Increased efficiency in robotic navigation by blending of local and global vision

Autonomous robotic navigation is a field of computer science with great potential for practical real-life application. Urban search-and-rescue work presents complex and demanding environments. In such situations, robots can prove to be invaluable and adaptable resources. Autonomous robots can also be of great value in everyday life, as intelligent aids for elderly or physically disabled persons. One factor remains equally important when designing an autonomous robot, namely a robust and accurate vision system. This research focused on the potential increase in efficiency in robotic navigation by utilizing a blended vision system, thereby resolving uncertainties related to incomplete or contradictory data. The first experiment considered an environment where the opening between obstacles is wide enough for the robot to use. The opening is gradually narrowed to a point where, due to low resolution, the global vision system falsely judges the opening too narrow. It was hypothesized that the local vision system would detect the error made by the global vision system, and allow the robot to use the opening. In the second experiment, the opening between obstacles is obscured to the global vision system. By blending local and global vision systems, it was hypothesized that the local vision would allow the robot to detect the opening and drive through when possible.