1990 Year End Report. Autonomous Planetary Rover at Carnegie Mellon

Abstract

This report describes progress during 1990 at The Robotics Institute, Carnegie Mellon University, in research on a six-legged autonomous robot, called the Ambler, designed for planetary exploration. The report summarizes the achievements during calendar year 1990, and lists personnel and publications. Also included are several papers that describe significant aspects of the research. Much of the research in 1990 focused on understanding the capabilities of the robot itself and developing a software system capable of autonomously walking the Ambler through natural terrain. We also investigated algorithms for perceiving and acquiring rock samples, which were demonstrated on a separate testbed. Understanding Mechanism Capabilities?haracteristics important to mobile robots in general, and walking robots in particular, include their mobility, stability, and dynamic performance. We investigated each of these aspects using a combination of mathematical and geometrical analyses, simulation, and experiments using robotic mechanisms. Results largely confirmed our expectations about the Ambler's high degree of mobility, rigidity, and large tipover safety margin. Autonomous Walking?his area combined four distinct efforts: real-time control of the Ambler, perception of rugged terrain using laser range scanning, algorithms for planning leg and body moves, and task-level control to integrate and coordinate the various components. We embarked on a comprehensive set of experiments, both in simulation and on the Ambler, that systematically added new components and increased the complexity of the environment traversed. These efforts culminated at year end in a successful demonstration to NASA consisting of some 25 autonomous steps walking on sandy, rolling terrain and crossing boulders, a wooden ramp, and stepping out of a meter deep trench. In total, the Ambler autonomously walked over a kilometer in 1990. Sample Acquisition?n preparation for demonstrating a complete mission scenario, we developed methods for perceiving the three dimensional shape of rocks in natural, cluttered surroundings using a light-striper. The shape descriptions were used to plan and execute grasp procedures to collect the rocks. A graphical user interface enabled humans to interact with the system, aid in the selection of rocks, and monitor progress. The methods were demonstrated using the Robot World testbed, and proved quite robust in practice.