A multi-source fusion perception system for an exoskeleton robot

Abstract

The application discloses a kind of multi-source fusion perception systems for exoskeleton robot, belong to robot technical field, system includes:foot bottom pressure acquisition module, is laid in exoskeleton foot component;Inertial measurement module, is laid in exoskeleton limb bar piece;Motor feedback module, is integrated in exoskeleton driving joint;Central processing unit is connected with each module, for receiving and processing multi-source data, method includes: by foot bottom pressure dynamic threshold algorithm real-time determination gait switching point;Fusion inertial measurement data and motor feedback data, carry out limb posture solution and man-machine interaction moment estimation;Based on the redundancy check of multi-source data realizes fault monitoring and safety control.The application solves the problems of high gait recognition delay, poor environmental adaptability and insufficient man-machine interaction transparency in the prior art through deep coupling and synchronous perception of mechanical and kinematic data, and realizes accurate, robust and real-time perception of the wearer's movement intention.

Description

Technical Field

[0001] This invention relates to the field of robotics, and more particularly to a multi-source fusion sensing system for exoskeleton robots. Background Technology

[0002] With the deep integration of rehabilitation medicine and robotics, lower limb exoskeleton robots have become an important tool for assisting patients with walking dysfunction (such as stroke patients and spinal cord injury patients) in gait rehabilitation training or daily walking. The core of achieving safe, natural, and efficient human-machine collaboration lies in accurately and in real time sensing and understanding the wearer's movement intentions.

[0003] Existing perception solutions for exoskeleton robots have the following limitations: 1) Single perception dimension: For example, relying solely on the motor encoder cannot know the true force state and spatial posture of the human body, resulting in inaccurate timing of assistance and easy generation of "parasitic torque" in human-machine confrontation; relying solely on the inertial measurement unit (IMU) for kinematic integral calculation has a lag and is difficult to cope with high-frequency gait switching.

[0004] 2) Poor environmental adaptability: Foot pressure triggering schemes based on fixed thresholds are difficult to adapt to pressure changes under different terrains (such as slopes and soft ground), resulting in low robustness.

[0005] 3) High system complexity and cost: Some solutions introduce bioelectrical signals such as surface electromyography (sEMG), but sEMG signals are susceptible to interference, have large individual differences, and need to come into contact with the skin, so their reliability needs to be improved. Other solutions employ complex deep learning models, which require high computing power from embedded hardware, hindering system miniaturization, low power consumption, and cost control.

[0006] 4) Lack of real-time interactive force perception: Many solutions fail to make full use of the high-frequency feedback signal of the current loop at the bottom of the motor, resulting in insufficient high-frequency human-machine interactive force perception and compensation at the microsecond level, which affects the transparency of "zero resistance" follow-up.

[0007] Therefore, this invention proposes a multi-source fusion sensing system for exoskeleton robots. Summary of the Invention

[0008] The purpose of this invention is to address the shortcomings of existing technologies by proposing a multi-source fusion sensing system for exoskeleton robots.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: In view of this, the purpose of the present invention is to provide a multi-source fusion perception system and method for exoskeleton robots, which aims to solve the problems of gait recognition delay, poor environmental adaptability and low human-computer interaction transparency caused by single perception dimension and insufficient data fusion in the prior art.

[0010] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a multi-source fusion sensing system for an exoskeleton robot, comprising: The plantar pressure acquisition module is installed on the foot component of the exoskeleton robot to collect real-time pressure distribution data of the wearer's feet. An inertial measurement module is installed on the limb links (such as thigh links and calf links) of the exoskeleton robot to measure the motion parameters such as angular velocity and acceleration of the limb links in real time, and to calculate the limb's attitude (heading, pitch, roll) data in space. The motor feedback module is integrated inside the drive joints (such as the hip joint and knee joint) of the exoskeleton robot to acquire the angle data (via encoder) and current data of the drive joint in real time. The central processing unit is communicatively connected to the plantar pressure acquisition module, the inertial measurement module, and the motor feedback module, respectively, and is used to receive and synchronously process data from the above multiple sources; The central processing unit is configured to perform the following core functions: Real-time gait event detection: Based on the plantar pressure distribution data, a dynamic threshold algorithm (non-fixed threshold) is used to determine key discrete switching points in the gait cycle (such as heel strike and toe lift) in real time. By analyzing dynamic features such as pressure jump slope, this algorithm can overcome the sensitivity of fixed thresholds to terrain and individual differences, achieving millisecond-level response.

