Research On Angle Measurement Method Of Soft-bodied Actuators Based On Multi-sensor Data Fusion

As the core power component of a soft robot,the soft-bodied actuators' natural softness makes it difficult to accurately measure the motion information of the softbodied actuators.The accurate measurement of the angle information of the soft-bodied actuators has important research significance for the motion control of the soft robot.Measuring the angle of the soft-bodied actuators often uses vision sensors,inertial sensors and curvature sensors.The vision sensors measurement is easily affected by factors such as obstruction,work area,and ambient light.The inertial sensors measurement is easily affected by the expansion of the airway embedded in the softbodied actuators,and the curvature sensors measurement has problems such as hysteresis and drift.Therefore,it is necessary to further study the sensor measurement scheme to improve the accuracy of the angle measurement of the software actuators.In this paper,based on the above three types of sensors,the angle of the softbodied actuators is easily affected by hysteresis,occlusion,and airway expansion.The soft-bodied actuators angle perception system based on inertial sensor,curvature sensor,and vision sensor is designed.That is,the curvature sensor senses the bending change of the soft-bodied actuators body,the attitude module senses the change of the angle of the end of the soft-bodied actuators,and the vision sensor senses the image information of the soft-bodied actuators' bending change.Based on multi-sensor data fusion method,it provides a platform for realizing high-precision angle measurement of soft-bodied actuators.In view of the contact sensors' measurement of the angle of the soft-bodied actuators is easily affected by hysteresis and airway expansion,this paper first uses inertial sensor and curvature sensor to measure the angle data of the soft-bodied actuators.The algorithm based on the combination of fuzzy inference and kalman filter can realize the data fusion of inertial sensor and curvature sensor,which can reduce the influence of the angle drift and the airway expansion of the soft sensor actuators.Based on the BP neural network and the long short-term memory network,the curvature sensor and the inertial sensor are fused respectively,which reduces the influence of the hysteresis and airway expansion of the contact sensors measurement soft-bodied actuators angle.The research results show that the long short-term memory network can better improve the angular accuracy of the contact sensors measurement softbodied actuators.In view of the problem of visual occlusion,this paper uses inertial sensor,curvature sensor and high-definition cameras to measure the soft-bodied actuators angle data.Based on support vector machine,long short-term memory network,kmeans clustering and BP neural network,fusion of high-definition camera and optimal contact sensors fusion value,respectively,to achieve visual occlusion data recognition.Through the method of kalman filter,the contact sensors compensates the angle data blocked by the vision sensor.The research results show that support vector machine can achieve better fusion effect.