Design of a Low Cost, General Purpose Robot

Abstract

This paper reports on current work at the University of California at Berkeley whose goal is the design and implementation of a relatively inexpensive, but versatile, experimental, computer-controlled robot suitable for use in either a research or educational setting. The Berkeley robot, dubbed Jason, is nearing completion and hardware tests are now being conducted. Jason is designed so as to permit it to navigate and manipulate simple objects in a real-world environment. It uses a variety of sensory-motor and communication devices; among these are an ultrasonic range, motion, and material detector, an isolated-word speech recognizer, a limited speech synthesizer, six inexpensive proximity detectors, and two arms for simple manipul ation, all of which are mounted on a platform chassis. The robot vehicle is remotely controlled, using radio telemetry, by a time-shared, virtual memory, HP-3000 mini-computer, utilizing adaptive learning programs. Jason was primarily constructed to explore: (1) how an inexpensive, real-world robot system might be designed, and (2) what problems a robot "encounters" and "creates" while performing tasks in a real-world environment populated by humans. The results of this research will hopefully enable us to design and build better (more reliable and safer) robots at a modest price that are still capable of performing a variety of interesting and useful tasks. Key Words and Phrases Robots, Artificial Intelligence, Problem Solving, Learning, Real-World Environment. CR Categories 3.62 (Learning and Adaptive Systems); 3.64 (Problem Solving); 3.69 (Miscellaneous-Robots).