Exoskeleton knee joint compliance control system and method based on piezoresistive sensing array

Abstract

The invention discloses an exoskeleton knee joint compliance control system based on a piezoresistive sensing array, and relates to the technical field of rehabilitation engineering and robot control. In the system, a sensing unit group consists of a flexible silica gel layer and a piezoresistive sensing array at a knee joint, a knee joint encoder, a plantar pressure sensor and an upper limb attitude sensor, and is used for collecting multi-source physical signals; the data processing and control module is used for realizing high-precision synchronous acquisition of multi-source data and hardware-level pressure feature extraction based on an FPGA (Field Programmable Gate Array), carrying out parallel decision making through a gait phase recognition unit arranged in a main controller, a fuzzy controller based on a muscle force model and a dynamic model and a PID (Proportion Integration Differentiation) corrector, and outputting an auxiliary torque; and finally, the torque synthesis module carries out safe amplitude limiting on the output torque in combination with the real-time pressure characteristics, and a driver drives a knee joint motor to execute. By means of multi-modal information fusion and prospective safety protection, compliance, self-adaption and safety control over the knee joint of the exoskeleton or the artificial limb is achieved.

Description

Technical Field

[0001] This invention relates to the fields of rehabilitation engineering and robot control technology, and in particular to an exoskeleton knee joint compliance control system based on a piezoresistive sensor array; and an exoskeleton knee joint compliance control method based on a piezoresistive sensor array. Background Technology

[0002] With the aging population and the increase in accidental injuries, the number of patients with lower limb motor dysfunction caused by stroke, spinal cord injury, or knee joint disease is rising year by year. As a key joint for weight-bearing and movement, the knee joint requires rehabilitation assistive devices that are highly intelligent and user-friendly.

[0003] Traditional knee rehabilitation or assistive devices, such as simple hinged braces or exoskeletons (robots) with fixed trajectory control, lack the ability to perceive changes in the user's intentions and state, and cannot provide adaptive, compliant assistive forces. This may lead to unnatural movements, muscle compensation, or even secondary injuries. Some more advanced systems use angle or torque sensors for control, but their sensing information sources are singular, lacking comprehensive perception and utilization of multi-dimensional information such as the interaction pressure between the knee joint and the device, and the coordinated movement of the whole body, resulting in poor adaptability to complex gaits (such as going up and down stairs, sitting up, and turning).

[0004] The main shortcomings of existing technologies are: 1. Limited Perception Dimension: Most systems rely solely on joint encoders or IMUs for control, which cannot accurately perceive the contact status of the human-machine interface. This can easily lead to localized pressure concentration, causing discomfort or injury.

[0005] 2. Rigid decision-making logic: Control strategies are mostly based on preset trajectories or fixed PID parameters, which are difficult to adapt to individual differences and dynamic changes in gait, and lack "compliance".

[0006] 3. Delayed safety mechanisms: Safety protection is usually triggered after kinematic thresholds (such as angle limits), which is a relatively delayed protection and cannot proactively suppress abnormal human-computer interaction forces at the moment they occur.

[0007] 4. Poor data synchronization: Multi-sensor data is synchronized at the software level, which has problems such as inconsistent timestamps and large processing delays, affecting the real-time performance and accuracy of closed-loop control.

[0008] Therefore, there is an urgent need for an exoskeleton compliance control system that can perceive movement intentions and human-computer interaction status in multiple dimensions and make intelligent, real-time, and safe decisions. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of the prior art and provide an exoskeleton knee joint compliance control system and method based on a piezoresistive sensor array. This system achieves accurate and real-time perception of the human-computer interaction state through multi-source heterogeneous sensor fusion and hardware-level synchronous data processing. Based on gait phase and pressure characteristics, it outputs an auxiliary torque that conforms to biomechanical laws and ensures user comfort and safety through parallel decision-making and safety limiting by multiple controllers.

