Organizing spatial data for robotics systems

Abstract Successful robotics systems require the organization and analysis of a tremendous amount of sensor data. Moreover, the current computer paradigm of “shape-from” algorithms requires multi-dimensional data from the sensors, e.g. 3-space location and surface normal. Such multi-dimensional vectors must be organized such that spatial searching is efficient and so that spatial proximity is easily determined. Unfortunately, even though methods of computational geometry have been quite successful in dealing with intersection and proximity problems in 2-D, they have not been so successful in dealing with data of dimension greater than two. We define the spatial proximity graph as a low-level organizational structure, and show how it can be built efficiently. Some examples are given.