The use of a laser imaging system for automated vehicle guidance and space servicing tasks

This thesis presents a simulation of a laser imaging system for automated vehicle guidance and space servicing tasks. Data generated by this system are used to command a robot navigating in an unconstrained natural terrain. A computer graphics simulation program which emulates laser range scanners was developed and used for generating the required range images in this research. This simulation eliminates the need for an expensive laser scanner for developing laser data processing algorithms. Because the precise geometric information of the artificial objects in the scene is available, the simulation program can also be used as a testbed for evaluating laser data processing algorithms. Techniques and algorithms for converting laser range data to terrain maps were investigated and are presented in this thesis. A noise insensitive edge detector was developed for extracting wireframes of objects from the range data. By combining this edge detector with the Locus algorithm developed at Carnegie Mellon University, topographical terrain maps were derived from the sensory data. A 3-D sensor-based direct search algorithm was developed for finding collision-free optimal navigation paths using the terrain maps. This path planner generates paths of any desired degree of smoothness which can be readily used by the steering mechanism of mobile robots. A procedure for using a laser range finder and infrared proximity sensor for locating a satellite in space is also presented. The ultimate goal of this research is to develop a real time laser imaging system for automated vehicle guidance and object recognition.