[0010] To achieve the above objectives, the present invention provides an exoskeleton knee joint compliance control system based on a piezoresistive sensor array, comprising: The sensor unit group is used to collect multi-source physical signals related to the user's knee joint movement; The data processing and control module is used to receive and process the multi-source physical signals and generate control commands; The sensing unit group includes: A flexible silicone layer and a piezoresistive sensor array are provided. The flexible silicone layer is attached to the front and back sides of the knee joint and has an array of thin-film piezoresistive sensors embedded inside it to detect the contact pressure of the knee joint. A knee joint encoder is installed on the shaft end of the DC servo motor of the knee joint to detect the angle and angular velocity of the knee joint. The plantar pressure sensor is installed on the forefoot, arch, and heel of the foot to detect the distribution of plantar pressure. An upper limb posture sensor, fixed to the forearm, is used to detect the acceleration and angular velocity of the upper limb. The data processing and control module includes: A multi-source data acquisition and synchronization module, built on FPGA, is connected to the sensing unit group and is used to synchronously acquire the multi-source physical signals and add a unified timestamp to all data. The pressure feature extraction module, integrated within the FPGA, is used to receive data from the piezoresistive sensing array and calculate pressure feature quantities in real time. The main controller is communicatively connected to the multi-source data acquisition and synchronization module, and integrates a gait phase recognition unit, at least two different types of controllers, and a torque synthesis module. The knee joint motor driver is communicatively connected to the main controller and is used to drive the knee joint DC servo motor according to torque commands.

[0011] Furthermore, the thickness of the flexible silicone layer is preferably 10 mm, and the Shore hardness is 28°; the piezoresistive sensing array is preferably a 4×8 channel thin-film piezoresistive sensor with an array spacing of 12 mm and a coverage area of ​​120 mm × 96 mm.

[0012] Furthermore, the preferred selection of the multi-source data acquisition and synchronization module includes: An analog signal acquisition unit is connected to the output terminal of the piezoresistive sensing array and the plantar pressure sensor. An orthogonal encoding counting unit is connected to the signal output terminal of the knee joint encoder; The I2C data reading unit is connected to the upper limb posture sensor; The timestamp addition unit is used to add a uniform timestamp to all collected data.

[0013] Furthermore, the pressure feature quantities calculated by the pressure feature extraction module preferably include: maximum pressure value, pressure center coordinates, pressure center offset, and pressure change trend slope.

[0014] Furthermore, the preferred options for at least two different types of controllers within the main controller include: a fuzzy controller based on a muscle force model, a fuzzy controller based on dynamics, and a PID attitude deviation correction unit.

[0015] Furthermore, the torque synthesis module is preferably used to perform vector superposition of the torque signals output by the fuzzy controller based on the muscle force model, the fuzzy controller based on dynamics, and the PID attitude deviation correction unit, and to perform safety limiting processing on the synthesized torque based on the real-time pressure features fed back by the pressure feature extraction module.

[0016] Furthermore, the preferred options for the safety limiting processing include: When a rapid increase in pressure is detected, the total output torque is instantaneously reduced; When an abnormal deviation in pressure distribution or the appearance of localized pressure hotspots is detected, the output torque is forcibly limited.

[0017] In addition, the exoskeleton knee joint compliance control method based on a piezoresistive sensor array designed in this invention includes the following steps: Step 1: Multi-source data acquisition and synchronization: Multi-source physical signals from the sensor unit group are acquired synchronously through FPGA, and a unified timestamp is added to all data; Step 2, Pressure Feature Extraction and Gait Phase Recognition: The FPGA hardware feature extraction unit extracts pressure features from the piezoresistive array data in real time; at the same time, the main controller identifies the current gait phase based on the knee joint angle and plantar pressure data. Step 3, Parallel Decision-Making by Multiple Controllers: Based on the identified gait phase, the main controller drives its internal multiple controllers to work in parallel, calculating the auxiliary torque, compensation torque, and fine-tuning torque respectively; Step 4, Torque Synthesis and Safety Limiting: The torques output from the multi-channel controller are synthesized, and the synthesized torques are subjected to safety limiting processing based on real-time pressure characteristics to generate the final torque command; Step 5, Drive Execution and Closed-Loop Operation: The final torque command is sent to the knee joint motor driver to drive the DC servo motor to output torque. The system executes steps one to five in a loop at a set frequency to form real-time closed-loop control.

