Collaborative Localization for Micro Aerial Vehicles

We present a framework for performing collaborative localization groups of micro aerial vehicles (MAV) that use vision-based sensing. The are each assumed to be equipped with monocular camera, and capable communicating other. This approach is developed as decentralized algorithm built in distributed fashion where individual relative pose estimation techniques combined the group localize against surrounding environments. MAVs initially detect match salient features between other create sparse reconstruction observed environment, which acts global map. Once map available, MAV individually performs feature detection tracking robust outlier rejection process estimate its own 6 degrees freedom. When needed, one or more can compute poses another through measurements, achieved by exploiting multiple view geometry concepts. These measurements then fused consistent result accurate estimates. results on image data from flights both simulation real life, discuss advantages improving accuracy.