Obstacle Identification for Vision Assisted Control Architecture of a Hybrid Mechanism Mobile Robot

This paper presents a monocular vision-based, unsupervised floor detection algorithm for semi-autonomous control of a Hybrid Mechanism Mobile Robot (HMMR). The paper primarily focuses on combining monocular vision cues with inertial sensing and ultrasonic ranging for on-line obstacle identification and path planning in the event of limited wireless connectivity. A novel, unsupervised vision algorithm was developed for floor detection and identifying traversable areas, in order to avoid obstacles in semi-autonomous control architecture. The floor detection algorithms were validated and experimentally tested in an indoor environment under various lighting conditions.