An onboard simultaneous localization and mapping system for unmanned aerial vehicles

点击翻译按钮获取中文摘要

The goal of this project is to develop a Simultaneous Localization and Mapping (SLAM) system for use on multi-rotor Unmanned Aerial Vehicles (UAVs) flying in indoor environments. This system will rely solely on onboard sensing and computing with no GPS. It will provide pilots with increased situational awareness by creating a constantly updating map of the environment being explored. UAVs are emerging as a viable, low-cost platform for indoor applications. There is a growing list of commercial applications, including surveillance, exploration (e.g., mining operations and disaster sites), agriculture, and filmmaking. While there has been significant work done in this area, existing systems have several limitations that make them unsuitable for many commercial applications. For instance, data is often processed offline. This manner of data processing is not ideal because it precludes feedback to the pilot during flight, and more generally prohibits real-time applications. While real-time data processing approaches exist, they do so by streaming data back to a powerful base station computer. In real-world applications, communication may often be hampered due to wireless signal interference. In this project, we will investigate an approach whereby the UAV operates in a semi-autonomous mode. In this mode, the UAV maintains an onboard map allowing it to avoid obstacles and localize itself within the environment. The pilot is tasked with providing the UAV macro commands. For example, the pilot may identify the UAV's desired destination, while the UAV computes the safest navigation route and proceeds along it. This approach not only increases situational awareness for the pilot as the onboard map is streamed back to the user interface but also is robust to wireless signal interference, as the communication between controller and UAV does not have to be in real-time. We will examine the viability of such a system as a commercial product. This entails optimizing between the speed of the map generation system, the map resolution and accuracy, and the monetary cost of the hardware.