GPS-independent localization for off-road vehicles using ultra-wideband (UWB)

At the present time, test and validation of autonomously driving vehicles cannot keep pace with their rapidly increasing popularity and development. This dilemma manifested in several incidents over the last years, revealing that suitable methodologies for test and validation of autonomous driving are needed. In order to develop and research test and validation strategies, a testbed for autonomous driving functionality is under development at our institute. This testbed contains an autonomous vehicle, which itself consists of different systems providing its autonomy. Key subsystems for this functionality are a path planning and a localization system, supplying the ability to determine the vehicle's position and future track as precisely as possible. In this paper we present a multi-sensorial localization solution suitable for test and validation of autonomous driving. Within this solution, a position is calculated based on time-of-flight measurement within a ultra wideband (UWB) network in addition to odometry and inertial measurements, using a tightly coupled extended Kalman filter with a constant turn rate and constant velocity (CTRV) model to determine the position. The main contribution of this paper is an experimental evaluation of a UWB-based GPS-independent localization solution achieving a localization accuracy of in median 0.25 m with three anchors and up to 0.17 m with seven anchors in dynamic driving situations.