Using unscented Kalman filter to improve accuracy of estimation and comparison with extended Kalman filter

In sensing and perception area, State estimation techniques like Kalman filters use estimates of the co-variance of process and measurement noise to improve the accuracy of the measurements that are fed into controller. In mobile robotics navigation, mapping techniques use sensor data to understand the environment, while localization techniques use sensor data to understand where the robot is in the environment. State estimation techniques are used to improve the accuracy of a controller to both disturbances in the plant and noise in sensor measurements. Klaman filters are widely used in this area. While traditional standard Kalman filter is designed only for linear system. When it comes to the nonlinear system, it is extended Kalman filter that will be used. However, there are some shortcomings existing in extended Kalman filter, so in this study, unscented Kalman filter will be used to improve the accuracy of the GPS measurement by replacing the extended Kalman filter. Besides, a comparison between unscented Kalman filter and extended Kalman filter will be made. The main tool is the MATLAB, which can be used to write the codes, build model and make the simulation. This work can verify whether the unscented Kalman filter can estimate the state of the robot more accurately and find out the characteristic of these two Kalman filters.