Nonlinear Control with Extended Kalman Filter Design for Vehicle Trajectory Tracking Muhammad Aizzat Zakaria 1 , Hairi Zamzuri 2* , Rosbi Mamat 3 and Saiful Amri Mazlan 4 1,2 UTM-Proton Active Safety Laboratory, Universiti Teknologi Malaysia, Kuala Lumpur 3 Control and Mechatronics Laboratory, Faculty of Electrical Engineering, Universiti Teknologi Malaysia, Skudai 4 Vehicle System Engineering Research Laboratory, Malaysia-Japan International Institute of Technology, Universiti Teknologi Malaysia, Kuala Lumpur E-mail: * hairi@ic.utm.my Abstract. This article describes the control design strategies for vehicle path following. Considering the noise from GPS sensor, robust control strategies is needed to ensure the autonomous vehicle is able to track the predefined path. GPS sensor is known for errors that affects the position reading of the vehicle such as multipath, receiver and sattellite clock errors. Thus, Extended Kalman filter is used in this research to overcome this problem in order to have a smooth data reading from GPS sensor. The Extended Kalman filter is fused with nonlinear tracking controller to achieve and solve the vehicle navigation problem. The nonlinear controller development is also described in this paper. 1. Introduction Outdoor navigation focusing on trajectory tracking is one of the topics that researchers widely investigate. Several solutions for tracking problem have been proposed by researchers [1–11]. The solutions proposed assumed ideal path tracking without concerning the input noise. However, during a real implementation noise generated from sensors is the main concern. A low cost sensor used is commonly inaccurate or not precise enough. Suitable method must be implemented to overcome this issue so that the noisy sensor reading would not affect the trajectory tracking. Several approaches were introduced to eliminate the noise. Muhammad Asif et al. explores a vision guidance for tracking where unconventional gray scale conversion technique to improve the image and Perona Malik filter is used to reduce the noise effect [12]. Somphop et al. [13] combines the GPS with compass, encoders together with machine vision to address GPS inaccuracy. The author used machine vision to extract the road region from the field of view in front of the vehicle. The road region data was compared with pre-defined map and candidate vehicle position to improve the localization posture. Researchers in [14] used Extended Kalman filter fusion with two sensors device, odometric and sonar by on line adjustment of the input and measurement noise covariances developed together with suitable estimation algorithm. In paper [15], fuzzy logic technique integration with Extended Kalman filter was used to adapt with initial statistic assumption for possible noise change in sensor device. George Taylor et al. [16] developed MMGPS technique where the GPS position is determined by accuracy selection method. The GPS reading is not taken when the accuracy is low and adapted with odometer to incorporate