Autonomous flight through unknown environments in the presence of obstacles is a challenging problem for micro aerial vehicles (MAVs). A majority of the current state-of-art research focuses on modeling obstacles as opaque objects that can be easily sensed by optical sensors such as cameras or LiDARs. Since obstacles may not always be opaque, particularly in indoor environments with glass walls and windows, robots (like birds) have a difficult time navigating to the unknown environments.

In this thesis, we describe the design, modeling, control and sensing for a new class of micro aerial vehicles that can navigate unknown environments and are robust to collisions. In particular, we present the design of the Tiercel MAV: a small, agile, light weight, collision-resistant robot powered by a cellphone grade CPU. The Tiercel is able to localize using a visual-inertial odometry (VIO) algorithm running on board the robot with a single downward facing wide angle camera. Next, we characterize the effects of impacts and collisions on the visual-inertial odometry running on board the robot. We further develop the system architecture and components to enable the Tiercel to fly autonomously in an unknown space, detect collisions using its on board IMU, and leverage that information to build a 2D map of the environment. Finally, we demonstrate the capability of a swarm of three Tiercel robots to navigate autonomously through an unknown, obstacle ridden space while sustaining collisions with the environment. Finally, our approach exploits contact to infer the presence of transparent or reflective obstacles like glass walls, allowing us to naturally integrate touch with visual perception for SLAM.