[0018] Furthermore, in step three, the preferred selection of the multiplexer includes: A fuzzy controller based on a muscle force model calculates auxiliary torque based on gait phase and pressure distribution; A dynamic fuzzy controller calculates compensation torque based on gait phase, knee joint kinematics, and upper limb posture. The PID attitude deviation correction unit outputs a fine-tuning torque for trajectory tracking.

[0019] Furthermore, in step four, the preferred options for the safety limiting processing include: If the pressure change trend indicates a rapid increase in pressure, the total output torque will be reduced proportionally. If the pressure center offset exceeds the threshold or a local pressure hotspot appears, the output torque will be limited to a preset safe range.

[0020] Compared with the prior art, the present invention has the following significant advantages: 1. Multimodal perception and deep fusion: By integrating four types of information—joint motion, human-computer interaction pressure, foot contact, and upper limb posture—a comprehensive environmental state perception capability is constructed, providing a rich data foundation for intelligent control.

[0021] 2. Hardware-level synchronization and real-time processing: Based on FPGA, synchronous acquisition and hardware feature extraction fundamentally solve the problem of multi-source data synchronization, and bring computationally intensive tasks forward, greatly reducing the load on the main controller and the overall system latency, and ensuring the real-time performance of closed-loop control.

[0022] 3. Bionic and Adaptive Decision Making: It combines bionic control based on muscle models and high-level dynamic compensation to make the auxiliary torque more in line with the natural movement law of the human body, while adapting to individual differences and dynamic changes of users through fuzzy logic.

[0023] 4. Proactive Safety Protection Mechanism: By using pressure characteristics as the direct basis for safety limits, it achieves a leap from "kinematic protection" to "interactive force protection." It can proactively reduce torque before discomfort or danger occurs (such as when pressure increases sharply), achieving proactive and flexible safety protection, and significantly improving comfort and safety.

[0024] 5. Modularity and scalability: The system architecture is clear, and the modules communicate with each other through standard interfaces, which facilitates the expansion and upgrading of functions (for example, new sensors or control algorithms can be easily added). Attached Figure Description

[0025] Figure 1 This is a control block diagram of an exoskeleton knee joint compliance control system based on a piezoresistive sensor array. Figure 2 This is a flowchart of an exoskeleton knee joint compliance control system based on a piezoresistive sensor array. Detailed Implementation

[0026] 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. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.

[0027] As one embodiment of the present invention, this embodiment provides an exoskeleton knee joint compliance control system based on a piezoresistive sensor array and Figure 1 The corresponding control block diagram shows the structure and connection relationships of each module as follows: The sensor unit group is used to collect multi-source physical signals related to the user's knee joint movement; The data processing and control module is used to receive and process the multi-source physical signals and generate control commands; The sensing unit group includes: Flexible silicone layer + piezoresistive sensor array: A flexible silicone buffer layer with a thickness of 10mm and a Shore hardness of 28° is attached to the front and back sides of the knee joint. 4×8 thin-film piezoresistive sensors (model FSR406) are embedded inside, with an array spacing of 12mm and a coverage area of ​​120mm×96mm. The sensor outputs are collected through distributed leads and sent to the analog input interface of the multi-source data acquisition and synchronization module. Knee joint encoder: It adopts a 2048-line incremental photoelectric encoder, which is integrated into the shaft end of the DC servo motor of the knee joint. The encoder signal output end is connected to the orthogonal encoding interface of the multi-source data acquisition and synchronization module. Foot pressure sensor: One thin-film piezoresistive sensor is installed in the forefoot, arch, and heel. The sensor output is connected to the analog input interface of the multi-source data acquisition and synchronization module. Upper limb posture sensor: A six-axis IMU (model MPU6050) is fixed to the user's upper limb forearm and connected to the multi-source data acquisition and synchronization module via an I2C interface; The data processing and control module includes: Multi-source data acquisition and synchronization module: Built on FPGA (XC7A35T), it completes analog signal acquisition, orthogonal encoding counting, I2C data reading, and adds a unified timestamp to all data. After acquisition, the FPGA's hardware feature extraction unit calculates and outputs the maximum pressure value (pressure peak), pressure center coordinates, pressure center offset, and pressure change trend slope in real time. Pressure feature extraction module: This is a hardware feature extraction unit built into the FPGA. It receives piezoresistive array data output from the multi-source data acquisition and synchronization module and calculates the maximum pressure value, pressure center coordinates, pressure center offset, and pressure change trend slope in real time. Main controller: It adopts an STM32H750 microcontroller (480MHz main frequency), receives synchronization data through the SPI interface, and has built-in fuzzy controller based on muscle force model, fuzzy controller based on dynamics and PID attitude deviation correction unit.

