This paper describes a novel adaptation and use of terrain mapping and optimal path-planning to fuse large and diverse time varying data sets into a common structure used for path extraction. The method described uses a system of map distortion and manipulation to encode anticipated future states of the environment into a single map. By doing so, it effectively encodes time into the map, allowing paths to be planned incorporating short-term accuracy and an approximate long-term path which accounts for anticipated movement of obstacles in the environment. The path is refined with recalculation as progress is made along the path. An application of the technique in cooperation planning for multiple physical agents is presented in the context of the RoboCup robot football competition.
展开▼