin, comprising a multimodal electrochemical electronic skin, a sensing unit, a signal processing unit, and a basal layer unit. The sensing unit is used to acquire stimulus signals and is in direct contact with the external environment, responsible for capturing various stimuli received by the multimodal electrochemical electronic skin. The signal processing unit is used to convert and transmit the acquired signals, simulating the neurotransmitter transmission and learning process in the biological brain through organic electrochemical synaptic transistors. The basal layer unit is used to integrate the sensing unit and the signal processing unit to form the electronic skin. The basal layer unit integrates all functional layers on a substrate material, forming an adaptive electrochemical synaptic electronic skin structure by fusing the multimodal sensing unit with the efficient signal processing layer through a robust and self-healing basal layer unit integrated system.
[0035] The sensing unit includes a shear electrode 2, an ion gel 3, a triboelectric layer 4, and a photosensitive OECT channel 5. All three components are disposed on the substrate unit. They are stacked within the ion gel 3, with the triboelectric layer 4 positioned below the shear electrode 2 and the photosensitive OECT channel 5 positioned below the triboelectric layer 4. The shear electrode 2 senses shear force, slip, and normal pressure, while the ion gel 3 detects multiple [factors / conditions]. The modal electrochemical electronic skin senses the pressure and strain of the target surface using triboelectric layer 4 to detect vibrations. It utilizes photosensitive OECT channel 5 to provide cues of target surface texture and reflectivity under visually limited conditions, forming a multimodal perception. It integrates photoelectrochemically coupled photosynapses with tactile / shear-triggered electrical synapses, and combines them with shear vectors and normal pressure to establish a mechanical-electrical-optical coupled physical model. This model predicts the critical conditions for slippage, improves early judgment criteria before and during slippage, detects slippage trends in advance, adjusts grip force, and prevents objects from slipping.
[0036] By using shear vector and normal pressure distribution as mechanical inputs, current and voltage responses of electrical synapses as electrical inputs, and photocurrent intensity of photosynapses as optical inputs, a constitutive model of a single synapse is first established, followed by a spatial distribution model of a multi-synapse array. This yields the transfer function matrices for mechanical-electrical coupling and electrical-optical coupling, resulting in the physical model shown below:
[0037] (t)= .exp( )+ .exp( ).H( )
[0038] It represents the initial peak current of the i-th synapse, which is related to the sensitivity of the synapse material and the contact area;
[0039] These are the fundamental parameters for converting optical signals into electrical signals, serving as the reference values for optical input;
[0040] ( , Let be the relaxation time constant of the i-th synapse, and be the normal pressure. Shear force Related;
[0041] .exp( ) represents the decay of the initial residual current;
[0042] .exp( ).H( ) represents the effect of N light pulse sequences;
[0043] H( () is the unit step function, representing the pulse at time (t). occur.
[0044] The mechanical-electrical coupling transfer function matrix is the relaxation time constant. ( , ), represented as:
[0045] =
[0046] Indicates the normal pressure-to-electrical signal conversion coefficient;
[0047] This represents the shear force-to-electrical signal conversion coefficient.
[0048] The electrical-optical coupling transfer function matrix is represented as follows:
[0049] =
[0050] This represents the electro-optic coupling coefficient.
[0051] Input normal pressure With shear force Input optical pulse sequence, output postsynaptic current and relaxation time constant By collecting the output postsynaptic current The signal is processed and converted into a slip judgment, first by measuring the postsynaptic current. The signal undergoes feature extraction, and the dynamic friction coefficient and current fluctuation variance are calculated. When an object begins to slip slightly, even a very small displacement will be detected in the tactile sensor. / The fluctuation of the friction coefficient, i.e., the change in dynamic friction coefficient, is used to extract the high-frequency micro-vibration signal unique to the slip before slippage, calculate the critical conditions, and output the critical slippage distance, grip force adjustment amount, and slippage probability. A dynamic evolution equation in the time dimension is introduced, and finally, a multi-field coupled simulation model is established using COMSOL to simulate the slippage process of objects of different materials and shapes, verifying the consistency between the model prediction and the simulation results. Simultaneously, the output critical slippage distance, grip force adjustment amount, and slippage probability data are used as feature streams. Through the input feature streams, a lightweight temporal network is used for inference. For example, RVFL-LSTM is used as the lightweight temporal network model architecture to map nonlinear temporal relationships, predict future states, and combine stochastic vector functional connections (SVMs) with long short-term memory (LSTM). When inputting temporal data from the past t-100ms to time t, the slippage velocity in the future t+50ms is predicted, and the slippage probability is output. When the slippage probability reaches a preset value, a pre-slippage alarm is triggered.
