Augmenting human performance in teleoperation tasks.

摘要

Human-guided teleoperation of complex machinery is an important and growing area of man-machine interactive tasks. The area is seriously hampered, though, by human difficulties with spatial reasoning when planning the motion of such complex objects. The result is frequent errors, costly repairs due to collisions with objects in the machine's environment, and unfinished tasks.; The objective of this thesis is to develop a new approach that would ameliorate and, eventually, eliminate the above problem. The idea is to augment human performance by machine spatial reasoning skills. The novelty of the approach is, first, in that human and machine intelligence are spliced in real time, thus resulting in a smooth and natural operation, and second, that the human operator and the machine are partners, each supplying complementary skills (unlike traditional master-slave systems).; The approach developed in this thesis is based on the use of arm configuration space (C-space) to simplify the problem presented to the human. The C-space approach transforms the problem into one humans are much better equipped to handle—moving a point in a maze—and results in significant improvement in performance, with a remarkable decrease in arm-obstacle collisions. Versions of the approach are described for two- and three-dimensional (3D) tasks, and tools are developed to efficiently interface the human and machine intelligence. Effectiveness of the C-space approach is demonstrated by a series of experiments with human subjects, showing a performance improvement on the order of magnitude in the 2D case and a factor of 2 to 4 in the 3D case, compared to usual work space control.; Finally, the C-space approach is extended to other 3 degree of freedom (DOF) motion planning tasks that involve translation and rotation of objects in the plane. This extension demonstrates that the C-space approach can be successfully applied to 3D and/or 3 DOF motion planning tasks where C-space is available. In addition, new tools are developed to facilitate testing of human performance in C-space motion planning over the Internet. These tools can be used in human-machine approaches for increasing (over the web) the test subject pool.