Electronic Embedded System for Stair Recognition Based on Possibilistic Modeling of Ultrasonic Signal

Safe navigation is a common challenge for humans, as well as for mobile robots. Embedded devices offering autonomous navigation are increasingly required. To fulfill this requirement, various navigation systems have been proposed. Most of them are designed to avoid collisions with objects from the scene. However, some other obstacles, such as stairs, are posing a real danger and unfortunately, they did not attract sufficient attention to get a reliable and practical solution. In this paper, we propose a new electronic system for staircase detection and recognition based on ultrasonic (US) signals. The designed system can be used for indoor and outdoor applications to detect stairs and to recognize their nature (ascending or descending). The challenge here is to process in real-time a small sample size of an ultrasonic signal. Various features in time and frequency are used to characterize as much as possible the ultrasonic signals acquired for each of the two classes of stairs. Possibilistic models of the selected features are built for each class. The paper presents the electronic architecture, as well as the ultrasonic signal processing schemes for the stair recognition task. A rate of up to 93% is reached for staircase recognition using the proposed system.