CN-122019975-A - Internet of things management method for rehabilitation robot

CN122019975ACN 122019975 ACN122019975 ACN 122019975ACN-122019975-A

Abstract

The invention discloses an Internet of things management method of a rehabilitation robot, which relates to the technical field of rehabilitation robot management, wherein an LSTM time sequence prediction algorithm is introduced to construct an LSTM model on the basis of perception regulation, time sequence data of a temperature sensor and a distance sensor are subjected to feature extraction through an anomaly prediction module, the anomaly risk level of the temperature sensor and the distance sensor in the future is predicted by utilizing the pre-trained LSTM model, differential linkage compensation is implemented by controlling the operation of a heating module and a displacement module based on the anomaly risk level, pitch angle, roll angle and acceleration vector of the rehabilitation robot are calculated in real time, safety early warning of corresponding levels is generated based on corresponding safety thresholds, the posture of the rehabilitation robot is recovered to be normal, a safety self-checking mode is automatically entered, and an initial regulation flow is recovered after no anomaly. The management method realizes the robustness and the security risk controllability of the rehabilitation robot under the abnormal working condition of the rehabilitation process through the deep fusion of the multi-mode perception and the intelligent algorithm.

Inventors

FAN SHIRAN
Cai Xukang
TANG JIA
LI TONGYAO

Assignees

青岛一跃具身机器人有限公司

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (10)

