Design and evaluation of an embedded system based SLAM applications

Simultaneous Localization And Mapping (SLAM) is a branch of algorithms widely used by autonomous robots operating in unknown environments. Research community has developed numerous SLAM algorithms in the last years. Several researches have presented optimizations into different approaches. However, they have not explored a system optimization from the algorithmic development level to the system hardware design. In some applications areas, such as indoor mapping, we would obviously benefit from low-cost sensors technology and SLAM implementations on a smart architecture. In this paper, a solution to the SLAM problem is presented. It is based on the co-design of a hardware architecture, a feature detector, a SLAM algorithm and an optimization methodology. Experiments were conducted with an instrumented vehicle. Results aim to demonstrate that our approach, based on low-cost sensors interfaced to an adequate architecture and an optimized algorithm, is good suitable to design embedded systems for SLAM applications in real time conditions.