In this thesis, I describe a robot localization algorithm and present two fancy quadrotor applications. I implement Multi-state Kalman filter which is a tightly coupled monocular visual-inertial odometry algorithm. This algorithm provides high-quality odometry with a low demanding on the computational resource. Based on this localization algorithm, I present a method allowing a quadrotor equipped with only onboard cameras and an IMU to catch a flying ball. Our system runs without any external infrastructure and with reasonable computational complexity. An online monocular vision-based ball trajectory estimator is designed in this system and applied to recover and predict the 3-D motion of a flying ball using only noisy 2-D observations. A multi-quadrotor system is designed fo...[ Read more ]

In this thesis, I describe a robot localization algorithm and present two fancy quadrotor applications. I implement Multi-state Kalman filter which is a tightly coupled monocular visual-inertial odometry algorithm. This algorithm provides high-quality odometry with a low demanding on the computational resource. Based on this localization algorithm, I present a method allowing a quadrotor equipped with only onboard cameras and an IMU to catch a flying ball. Our system runs without any external infrastructure and with reasonable computational complexity. An online monocular vision-based ball trajectory estimator is designed in this system and applied to recover and predict the 3-D motion of a flying ball using only noisy 2-D observations. A multi-quadrotor system is designed for interactive demonstration then. Our swarm system can recognize people's different gestures and reply with some specific actions. The architecture of our swarm system is well constructed and can work with a different number of quadrotors.