Computing C-space entropy for view planning with a generic range sensor model

We have recently introduced the concept of C-space entropy as a measure of knowledge of C-space for sensor-based path planning and exploration for general robot-sensor systems. The robot plans the next sensing action to maximally reduce the expected C-space entropy, also called the maximal expected entropy reduction, or MER criterion. The expected C-space entropy computation, however, made two idealized assumptions. The first was that the sensor field of view (FOV) is a point; and the second was that no occlusion (or visibility) constraints are taken into account, i.e., as if the obstacles are transparent. We extend the expected C-space entropy formulation where these two assumptions are relaxed, and consider a generic range sensor with non-zero volume FOV and occlusion constraints, thereby modelling a real range sensor. Planar simulations show that: (1) MER criterion results in significantly more efficient exploration than the naive physical space based criterion (such as maximize the unknown physical space volume), and (2) the new formulation with non-zero volume FOV results in further improvement over the point FOV based MER formulation. Preliminary experiments with the SFU eye-in-hand system, a PUMA 560 equipped with a wrist mounted range scanner corroborate the simulation results, however, for lack of space they are not reported here.