Simultaneous Localization and Mapping SLAM of 3D spaces through Multi functional depth sensor fitted to a stable MAV platform

Abstract

The popular research topic in robotics is simultaneously making the robot to localize itself newlineand to build the map of the environment. An extensive application in the area of localization newlineand mapping includes movable robotics, self driving cars, unmanned airborne vehicles and newlineautonomous undersea vehicles. The problem deemed in this research work is the design of a newlinevisual autopilot for miniature sized unmanned aerial vehicle. The planned system is based on newlinedepth sensor system for autonomous localization and mapping. Targets in manned and newlineunmanned aircraft have been achieved using conservative routing sensors such as standard newlineIMUs for direction, GPS for location, pressure sensors for elevation sensing, radar, newlineultrasounds and laser range finder for detection of obstruction. Our specific interests involve newlinesmall and micro UAVs soaring close to the floor in disorderly environments like outdoor and newlineindoor environments. The visual sensors are passive, lightweight and can provide rich newlineinformation about the aircraft self-motion and surroundings structure. Therefore, computer newlinevision can be used for autonomous localization, which is a crucial step for small aerial robots newlinecontrol and guidance. Though localization and mapping are intricately joined problems and newlinethe research in SLAM has proved that solving either requires solving both, in this work it is newlinefocused on localization and mapping as the major output of interest. A type of sparse map that newlineconsists of landmarks for localization has been constructed. Within a restricted environment, newlinetracking using real time cameras will have extended and looping motions (when a task is newlineperformed by human, a robot cleaning a house or view of a room from different angles with newlinegraphical augmentation). It is essential to have repeatable localization in which there is no newlineoccurrence of gradual drift from ground truth. The most important cases are new regions will newlinebe explored by moving cameras without returning. At this stage the fully probabilistic SLAM newlineapproach is used, wh