[0052] The signal processing unit includes a microdome PDMS array 1 and an electrochemical synaptic transistor 6. The microdome PDMS array 1 is disposed above the shear electrode 2. The microdome PDMS array 1 forms multiple regularly arranged micro dome-shaped protrusions on the ion gel 3 through microfabrication technology. When subjected to external pressure, the microdome structure deforms, changing the contact area with the shear electrode 2, thereby converting the pressure signal into an electrical signal. Due to the stress concentration effect generated by its dome shape, a small force can induce significant deformation. Therefore, the shear electrode 2 based on the microdome array has high sensitivity, enhancing the sensing ability of the multimodal electrochemical electronic skin. The electrochemical synaptic transistor 6 is disposed in the photosensitive OECT channel 5. Through electrochemical reaction, ions are injected into or extracted from the photosensitive OECT channel 5, thereby changing its doping state and conductivity. The electrochemical synaptic transistor 6 uses ions to regulate the conductivity of the semiconductor channel, while simultaneously transmitting ions and electrons, thereby achieving efficient conversion of ion signals to electronic signals.
[0053] The base layer unit includes an insulating film 7 and a flexible substrate 8. The insulating film 7 is disposed below the photosensitive OECT channel 5. The insulating film 7 is a copolymer film of tetrahydrofurfuryl acrylate and diethylene glycol divinyl ether synthesized by vapor-initiated chemical vapor deposition. The insulating film 7 serves as a gate dielectric layer or an isolation layer between devices in the electrochemical synaptic transistor 6 and the multimodal electrochemical electronic skin, allowing ions to pass through rapidly to regulate channel conductivity while strictly blocking electron leakage. The flexible substrate 8 is disposed below the insulating film 7. The flexible substrate 8 is used to support the sensing unit and the signal processing unit. The flexible substrate 8 utilizes a supramolecular polymer containing hydrogen bonds or metal coordination bonds, which can make the material have both high tensile strength and toughness. Moreover, the dynamic bonds can break and reform when the material is damaged, thereby realizing a self-healing function. Through interface treatment and continuous casting, the sensing unit, the signal processing unit and the base layer unit are seamlessly integrated together to form a three-layer electronic skin structure, ensuring that the layers do not separate when bent or stretched.
[0054] The upper and lower electrodes of the shear electrode 2 are arranged in an interdigitated shape. Utilizing the interdigitated electrodes and the tilted / misaligned structure, the shear electrode 2 cleverly converts mechanical signals into precisely measurable capacitance changes. It consists of an upper electrode, a deformable dielectric layer, and a lower electrode. The interdigitated electrodes are mirror images of each other, significantly increasing the facing area between the electrodes within a limited space. This amplifies the capacitance value and its response to external force changes, improving sensitivity. When subjected to shear force, the upper electrode undergoes horizontal displacement relative to the lower electrode, causing a change in the facing area of the corresponding interdigitated electrodes—one side increases, the other decreases. By measuring the difference in capacitance, the magnitude and direction of the shear force can be calculated. Furthermore, in the initial state, the interdigitated electrodes are not perfectly aligned; a certain pre-set displacement acts as a bias point for the sensor, making it more sensitive to minute initial slippage. Any minute horizontal force... It will immediately change the current facing area, thereby generating a clear capacitance change signal. Ionic gel 3 is used as the electrolyte of electrochemical synaptic transistor 6. Ionic gel 3 is used to sense the pressure and strain of multimodal electrochemical electronic skin and convert it into electrical signals. Ionic gel 3 is used as a deformable dielectric layer between shear electrode sheets 2. Ionic gel 3 is a zwitterionic gel electrolyte, prepared by solvent-free UV crosslinking method. The dual ion gel can provide a richer variety of ions and transport paths, and respond to photogenerated charge carriers and tactile electrical signals. Ionic gel 3 contains equal amounts of positive and negative charge groups on the same monomer unit. There is a dipole-ion interaction between the zwitterionic network and electrolyte ions, which promotes the dissociation of ion pairs and anchors ions in the gel network, inhibiting their leakage. Moreover, the zwitterionic gel will expand moderately in the ion environment, which makes its network structure more compact and further hinders ion escape.
