Spatial Representation for Pragmatic Navigation

As described here, pragmatic navigation attempts to harness simple facts about a two-dimensional environment to facilitate travel through it without an explicit map. It relies upon predefined spatial representations whose explicit instances are learned during a sequence of trips through a fixed maze. Once learned, any of these instances can be applied to subsequent travel. Some of the representations are heuristic, as are the procedures that employ them. The resultant performance of an implementation, particularly when contrasted with traditional AI techniques, argues for path-finding guided by representations like those detailed here.