[0011] Motion state fusion perception: This integrates the absolute limb posture data calculated by the inertial measurement module, the relative joint angle data acquired by the motor feedback module, and the human-computer interaction torque data calculated in real time from the current data combined with the motor model (such as torque constant). Through deep coupling of mechanics (pressure, torque) and kinematics (posture, angle), it achieves synchronous perception and prediction of the wearer's movement intentions, rather than triggering them afterward.

[0012] Redundancy verification and safety monitoring: Redundancy verification is performed by continuously comparing the absolute attitude data calculated by the inertial measurement module with the relative joint angle data obtained by the motor feedback module (both of which should have a definite kinematic relationship in physical terms). When the deviation between the two exceeds a preset safety threshold, it can be determined that the sensor is being interfered with or has experienced a mechanical failure, and a safety protection mechanism (such as switching to damped state) is immediately triggered, significantly improving system safety.

[0013] Preferably, the pressure sensors of the plantar pressure acquisition module adopt a triangular positioning layout, corresponding to the three key biomechanical bearing points of the heel, the first metatarsal head, and the fifth metatarsal head, respectively, in order to effectively capture the trajectory changes of the ground reaction force center during gait.

[0014] Preferably, the central processing unit is also configured to: when the wearer is determined to have entered the swing phase based on plantar pressure, control the exoskeleton to switch to a "zero-resistance" follow-up mode (such as impedance control). In this mode, the system combines the limb movement trend captured by the inertial measurement module with the angle change rate fed back by the motor to dynamically adjust the motor output torque, so as to actively counteract the inertia, gravity and frictional resistance of the exoskeleton mechanism itself, so as to minimize the resistance felt by the wearer.

[0015] Secondly, the present invention provides a multi-source fusion sensing method applicable to the above-mentioned system.

[0016] Thirdly, the present invention provides an exoskeleton robot incorporating the above-mentioned multi-source fusion sensing system.

[0017] The beneficial effects of this invention are as follows: High real-time performance and low latency: Gait events are directly detected by the dynamic threshold of plantar pressure, skipping the time accumulation delay caused by the traditional IMU integral calculation of displacement, and realizing a fast response to gait switching points.

[0018] High environmental adaptability and robustness: The dynamic threshold algorithm, fusion of multi-source data (pressure, posture, torque) and redundancy verification enable the system to effectively adapt to changes in different wearers, weights and terrains (such as flat ground and slopes), significantly improving the accuracy of gait recognition.

[0019] High transparency in human-computer interaction: Deeply integrated with the current feedback at the motor's underlying layer, it estimates the human-computer interaction torque in real time and combines it with kinematic data to achieve smooth control through a hybrid force-position mechanism. Especially in the oscillation phase, it can achieve highly transparent "zero-resistance" follow-up, greatly improving the comfort and naturalness of wearing the device.

[0020] The system boasts high safety: By redundantly comparing inertial attitude and joint angles, an inherent fault diagnosis and safety isolation mechanism is established, which can intervene in a timely manner to protect against sensor malfunctions or mechanical failures, thereby reducing the risk of use.

