The position of autonomous robot is of great significance for the purpose of navigation in an indoor environment. The goal of this project is improved localisation in Non Line of sight(NLoS) conditions.
The current technologies during Line of sight (LoS) such as GPS, Bluetooth has an uncertainty of greater than 20 cm, whereas Ultra Wide Band sensors are accurate with a lower uncertainty of less than 8 cm.
In this project, the tracking of an autonomous robot, assisted by Ultra Wide Band (UWB) sensors, in an indoor environment by deploying Unscented Kalman Filter (UKF) has been explored to mitigate effect NLoS. The autonomous robot
in this project is addressed as "Deepcar", provided by a company known as SMPL robotics. A machine learning classification model is employed to detect whether a robot is in NLoS or LoS with the beacon by analysing the power of the impulse responses received at the beacon. This classification model is developed using TREK 1000 evaluation kit.
In the proposed approach, during NLoS conditions, persons in the indoor environment, who also carry UWB tags are used as non stationary beacons and are assigned variable weights in UKF, hence aiding in position estimation of the Deepcar, along
with stationary beacons.