Design, and implement a computer vision method capable to detect worker bees in a beehive
Map Update for Long-Term Mobile Robot Visual Navigation
Design, implement and experimentally evaluate a method that adapts a map used by mobile robots to the environmental changes
The goal of this thesis is to develop methods for creating and updating maps for long-term mobile robot navigation in the changing environment, which is affected by illumination, weather conditions and seasonal changes. The main problem is to cooperate with uncertainty, whose accumulation in the map is necessary to be prevented. This uncertainty arises in the inaccuracy of sensory data or as a result of natural processes of a dynamic environment. In this work, a method of adaptive mapping is presented, which is able to integrate new information into the map, thanks to which it is able to fit its surroundings. At the same time, it also removes obsolete or no more relevant information from the map, which contaminates and worsens the final reliability of localization and robustness of navigation. This method is compared with other types of map and it is experimentally verified on the gathered dataset and evaluated according to the criteria related to environmental changes and the error of localization.
Assignment
This thesis will deal with methods that allow mobile robots to create, update and refine their environment models to keep up with the naturally occurring environment changes. These models will contain landmarks used in the teach-and-repeat method for visual navigation of mobile robots.The proposed method should be able to efficiently integrate the long-term observation of multiple landmarks into the representation, and retrieve the sufficient number of high-quality landmarks to enable visual teach-and-repeat navigation. This method will be integrated into Thethe Robotic Operating System (ROS) and experimentally verified on a real robotic platform.
Result:
The resulting thesis is available in CVUT's dspace system.