[0021] Easy to integrate and low cost: This invention is based on common and mature sensing solutions such as plantar pressure, IMU, motor encoder and current sensor. It achieves high-performance sensing through efficient fusion algorithm. It does not rely on unstable biological signals such as electromyography or computationally complex deep learning models, which reduces the requirements for hardware computing power and is conducive to embedded integration of the system, reducing cost and power consumption. Attached Figure Description

[0022] Figure 1 This is a diagram of a multi-source fusion perception system architecture for an exoskeleton robot proposed in this invention. Figure 2 This is a logic diagram of a multi-source fusion sensing system for exoskeleton robots proposed in this invention. Detailed Implementation

[0023] The technical solution of the present invention will be further described in detail below with reference to specific embodiments.

[0024] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "setting" should be interpreted broadly. For example, they can refer to a fixed connection or setting, a detachable connection or setting, or an integral connection or setting. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0025] Example 1: Example of dynamic ground contact detection based on piezoelectric ceramic sensor In this embodiment, the plantar pressure acquisition module has been specifically optimized based on the basic hardware architecture. The plantar pressure acquisition module specifically uses three ring-shaped piezoelectric ceramic sensors, which are respectively encapsulated in flexible silicone pads and embedded in the center of the heel, the first metatarsal bone, and the fifth metatarsal bone of the exoskeleton foot component. The piezoelectric ceramic sensors have the characteristics of high response frequency and sensitive dynamic pressure measurement, and are particularly good at capturing the impact force signal at the moment the heel hits the ground.

[0026] The central processing unit executes the following specific algorithm based on the characteristics of this sensor: High-frequency sampling: Acquire the raw voltage signal of the piezoelectric ceramic sensor at a frequency of not less than 2kHz.

[0027] Impact feature extraction: Real-time calculation of the energy integral or peak change rate of the heel sensor signal within a short time window (e.g., 5ms); Unlike schemes that use static pressure thresholds, this method does not rely on absolute pressure values, but captures the dynamic change features of force.

[0028] Dynamic event determination: When the extracted impact feature value exceeds a dynamic threshold that is adaptively adjusted based on the signal energy of the previous few steps, a "ground contact event" flag is immediately generated. This dynamic threshold can be finely adjusted according to walking speed and terrain (preliminarily judged by IMU attitude angle), thereby effectively distinguishing between real gait switching and occasional slight foot sway, significantly improving the robustness of recognition on soft surfaces or when gait is unstable, and overcoming the shortcomings of poor adaptability of traditional fixed threshold methods.

[0029] Example 2: Lightweight architecture for distributed preprocessing and fusion This embodiment aims to reduce the computational load on the central processing unit and improve system reliability. This system employs a distributed preprocessing architecture: The inertial measurement module integrates a low-power microprocessor (such as an ARM Cortex-M0 core). This microprocessor runs attitude calculation algorithms (such as Mahony complementary filtering) locally on the module, converting the raw gyroscope and accelerometer data into stable pitch and roll angle data in real time, and then packaging and sending it to the central processing unit; thus avoiding the central processing unit from performing a large number of floating-point operations.

[0030] Inside the driver of the motor feedback module, a current loop torque estimation unit is added. This unit directly reads high-precision current sensor data, combines it with the pre-calibrated motor torque constant and joint reduction ratio, calculates the estimated value of the human-machine interaction torque at the joint output end in real time, and sends it through the CAN bus. This provides direct utilization of the motor's underlying high-frequency dynamic information and realizes torque sensing that traditional technologies have not been able to fully utilize.

[0031] The main responsibility of the central processing unit has become advanced fusion and decision-making: receiving the pre-processed "attitude angle", "joint torque", "joint angle" and plantar pressure raw signals from each module, performing event detection as described in Example 1, and performing final multi-source information fusion and security verification; reducing the computing power requirements of the central processing unit, allowing the system to use more economical and lower power consumption processors, while each module has a certain degree of local intelligence, improving the modularity and reliability of the system.