[0028] Torque synthesis module: Integrated inside the main controller, used to superimpose torque signals output from the three controllers; Knee joint motor driver: The DS402 DC servo driver is used. It receives torque commands from the main controller via the CAN bus and drives the knee joint DC servo motor (model AKM2G-064) to output torque.

[0029] This embodiment provides a method for knee joint compliance control of an exoskeleton based on a piezoresistive sensor array, which includes the following steps: Step 1: Multi-source data acquisition and synchronization After the system powers on, its data flow originates from the multi-source data acquisition and synchronization module. This module, based on FPGA hardware, synchronously acquires physical signals from four independent sensor unit groups with high precision and speed. Specifically, it reads 32 analog voltage signals from the flexible silicone layer and piezoresistive sensor array, reflecting real-time pressure changes at various contact points of the knee joint. Simultaneously, it captures pulse signals output from the knee joint encoder via an orthogonal encoding interface to analyze the precise angle and angular velocity of the knee joint. Furthermore, it collects three pressure readings from the plantar pressure sensor and acceleration and angular velocity data from the upper limb posture sensor (IMU) via the I2C bus. The core function of this module is to assign a unified timestamp to all these heterogeneous, time-discrete sensor data, eliminating calculation errors caused by asynchronous sampling times and providing a strictly synchronized and complete environmental state dataset for subsequent control decisions.

[0030] Step 2: Pressure Feature Extraction and Gait Phase Recognition The synchronized data stream is split into two paths for preprocessing and feature extraction. In the first path, the raw piezoresistive array data is fed into a pressure feature extraction module. This hardware feature extraction unit calculates key contact features in real time, such as the maximum pressure value, the coordinates and offset of the pressure center, and the slope of the pressure change trend. In the second path, knee joint angle and plantar pressure data are fed into the gait phase recognition unit built into the main controller. This unit uses a rule-based state machine to analyze these kinematic signals in real time, accurately determining the user's current gait phase (e.g., early stance phase, late stance phase, early swing phase). At this point, the system obtains all the high-level information needed for intelligent decision-making: namely, "which stage of the gait is currently in" and "what is the pressure state at the knee joint."

[0031] Step 3: Parallel decision-making by multiple controllers After obtaining clear gait phase and stress characteristic information, the system enters the core decision-making stage. The main controller, based on the global context provided by the gait phase recognition unit, activates the corresponding control logic and drives its three internal controllers to work in parallel. The fuzzy controller (Tp) based on the muscle force model calculates an auxiliary torque conforming to human biomechanics principles, aiming to reduce muscle load, based on gait phase and stress distribution. The fuzzy controller (Tm) based on dynamics integrates the current gait phase, knee joint kinematics, and upper limb posture to calculate a compensating torque to overcome gravity and inertia and ensure smooth movement. The PID posture deviation correction unit (Ts) is responsible for more refined trajectory tracking adjustments, outputting a fine-tuning torque to eliminate errors. The outputs of these three controllers together constitute the auxiliary force that the system should provide under ideal conditions.

[0032] Step 4: Torque Synthesis and Safety Limitation The ideal torque output from the first three controllers is fed into the torque synthesis module for vector superposition. However, before the final output, the synthesized torque must undergo "filtering" based on safety and comfort principles. The torque synthesis module forcibly incorporates real-time feedback from the pressure feature extraction module: if the system detects a "rapid increase" in pressure, the total output torque is instantaneously reduced (e.g., reduced by 30%); if an abnormal shift in pressure distribution or the appearance of localized pressure hotspots is detected, the output torque is forcibly limited (e.g., limited to 50% of the baseline value). This process ensures that the final generated composite torque command not only meets the functional requirements of gait assistance but also strictly guarantees the user's comfort and safety.

