A Jacobian free approach for multi-robot relative localization

Abstract

This study presents a relative localization (RL) approach for an multi-robotics system (MRS), in which a robot detects and tracks one or more robots in its body-fixed coordinate system. A square-root cubature Kalman filter (SCKF) is employed to track the teammates' relative pose based on the high-frequency egocentric sensory data and the low-frequency inter-robot relative measurements (IRRM). This IRRM data consists of the relative range and the relative bearing between the tracking robot and its teammates. A series of Monte-Carlo simulations for a heterogeneous multi-robotic system is presented to evaluate the proposed SCKF-based RL scheme for different measurement noise configurations and different measurement update rates. To assess how the proposed SCKF-based RL scheme improves relative pose estimation, a comparison with the EKF and the general cubature Kalman filter-based RL schemes through numerical simulations are presented. The results suggest that the proposed SCKF-based RL scheme is a promising solution for relative pose estimation when an exteroceptive sensory system has high measurement uncertainty and/or low measurement update rate.