[0032] Example 3: Adaptive Fusion Perception Implementation for Ramp Walking This embodiment focuses on the system's perception and adaptation capabilities in non-horizontal terrain (such as slopes). The system hardware in this embodiment is the same as in Embodiment 1, but the algorithm of the central processing unit adds a terrain perception and parameter adaptation layer. Terrain recognition: The central processing unit continuously analyzes the attitude angles calculated by the inertial measurement modules fixed to the thigh and calf rods; by calculating the average pitch angle of the thigh and calf relative to the horizontal plane, the slope angle of the wearer can be estimated in real time (e.g., about 10 degrees uphill).

[0033] Adaptive sensing parameters: Pressure threshold adaptation: When going uphill or downhill, the distribution of plantar pressure differs significantly from that on flat ground; the system uses the slope angle as input to dynamically adjust the dynamic threshold used for ground contact detection in Example 1; for example, the impact of the heel may be reduced when going uphill, and the trigger threshold is lowered accordingly.

[0034] Expected posture mapping: The system has a pre-stored or online learning database of "typical gait postures" for different slopes; when the current slope is identified, the fusion algorithm will first compare the currently calculated limb posture with the expected posture for the corresponding slope, so as to more accurately determine the gait phase and correct the deviation of IMU data interpretation caused by the terrain.

[0035] Control mode fine-tuning: After identifying the uphill mode, the system can provide a larger assist torque curve in the support phase; in the downhill mode, the "follow-up" damping of the swing phase is enhanced to provide stability assistance; the terrain adaptive capability based on multi-source perception (pressure + attitude) is one of the core advantages of this invention compared with the environmentally unadaptable solutions in traditional technologies.

[0036] Example 4: Fusion Sensing Implementation Example of an Exoskeleton for Seated Hip and Knee Training This embodiment applies the multi-source fusion sensing system of the present invention to a seated lower limb rehabilitation training exoskeleton; in this scenario, the sole of the foot may not bear full weight, and the traditional sole pressure threshold method may fail.

[0037] Sensor configuration adjustments: The plantar pressure acquisition module is retained to detect slight foot contact or intent; the inertial measurement module is now prominently positioned on the thigh and calf joints. The motor feedback module becomes more critical for measuring joint torque.

[0038] Shift in perception fusion strategy: In this scenario, the system defines a new "motion intent" fusion logic: Active intention detection: When a patient attempts to move their leg, it first triggers muscle contraction, causing soft tissue deformation in the thigh / calf, which may be captured by a highly sensitive plantar pressure sensor or a contact force sensor on the strap (optional addition); at the same time, the patient's active force will cause the motor feedback module to detect small but continuous joint interaction torque.

[0039] Movement Triggering and Following: The central processing unit uses "continuous interactive torque" as the main triggering condition, supplemented by "subtle changes in limb posture (from IMU2)" and "foot contact signal" to comprehensively determine the patient's intention to initiate active movement; once confirmed, the system immediately enters the admittance control mode, mapping the interactive torque into compliant joint movements to assist the patient in completing the action.

[0040] Safety assurance: Throughout the process, the range of motion of the limbs calculated by the IMU2 is compared with the joint angles fed back by the motor encoder to ensure that the movement is carried out within a safe space and to prevent abnormal movements caused by patient spasms or other reasons.

[0041] This embodiment demonstrates the high scalability and scene adaptability of the multi-source fusion sensing framework of the present invention. By adjusting the weights and fusion logic of different sensor sources, the same hardware system can be applied to different rehabilitation stages from standing and walking to sitting posture training, which is difficult to achieve with existing technologies that rely on a single signal source or fixed logic.

