Vision Systems for Mobile Robots.

摘要

The first of the two objectives was to obtain detailed performance data on the ML/MD system currently in use. This system uses a pulsed laser diode and an eight-sided rotating mirror on a slowly rotating mast to scan the terrain in a 32 x 32 array once each mast revolution (about two seconds). A lens and 20 element linear photodiode array on the same mast, with 20 amplifiers, time gates, comparators, and latches serves as the detection scheme. Initial attempts to collect data indicated several deviations from expected results. These required some detailed analyses to confirm the calibration values, and some hardware troubleshooting and repair where necessary along with some improved software. Most of the actual measurements on artificial terrain were done by Clement. They show that the system, designed for a maximum range of three meters, only has a useful range in the laboratory of two meters, but that it does have sufficient resolution to reasonably insure the safety of a slow-moving vehicle in that both positive and negative obstacles are detected when their height approaches the radius of the wheels. Cross-path hazards were also reliably indicated. Ramp hazard detection was also investigated but the results were not quite so encouraging, although the basic difficulty is likely in the software. It is also clear that for some pitch angles, the field of view may be inadequate in the current system. Our second second objective was to design a similar system with extended range capability of use with a terrestrial vehicle with a maximum speed of about three meters per second. This design is documented by Hoogeveen (1983).