Study On Robot Joint Angle And Position Capture Technologies Based On Inertial And Magnetic Sensors

In the process of human-robot collaboration,it is of great significance to capture the real-time joint angle and position of serial revolute manipulator in a flexible and convenient way.Although most robots are equipped with an encoder system to measure the rotation angle of each joint,the encoder is usually installed at the center of the joint,which is difficult to achieve in some applications with limited mechanical structure.For robots whose sensing system is not open,since users cannot read the joint angle and position information through the application programming interface,it is difficult to achieve the system integration.At this time,the inertial-magnetic sensor measurement unit(IMMU)can provide a robotic joint angle and position measurement technology.For economical robots,it can provide a low-cost joint angle and position measurement method.Moreover,it can be used as a secondary measurement technology,which is independent and complementary with the high-resolution encoder to ensure the reliability of the joint angle measurement.Further,it can provide necessary information for robotic fault diagnosis to find existing problems during working process.The main research contents and achievements of this paper are as follows.The first is the studies of attitude estimation algorithms based on a single inertial-magnetic sensor measurement unit.It is the basis of robotic joint angle and position estimation,which has different performance requirements in different application scenarios.Aiming at the problems such as low accuracy,high complexity and poor real-time performance of attitude estimation,this paper conducts in-depth research on attitude estimation methods and technologies by using sensor fusion technologies.This paper proposes a decision-tree based multi-model unscented Kalman filter,and establishes the decision tree strategy of automatic switching filter model.It can select the most reliable sensor data at current time for data fusion.Moreover,a new attitude estimator with direct compensation structure is proposed.It directly compensates the attitude obtained by integrating gyroscope data based on the gradient descent optimization method.The decoupling method of accelerometers and magnetometers is proposed.It first obtains the non-heading quaternion by fusing the acceleration and angular velocity.Then the heading quaternion is acquired by fusing the magnetic field and angular velocity.The final attitude is estimated by fusing the non-heading attitude and heading attitude.This algorithm achieves that the magnetometer only affects the yaw determination but does not affect the pitch and roll estimation.The second is the studies of rotation angle and axis attitude estimation of single rotation based on the inertial-magnetic sensor measurement unit.Since the motion of serial revolute manipulator is formed by multiple single rotation of joints,this study is the basis of robotic joint angle and position estimation.In this study,an incremental quaternion algorithm is proposed.According to the relations between the incremental quaternion and the axis angle pair,the estimation model of the rotation angle and axis of the single rotation is established.This method can estimate the rotation angle and axis of the object even the rotation axis is unknown.Experiment results show that the proposed method can accurately track the rotation angle and axis attitude of single rotation whether in dynamic or static rotation.The third is the studies of joint angle and position estimation of serial revolute manipulator.Based on the study of rotation angle and axis attitude estimation of the single-axis rotation,this work systematically studies the method and technique of joint angle and position capture of multi-degree of freedom serial revolute manipulator.An IMMU is fixed on each joint of the robot arm or the connected link,which is used to measure the attitude variation of the joint during motion process.Firstly,this paper analyses the motion regulation of each robotic joint.By decomposing the compound motion of the manipulator,the compound attitude measured by the IMMU is decomposed into the single attitude variation caused by the single rotation of each joint.Then the joint angle of each joint can be estimated.Based on the Denavit-Hartenberg parameters of the manipulator,the position estimation of each joint can be further realized.The method is completely independent of the encoder system in estimating the joint angle and position of the manipulator,and has no restriction on the installation position and attitude of the IMMU,which simplifies the process and method of joint angle and position estimation of serial revolute manipulator.This algorithm can be generalized to multiple degrees-of-freedom serial revolute manipulator.The final is the system design of attitude and position capture system of collaborative robot.This paper designs a set of IMMU and develop a software to capture robot motion based on partial achievements in attitude estimation of single IMMU and robotic joint angle estimation.It achieves the real-time pose and position monitoring and controlling of the robot.The joint angle and position estimation experiments verify the feasibility of joint angle and position measurement with inertial and magnetic sensors,and prove the correctness of the proposed algorithms in this paper.