[0055] Triboelectric layer 4 utilizes the coupling effect of contact electrification and electrostatic induction to detect dynamic stimuli using a nanofiber membrane. As a nanofiber membrane prepared using electrospinning technology, the high specific surface area, high porosity, and tunable microstructure of electrospinning can significantly increase the triboelectric charge density, thereby improving output performance. When two different fiber membranes come into contact and separate, charge transfer occurs on the surface due to the different electron-acquiring capabilities of the materials, creating a potential difference between the electrodes. This drives electrons to flow through the external circuit, generating an electrical signal. When the electronic skin vibrates, it causes the two different fiber membranes inside to come into contact and separate, allowing for the detection of the vibration signal. The photosensitive OECT channel 5 is used to connect photoelectrochemically coupled photosynapses and tactile shear-triggered electrical synapses within the same channel. The photosensitive OECT channel 5 is a composite channel composed of an NDI-based n-type polymer and near-infrared sensitive perovskite quantum dots. The photosensitive OECT channel 5 is prepared using iodine-rich perovskite and can achieve a sensitive response to near-infrared light. The perovskite quantum dots absorb NIR / visible light, and photogenerated carriers are injected into the OECT channel to modulate the channel conductivity and update the photosynaptic weights. In addition, the shear electrode 2 and the triboelectric layer 4 generate electrical signals, which are coupled to the OECT gate through the electrolyte, allowing ions to be injected into the OECT channel to modulate the channel conductivity and update the electrical synaptic weights. The OECT channel acts as an ion-electron coupling converter, responding to both photogenerated carriers and electrophysiological ion injection, achieving the spatial and temporal summation of dual-mode signals.
[0056] A closed-loop anti-slip control system for multimodal adaptive electrochemical synaptic electronic skin, such as Figures 6-7As shown, the system includes a multimodal sensing layer, a shared-gate OECT synaptic array, a spatiotemporal encoding and weighting memory, a controller, and an actuator. The multimodal sensing layer captures stimulus signals through contact with the external environment and converts them into electrical signals. The multimodal sensing layer collects multimodal signals received by the electronic skin through sensing units, achieving optical-tactile-shear force three-modal coupling to prevent misjudgment and delay in complex scenarios. The shared-gate OECT synaptic array is electrically connected to the multimodal sensing layer. It allows multiple OECT channels to share the same common gate electrode and electrolyte layer. The shared-gate OECT synaptic array uses the signal generated by the multimodal sensing layer as a gate voltage input to the shared gate to regulate the synaptic weight distribution of the array. The shared-gate OECT synaptic array refers to multiple OECT synaptic units sharing a single gate electrode, significantly simplifying array wiring, reducing device costs, and providing a hardware foundation for the collaborative processing of spatiotemporal information. The spatiotemporal encoding and weighting memory is connected to the shared-gate OECT synaptic array. The ECT synaptic array is electrically connected. The spatiotemporal encoder and weight memory generate pulse sequences carrying time information by receiving signals from the shared-gate OECT synaptic array and store the weight information of the synaptic connections. The spatiotemporal encoder simulates the information processing method of brain neurons, encoding information simultaneously in the time and space dimensions. The weight memory is a core component of neuromorphic computing hardware. In the shared-gate OECT synaptic array, the weights are usually achieved by adjusting the channel conductivity of the OECT, and the long-term plasticity of biological synapses is corresponded by non-volatile conductivity modulation. The controller is electrically connected to the spatiotemporal encoder, weight memory, and shared-gate OECT synaptic array. The controller adopts a combination of threshold discrimination and lightweight temporal network to achieve closed-loop control of early slip warning and adaptive re-grabbing. The actuator is electrically connected to the controller. The actuator realizes mechanical movement by receiving signals from the controller. As the output terminal of the system, the actuator converts electrical signals into mechanical movement to realize tactile feedback or active operation.
[0057] The system uses a multimodal sensing layer to sense stimuli received by the electronic skin in a multimodal manner. It combines photoelectrochemically coupled photosynapses with shear vectors and normal pressure to predict critical conditions for slippage, providing diverse data for slippage determination. The multimodal signal is fed into a shared-gate OECT synapse array, where the signal generated by the multimodal sensing layer is used as a gate voltage input to the shared gate, regulating the synaptic weight distribution of the array. Spatiotemporal coding and a weight memory are then used to encode the signal simultaneously in both time and spatial dimensions. Furthermore, the signal is fed into the shared-gate OECT synapse array... By applying a gate voltage to the array, the flow direction of ions in the OECT channel is changed, thereby adjusting the conductivity and redistributing the synaptic weights. The synaptic plasticity is used to quickly enhance and desensitize the device, while lightweight self-training and threshold self-calibration are performed at the controller end to adapt to objects of different materials and weights. At the same time, the object characteristics are fed back to the shared gate OECT synaptic array to adjust the synaptic weights. The control strategy is dynamically generated according to the object characteristics, and the actuator is adjusted to change the gripping force of the actuator on the object, thereby realizing closed-loop anti-slip control of electrochemical synaptic electronic skin.
