EMB-SLAM: An Embedded Efficient Implementation of Rao-Blackwellized Particle Filter Based SLAM

Abstract

Simultaneous localization and mapping (SLAM) algorithms are an essential component for autonomous mobile robotics to be able to operate in a priori unknown environments. In the last two decades, plenty of SLAM algorithms have been developed and a number of optimizations have been done for those algorithms. However, rarely optimization approaches to low-cost and energy-efficient embedded systems that are suitable for indoor robotics have been done. The benefit of the development of embedded systems should be explored. With the emerging of new technologies (multi core, ARM® NEON™) which can greatly accelerate the processing speed, rethinking the implementation of algorithms should be done. In this work, a new embedded efficient Rao-Blackwellized particle filter based Simultaneous Mapping and Localization (EMB-SLAM) implementation is presented. It is based on the co-design with the multi-core embedded hardware, a SLAM algorithm and an optimization methodology. EMB-SLAM is tested with real datasets. Experiments show the real-time performance of this implementation, and demonstrate that the embedded system is suitable for realizing SLAM applications under real time constraints.