Fast local re-planning is necessary for safe autonomous navigation in the presence of unknown and/or moving obstacles.
Recent works have raised interest for active perception methods in motion planning. Such methods include perception awareness in the planning problem. However, how to represent perception awareness is an open research problem. It depends on the representation of the environment. Key requirements for a suitable environment representation are that it can be incrementally updated on-line, light-weight, and it contains confidence measurements on the occupancy.
Fast local re-planning is necessary for safe autonomous navigation in the presence of unknown and/or moving obstacles. Recent works have raised interest for active perception methods in motion planning. Such methods include perception awareness in the planning problem. However, how to represent perception awareness is an open research problem. It depends on the representation of the environment. Key requirements for a suitable environment representation are that it can be incrementally updated on-line, light-weight, and it contains confidence measurements on the occupancy.
In this thesis, we will evaluate commonly used mapping methods (e.g., voxel map, point-clouds, ...) for the local re-planning problem. The final goal is to based on such evaluation a recommendation or even implementation of a novel mapping solution.
Requirements: - Experience with mapping in robotics preferable but not required - Programming experience in C++ and Python.
In this thesis, we will evaluate commonly used mapping methods (e.g., voxel map, point-clouds, ...) for the local re-planning problem. The final goal is to based on such evaluation a recommendation or even implementation of a novel mapping solution. Requirements: - Experience with mapping in robotics preferable but not required - Programming experience in C++ and Python.