[0058] The multimodal sensing layer achieves differential readout of shear force through shear electrode sheet 2, detects vibration signals using triboelectric layer 4, and utilizes photosensitive OECT channel 5 to achieve fusion of photosynapses and electrical synapses in the same channel and provide additional optical features under low illumination. Pressure and strain signals are converted and transmitted through ion gel 3. The multimodal sensing layer performs multimodal sensing of stimulation received by electronic skin, combines photoelectrochemically coupled photosynapses with shear vector and normal pressure to predict critical conditions for slip occurrence, provides diverse data for slip occurrence judgment, and improves the accuracy and efficiency of the system in complex scenarios.
[0059] The shared-gate OECT synaptic array adopts a row-shared-gate-column source-drain cross-array structure. Each synaptic unit in the shared-gate OECT synaptic array consists of a source, drain, organic semiconductor channel, electrolyte layer, and shared gate. Multiple synaptic units share a single gate electrode. The gate signal modulates the channel conductance through ion transport in the electrolyte layer. The source and drain are used to read the synaptic weights of the synaptic units. In the shared-gate OECT synaptic array, each row of synaptic units shares a single gate electrode, and each column of synaptic units shares both a source and drain. A positive voltage is applied to the shared gate... Cations in the electrolyte migrate to the organic semiconductor channel, doping the channel material and increasing the channel conductivity. When a negative voltage is applied to the shared gate, the cations migrate back from the channel to the electrolyte, dedoping the channel material and reducing the channel conductivity. This conductivity change is non-volatile and can be stably maintained for several hours to several days. Furthermore, through the enhanced OECT structure and the regulation of ion trapping and detrapping kinetics, near-zero static power consumption is achieved without stimulation, and controllable release is available to accelerate reset. The standby power consumption is near-zero, and the long holding time after triggering and controllable reset meet the requirements of long-term robot tasks.
[0060] Spatiotemporal coding in weighted memory simulates the information processing of neurons in the brain, encoding information simultaneously in both temporal and spatial dimensions. By applying continuous gate voltage pulses, it simulates the short-term plasticity of synapses, achieving dynamic caching and time-series processing of information. Furthermore, by activating different shared gate groups at different times, it achieves spatial distribution modulation of synaptic weights. Combining the temporal pulse sequence with the spatial activation pattern forms an efficient spatiotemporal joint coding system. A lightweight convolutional neural network is used to extract the spatial distribution features of the sensor array, and a gated recurrent unit is used to extract the temporal dependence features of the signal. OECT is then utilized... Synaptic plasticity directly enables neuromorphic encoding of temporal features. Spatiotemporal encoding uses the initial conductance state of the shared gate OECT synaptic array as a reference value. Based on the spatial distribution of signals in the multimodal sensing layer, the conductance of the corresponding OECT device is adjusted. Through global adjustment of the shared gate voltage, the weighted fusion of features is achieved. By mapping programmable weights to the shared gate OECT synaptic array, spatiotemporal superposition and weight plasticity are realized at the device end. Spatiotemporal signals from different modalities are converted into neuromorphic codes that can be processed by the synaptic array, realizing efficient fusion and processing of multimodal information and simulating the closed-loop function of human skin in perception, decision-making, and response.
[0061] The weight memory in the spatiotemporal encoding and weight memory increases conductivity by applying a positive gate voltage and implanting ions into the channel. When a negative gate voltage is applied, ions are extracted from the channel, reducing conductivity. After the gate voltage is removed, the doping state of the channel can be maintained, achieving non-volatile storage of weights. Furthermore, by controlling the amplitude and time of the gate voltage, continuous conductivity modulation is achieved, simulating the multi-mode weights of biological synapses. By adjusting the gate voltage of the OECT device, the device conductivity is modulated, thereby storing weight information. Applying gate voltage pulses of different amplitudes or widths achieves different degrees of conductivity modulation. The spatiotemporal encoding and weight memory stores weights of different modes in different OECT devices. By selecting different devices, different mode weights can be programmed and read, achieving spatial multiplexing. Different mode weights can be programmed and read in different time intervals, achieving temporal multiplexing. By applying different shared gate voltages, different mode weights can be independently programmed and read, achieving voltage multiplexing, completing the mixed programming and reading of multi-mode weights.
