Ph.D., Purdue University, Dec. 1998
Co-Major Professors: Anthony A. Maciejewski and V. Balakrishnan
Operations in hazardous and/or remote environments are often performed by robots. The hostile nature of these environments, however, significantly increases the likelihood of failures in the robots' subsystems. The difficulty and delay in the detection and subsequent correction of these faults makes the post-fault performance of the robots particularly important. An approach to dealing with failures is to build failure-tolerant manipulators, where the damaged system can be operated with minimal performance degradation. Though there has been considerable work in the area of failure tolerance, existing methods rely on effective failure detection. However, the issue of coping with failures prior to detection, or if they remain undetected, remains unaddressed. This work focuses on the analysis and design of inverse kinematic control schemes for manipulators experiencing undetected locked joint failures.
The first part of this work focuses on the post-fault analysis of a manipulator, emphasizing the effects of undetected locked-joint failures on the global/convergence behavior of the manipulator. Tools for identifying failure-tolerant regions of a manipulator's workspace are developed so that task completion can be guaranteed with common control schemes even if failures remain undetected. The second part of this work concentrates on the design of inverse kinematic control schemes to minimize the performance degradation due to the local effects of undetected locked-joint failures. Specifically, three different failure-tolerance schemes are proposed. Using a set of proposed performance measures, these schemes are experimentally evaluated and compared with a commonly used control scheme for tasks performed under both computer and human-control (i.e., teleoperation). Results of this study indicate that the post-failure performance of a manipulator, even with undetected failures, can be significantly improved through the use of appropriate failure-tolerant control schemes such as those proposed here.