Embedded vision-based Monte-Carlo robot localisation without additional sensors

摘要: This paper presents a fast approach for vision-based self-localisation in the RoboCup middle size league without additional e.g. dead reckoning sensors. An omni-directional vision system extracts few features from image that are mapped to an sparse a-priori known map of environment using Monte Carlo filters. The filters also used model virtual odometry (mass-inertia model) which is maintained through filter itself. precision directly compared traditional identical data. We show stable and reactive while keeping processing time low.