With potential missions for quadrotor micro-air vehicles (MAVs) calling for smaller, more agile vehicles, it is important to implement attitude controllers that allow the vehicle to reach any desired attitude without encountering computational singularities, as is the case when using an Euler angle representation. A computationally efficient quaternion-based state estimator is presented that enables the Army Research Laboratory's (ARL) 100-gram micro-quadrotor to determine its attitude during agile maneuvers using only an on-board gyroscope and accelerometer and a low-power processor. Inner and outer loop attitude and position controllers are also discussed that use the quaternion attitude representation to control the vehicle along aggressive trajectories with the assistance of an outside motion capture system. A trajectory generation algorithm is then described that leverages the quadrotor's inherent dynamics to allow it to reach extreme attitudes for applications such as perching on walls or ceilings and flying through small openings.