We describe the use of multiple wide-angle sonar range measurements to map the surroundings of an autonomous mobile robot. A sonar range reading provides information concerning empty and occupied volumes in a cone (subtending 30 degrees in our case) in front of the sensor. The reading is modelled as probability profiles projected onto a rasterized map, where somewhere occupied and everywhere empty areas are represented. Range measurements from multiple points of view (taken from multiple sensors on the robot, and from the same sensors after robot moves) are systematically integrated in the map. Overlapping empty volumes re-inforce each other, and serve to condense the range of occupied volumes. The map definition improves as more readings are added. The final map shows regions probably occupied, probably unoccupied, and unknown areas. The method deals effectively with clutter, and can be used for motion planning and for extended landmark recognition. This system has been tested on the Neptune mobile robot at CMU.

We describe a second generation system that drives a camera-equipped mobile robot through obstacle courses. The system, which evolved from earlier work by Moravec [6], incorporates a new path planner and has supported experiments with interest operators, motion estimation algorithms, search constraints, and speed-up methods. In this paper we concentrate on the effects of constraint and on speed improvement. We also indicate some of our plans for a follow-on system.

A continuing trend in computer representation of three dimensional synthetic scenes is the ever more accurate modelling of complex illumination effects. Such effects provide cues necessary for a convincing illusion of reality. The best current methods simulate multiple specular reflections and refractions, but handle at most one scattering bounce per light ray. They cannot accurately simulate diffuse light sources, nor indirect lighting via scattering media, without prohibitive increases in the already very large computing costs.

A pipelined multiprocessor interconnection method functionally equivalent to a full crossbar, but with a per processor cost proportional to the square of the log of the total number of processors, is presented.