1. The Internet of things management method of the rehabilitation robot is characterized by comprising the following steps of: s1, acquiring data, transmitting the data to a controller, determining an optimal temperature range and a gradient change rule corresponding to different distance intervals by accessing a distance-temperature correlation algorithm model, and controlling a heating module and a displacement module by the controller in a linkage manner based on a PID algorithm; S2, an LSTM time sequence prediction algorithm is introduced to construct an LSTM model on the basis of perception regulation, time sequence data of the temperature and distance sensors are extracted through an anomaly prediction module, the pre-trained LSTM model is used for predicting the anomaly risk level of the future temperature and distance sensors, and the heating module and the displacement module are controlled to operate based on the anomaly risk level to implement differential linkage compensation; And S3, calculating pitch angle, roll angle and acceleration vector of the rehabilitation robot in real time, generating safety early warning of corresponding grades based on corresponding safety thresholds, and automatically entering a safety self-checking mode when the posture of the rehabilitation robot is recovered to be normal, and recovering an initial regulation and control flow after no abnormality exists.
2. The method for managing the Internet of things of the rehabilitation robot according to claim 1, wherein the method for controlling the operation of the heating module and the displacement module to implement differential linkage compensation based on the abnormal risk level is characterized by comprising the following steps: If the predicted distance sensor is abnormal in degree, the running state data of the motor is called, the actual displacement distance of the moxibustion head and the deviation direction from a theoretical value are reversely pushed by combining the step angle and the transmission ratio of the stepping motor, the temperature compensation quantity corresponding to the deviation is reversely deduced by a distance-temperature correlation algorithm model, and the heating power is adjusted in advance; If the predicted temperature sensor is moderately abnormal, calculating a real temperature interval of a moxibustion point through a distance-temperature correlation algorithm model based on three-dimensional correlation data of historical temperature, distance and environmental temperature, current motor displacement parameters and an environmental temperature compensation value, and dynamically adjusting heating power to a corresponding gear.
3. The method for the Internet of things management of the rehabilitation robot according to claim 2, wherein the method is characterized in that the time sequence data of the temperature and distance sensor are extracted through an anomaly prediction module, and the anomaly risk level of the temperature and distance sensor in the future is predicted by using a pre-trained LSTM model, and the method comprises the following steps: Acquiring a continuous time sequence data stream with a fixed sampling period to form a sliding time window, and sliding a sampling point forward each time to generate training and reasoning samples; Carrying out multidimensional feature construction on the original sequence in each time window, wherein the multidimensional feature construction comprises a distance data feature, a temperature data feature and a cross-modal correlation feature; Performing supervised learning by using a historical anomaly dataset with labels, wherein anomaly tags are divided into three levels, including mild anomalies, moderate anomalies and severe anomalies; the LSTM network structure adopts a double-layer hidden layer, inputs a multidimensional feature vector sequence, and outputs probability distribution of a future abnormal risk level.
4. The method for the Internet of things management of the health maintenance robot according to claim 2, wherein when LSTM prediction or real-time rules determine that a distance sensor enters a moderate abnormality, a cross-mode compensation mechanism is started, and the actual distance is reversely deduced by using motor operation data and temperature linkage correction is completed: Reading running state data of the stepping motor, including forward and reverse rotation direction identification, pulse counting and accumulated running time length, and calculating theoretical displacement according to the known stepping angle and transmission ratio of the stepping motor: Theoretical displacement = pulse number x step angle/360 degrees x lead, comparing the theoretical displacement with a value reported by a distance sensor to obtain a deviation direction and an amplitude, substituting the deviation distance into a distance-temperature correlation algorithm model, and calculating the temperature variation to be compensated according to a gradient variation rule learned by the model; continuously comparing actual displacement fed back by the motor encoder with theoretical displacement calculated by the model, taking the difference value as a new PID control error to be input into a displacement control module, and dynamically adjusting the position of the stepping motor.
5. The method for Internet of things management of a health robot according to claim 2, wherein if the temperature sensor is moderately abnormal, calculating a real moxibustion point temperature by using historical associated data and cross-modal information, and performing dynamic power adjustment and later calibration: invoking three-dimensional associated data of temperature, distance and environment temperature to form a training/reference library, wherein each sample record comprises actual measured temperature, actual distance, environment temperature and heating power gear; According to the current motor displacement parameter and the ambient temperature compensation rule, a distance-temperature correlation algorithm model is called to reversely calculate the interval of the real temperature of the moxibustion point; The controller recalculates the target temperature according to the calculated real temperature interval and generates a corresponding PWM duty ratio to drive the heating module to output power so that the moxibustion point temperature reaches the expected comfortable interval.
6. The method of claim 3 wherein the distance data features include calculating a step change rate, a mean shift amplitude and a fluctuation variance, the temperature data features include calculating a variance fluctuation, a short-time gradient change and a sustained deviation from an ambient temperature difference, and the cross-modal correlation features include introducing a time delay correlation of the distance change and the temperature change.
7. The method for the Internet of things management of a health robot according to claim 1, wherein the method for generating the safety precaution of the corresponding level based on the corresponding safety threshold comprises the following steps: In the primary early warning, the pitch angle theta is larger than a first pitch angle threshold value or the roll angle phi is larger than a roll angle threshold value, and a user is prompted to adjust the posture of the rehabilitation robot; In the second-level intervention, the pitch angle theta is larger than a second pitch angle threshold value or the linear acceleration component a_z in the Z-axis direction is larger than a first component, and attitude compensation is started; in the third-stage emergency, the pitch angle theta is greater than a third pitch angle threshold value or the linear acceleration component a_z in the Z-axis direction is greater than a second component, the accidental overturning/large-amplitude abnormal movement is judged, and the three-break one-lock one-alarm operation is executed.
8. The method for managing the Internet of things of the health robot of claim 7, wherein the three-break-one-lock-one alarm operation is performed, comprising the steps of: The heating plate power supply loop is disconnected through the relay, the band-type brake signal is sent to the stepping motor driver, the steering engine control signal is disabled, the current position is maintained, emergency alarm information is sent to the cloud health management platform through the MQTT protocol, the emergency alarm information comprises a health maintenance robot ID, an abnormal time stamp, an attitude data snapshot and a temperature/distance curve, and the emergency alarm information is synchronously pushed to a user APP and a medical staff terminal and starts local acousto-optic alarm.
9. The method for managing the Internet of things of the rehabilitation robot according to claim 8, wherein the posture of the robot to be rehabilitation is recovered to be normal, and the robot automatically enters a safe self-checking mode, and the method comprises the following steps: checking the validity of a zero point of the temperature sensor and a distance sensor; gradually unlocking the motor and the heating module, and verifying displacement precision and temperature stability through test operation; And recovering the initial regulation and control flow after no abnormality.
10. The method for managing the Internet of things of the rehabilitation robot according to claim 1, wherein the controller controls the heating module and the displacement module in a linkage manner based on a PID algorithm, and the method comprises the following steps: the on-off frequency of the solid state relay is regulated by PWM signals for the heating module, and the power of the heating plate is dynamically changed; and for the displacement module, driving a stepping motor/steering engine to adjust the position of the moxibustion head, so that the distance and the temperature are in a comfortable coupling interval defined by a distance-temperature correlation algorithm model.

