Monocular Vision-based Exploration with an Outdoor Non-Holonomic Mobile Robot

M. Alencastre-Miranda, L. Muñoz-Gomez, and R. Swain-Oropeza (Mexico)


Outdoor mobile robotics, exploration, computer vision, reactive navigation, unknown natural environments


This paper presents an approach to explore unknown natural environments in real time with an autonomous outdoor mobile robot using: a fast monocular vision algorithm, reactive navigation, non holonomic kinematics constraints and bezier curves. In order to find traversable areas and recognize different elements in natural scenes (grass, trees, sky, etc.), a fast monocular vision algorithm is used; this algorithm relies on a color segmentation and Markov Random Fields (MRFs) classification procedures. To navigate, our mobile robot executes bezier curves trajectories through traversable areas obtained from vision algorithm, considering its non holonomic kinematics constraints and odometry data. To perform exploration of unknown natural environments, we use a reactive navigation method. On-line (physical robot tests) results are presented.

Important Links:

Go Back