Quadrotor perception and planning in confined spaces

点击翻译按钮获取中文摘要

Small aerial vehicles are seeing explosive growth, due to their flexibility, ease of use and ability to collect valuable sensor data from an aerial perspective. Both public safety and precision agriculture have seen the most rapid adoption of the new technology, as these applications occur in open spaces far away from obstacles in the environment. There are many more applications, such as infrastructure inspection, target tracking and drone delivery, that will further expand the market for small UAVs, but are currently limited due to the difficulty of automatically determining vehicle motion, detecting obstacles and planning safe routes through the environment.This project seeks to address the current limitations of onboard autonomy by advancing vision based algorithms for localization and obstacle detection, and for safe motion planning onboard the vehicle. We will incorporate both static and gimballed camera information, construct dense maps of the environment and plan safe routes using a fixed set of efficient vehicle motions. The complete pipeline for autonomous operation near static obstacles will be developed and tested onboard our existing quadrotor aerial vehicle research platforms. The results will open the door to much safer operation of aerial vehicles in missions where their proximity to objects is of primary importance.