[0062] The controller performs lightweight feature extraction and encoding on the parameters of the multimodal perception layer, and fine-tunes the policy network using a lightweight PPO algorithm. It constructs a three-dimensional threshold space for object characteristics based on the multimodal perception layer parameters, and dynamically generates control strategies based on these characteristics to achieve compliant grasping. By loading pre-trained basic control strategies, the controller controls the actuator to contact the object, collects multimodal perception layer data, and fine-tunes the policy network using the lightweight PPO algorithm at the controller end. The optimized strategy is then deployed to the actuator. Calibration data is collected using standard weight blocks and material samples. A three-dimensional threshold model of weight, material, and contact area is established using Gaussian process regression. A default threshold range for stimulation of the electronic skin is set, and the actuator is controlled to contact the object with a small force. Initial feedback is collected, and the object category is determined based on the feedback characteristics. The corresponding control strategy is matched from the policy library, and the control parameters are optimized based on real-time feedback.
[0063] Principle of this invention:
[0064] This invention utilizes a multimodal sensing layer to perceive stimuli received by electronic skin. It combines photoelectrochemically coupled photosynapses with shear vectors and normal pressure to predict critical conditions in the early stages of slippage, providing diverse data for slippage detection. Based on the multimodal signals, a gate voltage is applied to the shared-gate OECT synaptic array, altering the ion flow direction in the OECT channel and regulating conductivity. This redistributes synaptic weights, while simultaneously feeding back object characteristics to the shared-gate OECT synaptic array for further synaptic weight adjustment. A control strategy is dynamically generated based on the object's characteristics. The system provides feedback adjustments to the actuators, altering their gripping force on objects to achieve closed-loop anti-slip control of the electrochemical synaptic electronic skin. Based on early critical conditions and synaptic weight updates in the multimodal phase, the actuators are adjusted in real time to improve the accuracy and efficiency of the system's judgment in complex scenarios. Furthermore, through an enhanced OECT structure and ion trapping and detrapping dynamics regulation, near-zero static power consumption is achieved without stimulation, along with controllable release to accelerate reset. Standby power consumption is near-zero, and the system maintains a long hold time and controllable reset after triggering, meeting the robot's long-duration task requirements.
[0065] 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 multimodal adaptive electrochemical synaptic electronic skin, characterized in that, The invention includes a multimodal electrochemical electronic skin, which is composed of a sensing unit, a signal processing unit, and a basal layer unit. The sensing unit is used to acquire stimulation signals, the signal processing unit is used to convert and transmit the acquired signals, and the basal layer unit is used to integrate the sensing unit and the signal processing unit to form an electronic skin. The sensing unit includes a shear electrode sheet (2), an ion gel (3), a triboelectric layer (4), and a photosensitive OECT channel (5). The shear electrode sheet (2), the ion gel (3), the triboelectric layer (4), and the photosensitive OECT channel (5) are all disposed on the base layer unit. The shear electrode sheet (2), the triboelectric layer (4), and the photosensitive OECT channel (5) are stacked in the ion gel (3). The triboelectric layer (4) is disposed below the shear electrode sheet (2), and the photosensitive OECT channel (5) is disposed below the triboelectric layer (4). The signal processing unit includes a microdome PDMS array (1) and an electrochemical synaptic transistor (6). The microdome PDMS array (1) is disposed above the shear electrode sheet (2). The microdome PDMS array (1) forms multiple regularly arranged microdome-shaped protrusions on the ion gel (3) through microfabrication technology. The electrochemical synaptic transistor (6) is disposed in the photosensitive OECT channel (5). The upper and lower electrodes of the shear electrode sheet (2) are arranged in an interdigital shape. The ion gel (3) is used as the electrolyte of the electrochemical synaptic transistor (6). The ion gel (3) is used to sense the pressure and strain of the multimodal electrochemical electronic skin and convert it into an electrical signal. The triboelectric layer (4) uses a nanofiber membrane to detect dynamic stimulation based on the coupling effect of contact electrification and electrostatic induction. The photosensitive OECT channel (5) is used to fuse photoelectrochemically coupled photosynapses and tactile shear-triggered electrical synapses in the same channel.
