Safe Path Planning in an Uncertain-Configuration Space using RRT

This paper addresses the problem of safe path planning in an uncertain-configuration space. We consider the case of a car-like robot moving in an indoor environment (three-dimensional space). The extended Kalman filter (EKF) is a popular way to localize such a robot and to estimate its configuration uncertainty during navigation. Consequently, we supply an EKF with simulated measurements in order to compute realistic uncertainties (in a four-dimensional space) during path planning. We show that our safe-RRT algorithm, based upon rapidly-exploring random trees (RRT), is an efficient way to find a path in the resulting seven-dimensional uncertain configuration space