Localizing a quadrotor with collisions: novel sensor design and applications to particle filtering
MetadataShow full item record
Collisions are typically seen as adverse events for quadrotors, with numerous researchers designing cages for minimizing the effect of impacts on the vehicles. In this thesis, we reverse this paradigm, treating collisions as an additional opportunity for localization. In order to gather collision information, a touch-only sensor with a protective frame is designed to sense the 2-D relative displacement due to inertial force between vehicle and the frame while collision happens. We provide a mathematical model that represents the displacement in terms of the poses of the protective frame and quadrotor is constructed and solved numerically, which helps analyze the distance from obstacles to the vehicle and collision direction. At last, a particle filter based localization observing the collision status is simulated, which verifies the theoretical feasibility utilizing collision information to perform localization in a known environment.