Obstacle avoidance using optical flow and occupancy grids: This video shows a simulation of obstacle avoidance for a small autonomous aircraft. Optical flow from a monocular camera is used to populate an occupancy grid (aka evidence grid), which is used to plan trajectories for obstacle avoidance. The occupancy grid serves two purposes: robustness to noisy optical flow measurements and the ability to plan a local trajectory rather than simply using differences in optical flow to compute a steering command. See my Projects page for more information. [17.8Mb AVI]

GPS-free navigation: This 3D visualization of a 2D simulation shows a robotic aircraft (unmanned aerial vehicle, or UAV) flying autonomously through a previously unknown forest. The only sensing onboard is a monocular camera and an inertial measurement unit, or IMU. The aircraft simultaneously estimates the locations of trees and its own location and velocity, allowing it to plan a path around the obstacles and to the goal posts. The translucent red cyclinders represent the uncertainty associated with the estimated position of each tree. This shows GPS-free navigation using only passive sensing, and has also been implemented on hardware. See my old projects page for more information. [5.1Mb MPEG]