An Improved RGB-D VFH plus Obstacle Avoidance Algorithm with Sensor Blindness Assumptions

Early obstacle avoidance algorithms make a lot of effort to cope with the unreliability of ranging sensors. While with modern high-quality sensors like RGB-D cameras, the algorithms turn out to be more effective. However, all these sensors still have the problem that they cannot tell the areas occluded by the front objects. This leads to some undesirable planning results to the goal. In the paper, we propose three reasonable assumptions to predict the occluded environment, making the algorithm more sophisticated to detect some potential paths. Our work is based on the original vector field histogram+ (VFH+) obstacle avoidance algorithms. Applying with the assumptions, improvement has been experienced in most scenarios with the RGB-D camera. Foreseeing penalty is also taken into account for early misleading prevention.