Research On Information Fusion Algorithm For Attitude Estimation Of Unicycle Robot

Attitude control is the key technology for the balance movement of the unicycle robot, and attitude estimation is very important for attitude control. There are three kinds of sensors that the unicycle robot always used, which are gyroscope、accelerometer and magnetometer. Whether the performance of sensors is good or not will directly affect the performance of balance controlling for the unicycle robot.Considered that the high-performance sensors are expensive, while MEMS inertial sensors have some advantages such as it is small、light、cheap and energy saving. so we chose MEMS inertial sensors to make up our attitude estimation system.However, the low-cost MEMS inertial sensors also have some shortages such as accelerometer has a heavy measurement noise and gyroscope has great random drift errors, which lead to a great difficulty to obtain accurate attitude information by each of them.To resolve the problem above, this paper sets up the posture detection system of the unicycle robot. And analyzes the strong points and shortcomings of different types of sensors in the system. Hardware foundation is built for unicycle robot attitude estimation.Deeply study the principle and process of extended Kalman filter(EKF) and unscented Kalman filter(UKF). Build the system model of unicycle robot by Euler method and simulation is made based on two kinds of algorithms.We installed the sensors on a PAN-TILT UNIT which have two degrees of freedom. We have done static and dynamic tests of roll angles and pitch angles estimation on the experimental platform.Taking into account that the accelerometer is sensitive to external vibration and non-gravitational, we get the relationship curve between vibration intensity and the influence to accelerometer through experiment. Through the deeply analysis to Kalman filter algorithm and coupled with a strong validation of experimental results,this paper extends and modifies the unscented Kalman filter algorithm to deal with the problem. The experimental results indicate that the extended Kalman filter algorithm after modification(AUKF) has a good effect for solving unicycle robot posture detecting problems and eventually gets the more accurate attitude angles.