2. The multimodal adaptive electrochemical synaptic electronic skin according to claim 1, characterized in that, The base layer unit includes an insulating film (7) and a flexible substrate (8). The insulating film (7) is disposed below the photosensitive OECT channel (5), and the flexible substrate (8) is disposed below the insulating film (7). The flexible substrate (8) is used to support the sensing unit and the signal processing unit.
3. A closed-loop anti-slip control system for multimodal adaptive electrochemical synaptic electronic skin, characterized in that, Using the multimodal adaptive electrochemical synaptic electronic skin according to any one of claims 1-2, comprising: A multimodal sensing layer, which captures stimulus signals by contacting the external environment and converts them into electrical signals; A shared gate OECT synaptic array is electrically connected to a multimodal sensing layer. The shared gate OECT synaptic array is used to share the same common gate electrode and electrolyte layer among multiple OECT channels. The shared gate OECT synaptic array controls the synaptic weight distribution of the array by inputting the signal generated by the multimodal sensing layer as a gate voltage to the shared gate. A spatiotemporal encoding and weighting memory is electrically connected to a shared gate OECT synaptic array. The spatiotemporal encoding and weighting memory generates a pulse sequence carrying time information by receiving signals from the shared gate OECT synaptic array and stores the weight information of synaptic connections. The controller is electrically connected to a spatiotemporal coding and weight memory and a shared gate OECT synaptic array. The controller adopts a combination of threshold discrimination and lightweight timing network to realize closed-loop control of early warning of slip and adaptive re-grabbing. An actuator, which is electrically connected to a controller, performs mechanical motion by receiving signals from the controller.
4. The closed-loop anti-slip control system for a multimodal adaptive electrochemical synaptic electronic skin according to claim 3, characterized in that, The multimodal sensing layer achieves differential readout of shear force through shear electrode sheet (2), detects vibration signal using triboelectric layer (4), achieves fusion of photosynapses and electrical synapses in the same channel using photosensitive OECT channel (5) and provides additional optical features under low illumination, and converts and transmits pressure and strain signals through ion gel (3).
5. The closed-loop anti-slip control system for a multimodal adaptive electrochemical synaptic electronic skin according to claim 4, characterized in that, The shared-gate OECT synaptic array adopts a cross-array structure of row shared gate and column source-drain. The synaptic unit in the shared-gate OECT synaptic array consists of a source, a drain, an organic semiconductor channel, an electrolyte layer, and a shared gate. Multiple synaptic units share a single gate electrode. The gate signal modulates the channel conductance through ion transport in the electrolyte layer. The source and drain are used to read the synaptic weight of the synaptic unit. Through the OECT structure and ion dynamics regulation, near-zero static power consumption is achieved in the unstimulated state.
6. The closed-loop anti-slip control system for a multimodal adaptive electrochemical synaptic electronic skin according to claim 5, characterized in that, The spatiotemporal coding and weight memory simulate the information processing of brain neurons, encoding information simultaneously in both time and space dimensions. By applying continuous gate voltage pulses, the short-term plasticity of synapses is simulated, enabling dynamic caching and time-series processing of information. Furthermore, by activating different shared gate groups at different times, spatial distribution modulation of synaptic weights is achieved. Finally, the time pulse sequence is combined with the spatial activation mode to form an efficient spatiotemporal joint coding.
7. The closed-loop anti-slip control system for a multimodal adaptive electrochemical synaptic electronic skin according to claim 6, characterized in that, The weight memory in the spatiotemporal encoding and weight memory adjusts the conductivity by applying positive and negative gate voltages to change the flow direction of ions in the OECT channel. After the gate voltage is removed, the doping state of the channel can be maintained, realizing non-volatile storage of weights. Furthermore, by controlling the amplitude and time of the gate voltage, continuous conductivity modulation is achieved, simulating the polymorphic weights of biological synapses.
8. The closed-loop anti-slip control system for a multimodal adaptive electrochemical synaptic electronic skin according to claim 7, characterized in that, The controller performs lightweight feature extraction and feature encoding on the parameters of the multimodal perception layer, and fine-tunes the policy network using a lightweight PPO algorithm. It constructs a three-dimensional threshold space of object characteristics based on the parameters of the multimodal perception layer, and dynamically generates control strategies based on the object characteristics to achieve compliant grasping.