Pseudo-linear measurement approach for heterogeneous multi-robot relative localization

Abstract

The purpose of relative localization (RL) is to locate and track one or more robots in another moving robot body-fixed coordinate frame using relative range and/or bearing measurements. Most available RL methods assume known initial conditions at the first encounter of an arbitrary robot, and the tracking is then followed using an extended Kalman filter (EKF). In case of poor filter initialization, these EKF based methods sometimes cause instability or demand longer settling time. To overcome this issue, this paper proposes a pseudo-linear measurement (PM) based technique for RL where true nonlinear measurements are algebraically transformed into PM. The proposed RL scheme is tested in Monte Carlo simulations for a heterogeneous multi-robot system comprising both aerial and ground robots. Results demonstrate that the proposed method performs RL with 5~10 cm positional accuracy and 0.075~0.1 rad orientational accuracy. The performance of the PM based RL is then compared against traditional EKF based methods with unknown filter initialization. The results demonstrate that the proposed method able to achieve both the positional and orientational accuracy within 12 iterations, whereas the traditional methods requires more than 250 iterations to achieve the same accuracy. The experiment validation of the proposed method was performed and results are congruent with the simulations.