Indonesia Conference Directory

The combination of inertial measurement unit and global navigation satellite system is widely used in the localization of autonomous cars. However, global navigation satellite systems are highly dependent to the external conditions and have a low sampling rate. In order to make the localization of autonomous cars more reliable in various external condition, a virtual sensing system using Error State Kalman filter and Diagonal Recurrent Neural Network (DRNN) approach is proposed in this paper. In this proposed system, DRNN served as an estimator for the location of autonomous car. DRNN is applied due to its independency against external condition, the ability to learn, and also its faster sampling rate compared to global navigation system. Implementation and testing of this new approach using Carla Simulator shows that the proposed system could correct the deviation caused by the absence of absolute position measurement. By having this alternative sensing method, it is expected that it would be able to replace the existing global navigation satellite systems and unlock the possibility for offline localization.