[0033] Step 5: Drive Execution and Closed-Loop Operation The final torque command, after synthesis and safety verification, is sent to the knee joint motor driver via the CAN bus. The driver converts this digital command into a high-precision current signal, driving the knee joint DC servo motor to output precise physical torque. This torque acts on the human body through the mechanical structure of the exoskeleton, completing the "execution" phase of one control cycle. The system continuously runs steps one through five at a frequency of 100Hz, forming a real-time closed-loop control of perception, decision-making, execution, and re-perception. This dynamically adapts to the user's movement intentions and state changes, achieving truly compliant and safe rehabilitation assistance.

[0034] When using the exoskeleton (robot) combining the above-described exoskeleton compliance control system and method, a flexible silicone sleeve with a piezoresistive sensor array is wrapped around the user's knee joint, foot pressure sensors are embedded in the insoles, the upper limb IMU is attached to the forearm, and the entire exoskeleton structure is worn on the lower limb. After the system is powered on, refer to... Figure 2 The process runs. Below are two common use cases: Scenario 1: The FPGA begins synchronous data acquisition. The pressure feature extraction module outputs the pressure center coordinates in real time. Assuming the pressure center should normally be located in the center of the array during the mid-walking support phase, if the system detects a continuous backward shift of the pressure center (towards the popliteal fossa) and the maximum pressure value exceeds the limit, the gait phase recognition unit determines that the user may be in a backward leaning posture. At this time, the dynamics-based fuzzy controller calculates an additional backward compensation torque to prevent uncontrolled knee flexion. However, after torque synthesis, the safety limiting unit, due to the characteristics of "local pressure hotspots" and "abnormal pressure center shift," forcibly limits the final output torque to a lower level. This provides necessary support while avoiding severe compression discomfort in the user's popliteal fossa due to over-assistance.

[0035] Scenario 2: In the early phase of the swing phase, the user needs to bend their knee to cross an obstacle. If the lower leg suddenly hits an obstacle (such as a low stool) at this time, the sensor in front of the piezoresistive sensor array will detect a rapid increase in pressure within a very short time (the slope of the pressure change trend is extremely high). This information is quickly extracted by the FPGA and transmitted to the safety limiting unit of the main controller. The unit immediately acts, significantly reducing the output knee flexion torque (e.g., reducing it by 50%), thereby instantly weakening the output torque of the knee joint servo motor, avoiding a hard collision with the obstacle, and protecting the user and the equipment. This response speed is unattainable by traditional angle-over-limit-based protection mechanisms.

[0036] This invention is not limited to the preferred embodiments described above. Anyone can derive other products in various forms under the guidance of this invention. However, regardless of any changes in shape or structure, any technical solution that is the same as or similar to this application falls within the protection scope of this invention.

Claims

1. A knee joint compliance control system for an exoskeleton based on a piezoresistive sensor array, characterized in that, include: The sensor unit group is used to collect multi-source physical signals related to the user's knee joint movement; The data processing and control module is used to receive and process the multi-source physical signals and generate control commands; The sensing unit group includes: A flexible silicone layer and a piezoresistive sensor array are provided. The flexible silicone layer is attached to the front and back sides of the knee joint and has an array of thin-film piezoresistive sensors embedded inside it to detect the contact pressure of the knee joint. A knee joint encoder is installed on the shaft end of the DC servo motor of the knee joint to detect the angle and angular velocity of the knee joint. The plantar pressure sensor is installed on the forefoot, arch, and heel of the foot to detect the distribution of plantar pressure. An upper limb posture sensor, fixed to the forearm, is used to detect the acceleration and angular velocity of the upper limb. The data processing and control module includes: A multi-source data acquisition and synchronization module, built on FPGA, is connected to the sensing unit group and is used to synchronously acquire the multi-source physical signals and add a unified timestamp to all data. The pressure feature extraction module, integrated within the FPGA, is used to receive data from the piezoresistive sensing array and calculate pressure feature quantities in real time. The main controller is communicatively connected to the multi-source data acquisition and synchronization module, and integrates a gait phase recognition unit, at least two different types of controllers, and a torque synthesis module. The knee joint motor driver is communicatively connected to the main controller and is used to drive the knee joint DC servo motor according to torque commands.

