SISTEM PENENTU LOKASI KENDARAAN OTONOM MENGGUNAKAN TEKNIK SENSOR FUSI DENGAN MODEL ACKERMANN STEERING

• Main Author: Tartila Sahid, Muhamad
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/53730
• Institution: Institut Teknologi Bandung

Human can easily moves from a distant place to another by driving vehicle as a transportation method. Within a day, driving takes human time from doing activities with the chance of having road accident that could happen. The time spent from driving could be used for something more productive such as work and rest. This research is made by developing a localization system for autonomous vehicle for the purpose of increasing daily human productivity while maintaining a lower rate of possible accidents. The localization system is made for autonomous vehicle to work and able to predict the position of vehicle in global coordinate to reach the destination safely. This system is developed because the GPS system that are currently used by vehicles has a low frequency rate in spite of having high accuracies given. Therefore, multiple sensors are used to accomodate and predict the location of vehicles during the time GPS doesn’t send its signal. The combination of sensors with GNSS GPS are done by using Kalman Filter algorithm. The algorithm works by estimating the state of position, velocity and estimation of a vehicle with a selected model. Ackermann kinematics model can be used for nonholonomic vehicles such as car to determine the linear and angular velocity of a vehicle. Unfortunately, the model is nonlinear and therefore ineligible to be performed by Kalman Filter. To overcome this problem, a modified version of the algorithm, namely Unscented Kalman Filter (UKF) is used to solve the nonlinearity of the kinematic model. Within this research, three sensors of GNSS, IMU, and absolute encoder are used to increase the accuracy of the localization system. Inertial sensor IMU are used to measure and determine the velocity, acceleration, and orientation of a vehicle. Further, the absolute encoder is used to measure the steering angle. Meanwhile, the GNSS sensor are used to locate the absolute position of the vehicle in the form of Universal Transverse Mercator (UTM). These sensors are fused within UKF for predicting and correcting to determine the state of a vehicle. Particle Swarm Optimization (PSO) are used to increase the accuracy of the system by finding the optimum parameter of the UKF. The localization system designed within this study has an accuracy with RMSE of 0.056 m for position and 3.01° for orientation.