Multiple robots localization via data sharing

Abstract

This thesis applies a systems engineering approach to identify the critical issues in using a robot localization technique for a swarm of unmanned systems operating in an urban environment. It starts by presenting a concept of operations requiring data sharing between multiple robots operating in a confined environment, and proceeds with the development of a localization technique based on observing the relative position of neighbor vehicles and then sharing this information with them. The centroids of the measured positions are fed into a Kalman filter as the measurement inputs. The Kalman filter merges measurement data with a predicted state from a simple kinematic model. A simulation developed in Python is used to compare the performance of developed data-sharing localization technique with the individual robot odometry. The simulation results show a significant improvement of robot localization precision while the simple odometry technique results with continuing growth of the estimation error.