Topological mapping for a warehouse robot
The goal of the project is to build a topological map of the warehouse given a metric map (i.e. position of landmarks/objects in the environment) in addition to localization of the robot. The task would require extracting a graph from the metric representation. This graph would then need to be converted into a set of graphs by partitioning into smaller sub-graphs (e.g. Kd-tree) called nodes. Each node would contain the location as well as relationship information with other neighboring nodes such as similarity. When the robot next visits a location, a smaller graph is created with the current objects present in the robot's field of view and is matched in the set of sub-graphs of the main graph thereby localizing the robot into a topological map. The data and initial map of the environment will be provided along with the graph matching method. The task would be to abstract the given metric map into a topological map and then be able to infer the position of the robot in that topological map.
Contact: Rafid Siddiqui