This project investigates an application of the Distance Transformation method to the n-dimensional path planning problem.
The capability of the Self-Organizing Map (SOM) to create topology preserving mappings of high-dimensional data has been exploited for temporal sequence processing. The SOM combined with the U-Matrix allows state transitions to be visualized as trajectories on the mapping. These states often form a smooth manifold in the high-dimensional space such that the process of navigating these manifolds is equivalent to the occurrence of a state transition. Although SOMs are able to produce an approximate representation of such manifolds, this fact has not been exploited in temporal sequence processing applications to extrapolate from the representation the intermediate states that could have been reached during a state transition.
This project investigate an approach where the Geodesic Self-Organizing Map (SOM) is used to approximate the smooth manifolds and apply distance transformations to construct more detailed trajectories which would allow users to gain more insight into the state transitions.
The opportunity ID for this research opportunity is 387