DESIGN OF AN INDOOR LOCALIZATION SYSTEM BASED ON LIDAR

1 INTRODUCTION Many different technologies like GPS and radar have been used to localize, to map environments and to detect objects over time those let vehicles to move. In this thesis, I analyze Lidar, especially in indoor, to let robots, drones, cars, etc. to move without a driver. One of the most popular sensor, the GPS, doesn’t give a good accuracy in dense urban environments because the signal availability changes over time, thus its potential for autonomous vehicles is limited; moreover, when it’s integrated with IMU, errors can be accumulated when the vehicle travels because the measured positions don’t follow the true position. Camera based localization has the drawback to be sensitive to the changes of illumination, to observation angle and it can be affected by weather too; furthermore, when it’s equipped with RANSAC algorithm, it’s not good to distinguish stationary objects from objects in movement. Ultrasonic based localization requires instead a very long processing time because its extractions of feature points aren’t accurate and, when it’s integrated with other sensors, errors can be accumulated; in addition, it’s incompatible for autonomous vehicles also due to its short range to detect obstacles. Radar based localization has the drawback to have many sources of noise, those make it a sensor with low robustness; moreover, it has some problems to recognize target’s surface properties, out of its data association. UWB sensor works well in indoor environment, but needs a reference node and it’s bad in outdoor environment because its signals have short range. Actually there isn’t a dominant sensor technology based on autonomous navigation. One of the candidates is Lidar, which guarantees smaller positioning errors than the other localization techniques like GPS, radar, sonar, camera, etc., but it’s also more expensive than to those ones. In autonomous navigation, it’s important that a car/robot/drone is able to estimate its position, all near objects and the surrounding environment in every instant, to move and follow a path toward a destination and to avoid every object/obstacle in every instant. In common words, it’s very important to make a safe journey without accidents. 1.1 OBJECTIVE In the past, people explored unknown environments risking to be injured; then they created vehicles, which were at first non autonomous and due to that reason there were accidents, especially when the driver is distracted; only in the last years people developed robots, drones and autonomous cars which are finally able to avoid obstacles not only when they follow a path, but also when they explore in an unknown environment. The aim of this thesis is to describe the Lidar, explaining the reasons for which it was created and used, the differences respect to other technologies, why is so important and how can be used for autonomous navigation. 1.2 OVERVIEW Chapter 2 explains the role of Lidar in autonomous navigation systems, their problems and their respective solutions through obstacle avoidance algorithms. Chapter 3 describes Lidar’s technology, history, characteristics, tipologies, advantages, disadvantages, use cases (especially for autonomous navigation), models (I will not include everything), context, measured data and its errors. Chapter 4 compares different localization algorithms and map extraction techniques used in Lidar and describes SLAM. Chapter 5 describes Matlab’s Robotics System Toolbox explaining its commands too.