Comparison of the effectiveness of Extended Kalman Filter and machine learning algorithms in sensor fusion

On account of the wide application of the autonomous vehicle system, accurate and reliable location information is of great significance. Since the GPS and the IMU information doesn't perform well on their own respectively, however when merging the GPS and the IMU data from the sensors, the high precision of the navigation system is achieved. While the Extended Kalman Filter is the most prevalent in fusing the data, in this paper, algorithms of machine learning, the Random Forest Regression, the Decision Tree Regression, the Multiple Linear Regression and the Extreme Gradient Boosting for instance, algorithms of neural network, the Multiple Perceptron Neural Network for instance is applied for sensor fusion. The results from the experiment demonstrate a promising direction for machine learning and neural network in application in the sensor merging field.