Unmanned aerial systems (UAS's) and in particular intelligent, autonomous rotorcraft and fixed-wing aircraft have the potential to significantly impact modern society. A few examples of their utility include aerial surveying in difficult-to-access terrain, precision agriculture, package delivery, moviemaking, infrastructure inspection, fire fighting, search and rescue, etc. Recently there has been a lot of interest in autonomous air vehicles for cargo delivery to improve cost and time associated with shipping goods. Finally, much of the technology for autonomy could be used as a pilot's aid to help in difficult tasks such as landing a helicopter on an oil rig in the high seas or in the personal air vehicles of the future which are envisioned to be operated by people without significant pilot training. While the technology for unmanned air vehicles operating day in and day out without constant human supervision is maturing steadily, much remains to be done to make these vehicles commonplace. We have identified a number of challenges that must be addressed for these vehicles to safely and efficiently conduct their tasks in the National Airspace System (NAS). Civilian applications of UASs must ensure that they can: 1. sense and avoid other vehicles and follow air traffic commands, 2. avoid the terrain and land without operator intervention, 3. react to contingencies such as engine out and lost link scenarios, and 4. be reliable and cost-effective. We propose to a combination of software algorithms and low-cost, low SWAP sensors that simultaneously solves the navigation and obstacle detection problem, especially as relates to operation in cluttered environments. That is, in this program we will show that it is possible for small autonomous air vehicles to reliably and safely fly in the first and last 50 feet of operation.