Probabilistic Mapping with Ultrasonic Distance Sensors

This paper proposes a probabilistic robotic mapping approach to merge ultrasonic distance readings by modelling them as Gaussian random variables and using scan matching to reduce uncertainty in mapping process. To account for the high angular uncertainty of ultrasonic distance sensors, both positive readings (detected objects) and negative readings (the lack of detection) are taken into account to update the measurements and create the updated environment map. To support this approach, the map consists simultaneously from two parts - free space scans are stored in occupancy grid and obstacle readings are represented as Gaussian variable feature set.