2. The exoskeleton compliance control system according to claim 1, characterized in that, The flexible silicone layer has a thickness of 10mm and a Shore hardness of 28°; the piezoresistive sensing array is a 4×8 channel thin-film piezoresistive sensor with an array spacing of 12mm and a coverage area of ​​120mm×96mm.

3. The exoskeleton compliance control system according to claim 1, characterized in that, The multi-source data acquisition and synchronization module includes: An analog signal acquisition unit is connected to the output terminal of the piezoresistive sensing array and the plantar pressure sensor. An orthogonal encoding counting unit is connected to the signal output terminal of the knee joint encoder; The I2C data reading unit is connected to the upper limb posture sensor; The timestamp addition unit is used to add a uniform timestamp to all collected data.

4. The exoskeleton compliance control system according to claim 1, characterized in that, The pressure feature extraction module calculates the following pressure features: maximum pressure value, pressure center coordinates, pressure center offset, and pressure change trend slope.

5. The exoskeleton compliance control system according to claim 1, characterized in that, The main controller includes at least two different types of controllers: a fuzzy controller based on a muscle force model, a fuzzy controller based on dynamics, and a PID attitude deviation correction unit.

6. The exoskeleton compliance control system according to claim 5, characterized in that, The torque synthesis module is used to perform vector superposition of the torque signals output by the fuzzy controller based on the muscle force model, the fuzzy controller based on dynamics, and the PID attitude deviation correction unit, and to perform safety limiting processing on the synthesized torque based on the real-time pressure features fed back by the pressure feature extraction module.

7. The exoskeleton compliance control system according to claim 6, characterized in that, The safety limiting process includes: When a rapid increase in pressure is detected, the total output torque is instantaneously reduced; When an abnormal deviation in pressure distribution or the appearance of localized pressure hotspots is detected, the output torque is forcibly limited.

8. A method for knee joint compliance control in an exoskeleton based on a piezoresistive sensor array, characterized in that, The system applied to any one of claims 1-7 includes the following steps: Step 1: Multi-source data acquisition and synchronization: Multi-source physical signals from the sensor unit group are acquired synchronously through FPGA, and a unified timestamp is added to all data; Step 2, Pressure Feature Extraction and Gait Phase Recognition: The FPGA hardware feature extraction unit extracts pressure features from the piezoresistive array data in real time; at the same time, the main controller identifies the current gait phase based on the knee joint angle and plantar pressure data. Step 3, Parallel Decision-Making by Multiple Controllers: Based on the identified gait phase, the main controller drives its internal multiple controllers to work in parallel, calculating the auxiliary torque, compensation torque, and fine-tuning torque respectively; Step 4, Torque Synthesis and Safety Limiting: The torques output from the multi-channel controller are synthesized, and the synthesized torques are subjected to safety limiting processing based on real-time pressure characteristics to generate the final torque command; Step 5, Drive Execution and Closed-Loop Operation: The final torque command is sent to the knee joint motor driver to drive the DC servo motor to output torque. The system executes steps one to five in a loop at a set frequency to form real-time closed-loop control.

9. The exoskeleton compliance control method according to claim 8, characterized in that, In step three, the multiplexer includes: A fuzzy controller based on a muscle force model calculates auxiliary torque based on gait phase and pressure distribution; A dynamic fuzzy controller calculates compensation torque based on gait phase, knee joint kinematics, and upper limb posture. The PID attitude deviation correction unit outputs a fine-tuning torque for trajectory tracking.

10. The exoskeleton compliance control method according to claim 8, characterized in that, In step four, the safety limiting process includes: If the pressure change trend indicates a rapid increase in pressure, the total output torque will be reduced proportionally. If the pressure center offset exceeds the threshold or a local pressure hotspot appears, the output torque will be limited to a preset safe range.

Images