Range-Based Collaborative MSCKF Localization

In this paper, a framework for collaborative localization of heterogenous systems is presented. Making advantage of the original MSCKF framework, we design a collaborative MSCKF filter that operates in two levels and allows a decentralized 3D collaborative localization without use of external computation systems. To achieve that, based on MSCKF localization, we first propose a range based collaboration that we optimize using the extracted environment constraints, an operation allowed by the use of a truncated unscented kalman filtering updates. The collaborative filtering is managed to not impact the original MSCKF odometry properties. The framework is applied to collaborative localization of aerial and ground robots; experimental results show the effectiveness of the proposed method.