Development for Space Robotics with Emphasis on Fault-Tolerance, page 4

Level III Fault Tolerance – A serial robot with 10 degrees of freedom may have fault-tolerance at Levels II and III. This kinematic arrangement affords the robot complete dexterity even if any one of the joints failed and is locked. The manipulator has a regional structure with 6 degrees of freedom (that is built on a regional sub-structure with 3 degrees of freedom) and an orientation structure with 4 degrees of freedom. The use of two-roll joints in the regional structure provides positional fault-tolerance and the 4 DOF orientation structure retains all joint axes intersecting at right angles even if any one of the wrist joints fails. In the decision-making and control algorithm for this system, the fault-detection and identification sub-system identifies the faulty joint and informs the decision making subsystem. The decision making sub-system selects any six joints of the robot for application of the inverse kinematics subsystem. The inverse kinematics sub-system then returns the joint displacements to the decision-making sub-system or an error flag if the calculations were unsuccessful due to mathematical singularity or workspace violations. The entire decision-making and control system executes at approximately 300 Hz. on an Indigo R4400.

Dual Arm Operations - An advanced 17 degree-of-freedom dual arm robotic system is being integrated into a technology demonstration testbed. The robot is capable of demonstrating Level III fault tolerance with each of its redundant DOF arms and Level IV fault tolerance with its dual arm configuration. is a K/B-2017 Dexterous Manipulator manufactured by Robotic Research Corporation. It incorporates two seven-axis manipulators on a three-axis torso assembly providing a total of 17 DOF. Grippers attached to the end of the manipulators each provide an additional DOF, making the entire system 19 DOF. Each joint drive module is composed of an electric servomotor, harmonic drive gear reducer, and joint position and torque transducers. Control of the system will be based on a task oriented architecture as application of a large force or precise motion of the end-effectors would be identified by the operator for a given operation. The operator may then concentrate on task completion while the robot's command and control architecture manages the robot's redundant resources and makes compensations for any system faults. The control system determines the best robot motion or arm poise to meet the physical task requirements by optimizing one or more of a set of dual arm performance criteria. The program has identified and mathematically formulated 25 dual arm criteria. The criteria are addressed at three separate levels of control: Operation level, End-Effector level, and Manipulator level. At each level, criteria are used to characterize different distinct kinematic and dynamic attributes' Multiple criteria have been -successfully demonstrated in computer simulated dual arm task. The task include heavy payload lifting, fixtureless assembly on-the-fly, and work piece deformation.

Failure Detection and Isolation - The subject of machine self-diagnostics and failure detection and identification (FDI) in dynamic systems is not new, but the integration of such schemes into a real-time failure responsive control system is yet to be realized. A key element of such an FDI scheme is its capability to detect incipient failure, which gives the reconfiguration and recovery stages ample time to respond to the failure in such a manner as to minimize error excursions and reconfiguration shock.

Two basic types of FDI schemes exist: model-free and model-based methods. The latter methods are more suited for real-time autonomous systems as the former invariably tend to be off-line and require operator assistance. Model-based methods rely on the use of mathematical descriptions of the monitored system, called analytical redundancies. The methods involve the generation of a residual, whose on-line statistics are compared against nominal ones in a decision unit where fault detection, isolation and identification take place. The decision unit then generates a control impairment status (CIS) report, that provides quantitative and qualitative information about the status of the monitored system, based on which the failure responsive control system takes subsequent actions. Next Page ->