On-Board Visual Odometry And Autonomous Control Of A Quadrotor Micro Aerial Vehicle

This paper presents the autonomous control of a fully customized quadrotor MAV in UPS-denied environments based on onboard visual odometry. To achieve the best integration and efficiency, key modules of the MAV hardware system have been investigated and optimized iteratively, inclusive of mechanical frame, propulsion system, flight controller and the visual sensor. An optical-flow module is developed with the low-cost and small-scaled visual sensor as well as height estimation. In order to achieve real-time and high-accuracy position feedback, vision odometry is estimated based on pixel intensity block matching with rotational motions compensated. Subsequently, autonomous path tracking is achieved from the fusion of corresponding sensing feedback with proper design of a Kalman Filter.