Technology is transforming almost all aspects of our lives, one of them is automation. The main motivation of automation is to help humans avoid performing tedious, high risk jobs. Automated driving, also known as autonomous driving, has been at the center of industrial and academic attention since a few decades now, thanks to its potential of making driving risk-free by enabling a highly efficient machine control the vehicle on roads. Apart from the common outdoor use-cases, several applications in indoor environments have also been extensively investigated. The primary ones include process automation and management in large factories and warehouses.

Localization of the autonomous vehicle is crucial to determine the path to be followed to reach the desired destination. Sensor fusion techniques are extensively investigated for this. However, the major challenge arising in indoor environment localization is obtaining accuracy in the scale of a few centimeters in real-time. In this thesis, we intend to address this challenge. The contributions of this thesis are two-fold. Firstly, we develop a low-cost testbed – Autonomous Ground Drone (AGD) – that enables us to develop sensor fusion and localization scheme for autonomous driving. Secondly, we employ Extended Kalman Filter (EKF) on the sensor combination of UWB, IMU, and Radar, and achieve a localization accuracy of 8 cm. Our localization scheme outperforms state of the art in this field in terms of accuracy, latency, and power consumption.

Keywords: Localization, Sensor Fusion, EKF, AGD, low-cost, real-time