Description

Internet of things management method for rehabilitation robot Technical Field The invention relates to the technical field of health robot management, in particular to an Internet of things management method of a health robot. Background Moxibustion is a traditional Chinese medicine external treatment method which uses heat and medicine effect generated by burning moxa as a stimulus source and acts on specific acupoints or parts of a human body to achieve the purposes of warming and invigorating yang and qi, dispelling cold and removing dampness and dredging channels and collaterals. The traditional Chinese medicine health care preparation is relatively simple and convenient to operate and wide in application, is particularly suitable for people with cold constitution, weakness and chronic cold-dampness pain, and becomes a common means for traditional Chinese medicine health care and home health care. In the traditional Ai Jiukang robot, as the sensor is easily interfered by dust shielding, element aging and other factors in the long-term use of the robot, data drift or response delay occurs, and the existing scheme lacks effective abnormal pre-judging and cross-mode compensation mechanisms, so that the rehabilitation precision is obviously reduced along with the use time, and the comprehensive requirements of modern rehabilitation scenes on equipment intelligence, reliability and safety are difficult to meet. Disclosure of Invention The invention aims to provide an Internet of things management method of a health robot, which aims to solve the problem of the deficiency in the background technology. In order to achieve the purpose, the invention provides the technical scheme that the Internet of things management method of the rehabilitation robot comprises the following steps of: s1, acquiring data, transmitting the data to a controller, determining an optimal temperature range and a gradient change rule corresponding to different distance intervals by accessing a distance-temperature correlation algorithm model, and controlling a heating module and a displacement module by the controller in a linkage manner based on a PID algorithm; S2, an LSTM time sequence prediction algorithm is introduced to construct an LSTM model on the basis of perception regulation, time sequence data of the temperature and distance sensors are extracted through an anomaly prediction module, the pre-trained LSTM model is used for predicting the anomaly risk level of the future temperature and distance sensors, and the heating module and the displacement module are controlled to operate based on the anomaly risk level to implement differential linkage compensation; And S3, calculating pitch angle, roll angle and acceleration vector of the rehabilitation robot in real time, generating safety early warning of corresponding grades based on corresponding safety thresholds, and automatically entering a safety self-checking mode when the posture of the rehabilitation robot is recovered to be normal, and recovering an initial regulation and control flow after no abnormality exists. Preferably, the differential linkage compensation is implemented by controlling the operation of the heating module and the displacement module based on the abnormal risk level, and the method comprises the following steps: If the predicted distance sensor is abnormal in degree, the running state data of the motor is called, the actual displacement distance of the moxibustion head and the deviation direction from a theoretical value are reversely pushed by combining the step angle and the transmission ratio of the stepping motor, the temperature compensation quantity corresponding to the deviation is reversely deduced by a distance-temperature correlation algorithm model, and the heating power is adjusted in advance; If the predicted temperature sensor is moderately abnormal, calculating a real temperature interval of a moxibustion point through a distance-temperature correlation algorithm model based on three-dimensional correlation data of historical temperature, distance and environmental temperature, current motor displacement parameters and an environmental temperature compensation value, and dynamically adjusting heating power to a corresponding gear. Preferably, feature extraction is performed on time sequence data of the temperature and distance sensor through an anomaly prediction module, and the pre-trained LSTM model is used for predicting the future temperature and the anomaly risk level of the distance sensor, and the method comprises the following steps: Acquiring a continuous time sequence data stream with a fixed sampling period to form a sliding time window, and sliding a sampling point forward each time to generate training and reasoning samples; Carrying out multidimensional feature construction on the original sequence in each time window, wherein the multidimensional feature construction comprises a distance data feature, a temperature data