[0042] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A multi-source fusion sensing system for exoskeleton robots, characterized in that, include: The plantar pressure acquisition module is installed on the foot component of the exoskeleton robot to collect real-time pressure distribution data of the wearer's feet. An inertial measurement module is installed on the limb links of the exoskeleton robot to measure the motion parameters of the limb links in real time and calculate the posture data of the limb in space. The motor feedback module is integrated inside the drive joint of the exoskeleton robot to acquire angle and current data of the drive joint in real time. The central processing unit is communicatively connected to the plantar pressure acquisition module, the inertial measurement module, and the motor feedback module, respectively, and is used to receive and process data from each module. The central processing unit is configured as follows: Based on plantar pressure distribution data, a dynamic threshold algorithm is used to determine the discrete switching points of the gait cycle in real time. By integrating limb posture data, joint angle data, and human-computer interaction torque data derived from current data, a comprehensive perception and prediction of the exoskeleton's motion state can be achieved. Redundancy verification and fault monitoring are performed by comparing the absolute attitude data calculated by the inertial measurement module with the relative joint angle data obtained by the motor feedback module.

2. The multi-source fusion sensing system for an exoskeleton robot according to claim 1, characterized in that, The plantar pressure acquisition module includes multiple pressure sensors, which are arranged in a triangular positioning manner at the corresponding positions of the heel, first metatarsal head, and fifth metatarsal head of the foot component.

3. The multi-source fusion sensing system for an exoskeleton robot according to claim 1, characterized in that, The dynamic threshold algorithm specifically includes: calculating the jump slope of the pressure data at the heel in real time, and when the jump slope exceeds a threshold value dynamically adjusted according to historical data, it is immediately determined as the starting point of the gait support phase.

4. A multi-source fusion sensing system for an exoskeleton robot according to claim 1, characterized in that, The inertial measurement module includes a three-axis gyroscope and a three-axis accelerometer, and optionally also includes a three-axis magnetometer, for calculating the heading angle, pitch angle and roll angle of the limb members in space.

5. A multi-source fusion sensing system for an exoskeleton robot according to claim 1, characterized in that, The motor feedback module includes a joint encoder and a current sensor. The central processing unit calculates the human-machine interaction torque data in real time based on the current data and the motor torque constant.

6. A multi-source fusion sensing system for an exoskeleton robot according to claim 5, characterized in that, The central processing unit is also configured to: when it is determined based on the plantar pressure distribution data that the wearer has entered the swing phase, control the exoskeleton robot to switch to impedance control mode, and dynamically adjust the motor output based on the rate of change of the limb posture data and the joint angle data to counteract the inertial resistance and frictional resistance of the mechanical structure.

7. A multi-source fusion sensing system for an exoskeleton robot according to claim 1, characterized in that, The central processing unit is also configured to: when the deviation between the absolute posture data and the relative joint angle data in the time domain is found to exceed a preset safety threshold through redundancy verification, determine that it is a sensor abnormality or mechanical failure, and control the exoskeleton robot to enter a safety damping state.

8. A multi-source fusion sensing method for exoskeleton robots, characterized in that, The multi-source fusion sensing system for an exoskeleton robot, as described in any one of claims 1-7, comprises: Simultaneously collect plantar pressure distribution data, limb inertial measurement data, and motor feedback data driving the joints; The plantar pressure distribution data is processed, and a dynamic threshold algorithm is used to identify gait switching events in real time. The inertial measurement data is processed to calculate the absolute posture of the limb in space; The motor feedback data is processed to obtain the relative angle of the joint, and the human-machine interaction torque is calculated based on the motor current. By integrating the gait switching event, the absolute posture, the relative angle of the joints, and the human-machine interaction torque, a comprehensive perception result of the exoskeleton robot's current motion state and the wearer's motion intention is generated; By continuously comparing the absolute posture with the relative angle of the joint, system-level redundancy verification and security monitoring are achieved.

9. The multi-source fusion sensing method for exoskeleton robots according to claim 8, characterized in that, The fusion step further includes: predicting the limb movement trend at the next moment based on the gait switching event and the absolute posture, and combining the human-machine interaction torque to provide feedforward or feedback input for the assist control or follow-up control of the exoskeleton robot.

10. An exoskeleton robot, characterized in that, The system includes a multi-source fusion sensing system for exoskeleton robots as described in any one of claims 1-7.

Images