To Örebro University

Industrial environments are rarely static and oftentheir configuration is continuously changing due to the materialtransfer flow. This is a major challenge for infrastructure freelocalization systems. In this paper we address this challengeby introducing a localization approach that uses a dualtimescaleapproach. The proposed approach - Dual-TimescaleNormal Distributions Transform Monte Carlo Localization (DTNDT-MCL) - is a particle filter based localization method,which simultaneously keeps track of the pose using an aprioriknown static map and a short-term map. The short-termmap is continuously updated and uses Normal DistributionsTransform Occupancy maps to maintain the current state ofthe environment. A key novelty of this approach is that it doesnot have to select an entire timescale map but rather use thebest timescale locally. The approach has real-time performanceand is evaluated using three datasets with increasing levels ofdynamics. We compare our approach against previously proposedNDT-MCL and commonly used SLAM algorithms andshow that DT-NDT-MCL outperforms competing algorithmswith regards to accuracy in all three test cases.