This is a past event
Many complex systems are described locally by linear mappings.
Many complex systems are described locally by linear mappings. For instance, in a robot manipulator, the mapping from joint velocities to end effector velocities is linear. Similarly, in a visual servoing application, the mapping from image feature velocities to object velocities is linear. Failures in these types of systems can often be modeled by removal of a row or column of the matrix describing the mapping. For instance, if a robot joint becomes stuck in one position, the column of the Jacobian matrix corresponding to that joint is removed. Similarly, if an image feature becomes occluded, the column of the visual Jacobian matrix corresponding to that feature is removed. The ratio of the Jacobian’s nominal and failed determinant measures the degradation due to failure, and it is termed the relative manipulability for robotic systems, and the relative perceptibility for visual systems. This talk will show that the mean squared relative manipulability/perceptibility is always constant, irrespective of the geometry. As a result of this principle, optimal fault tolerant manipulability / perceptibility is defined for manipulators as that geometry which maximizes the minimum manipulability / perceptibility. Examples of robotic manipulators and visual systems which have been designed using this technique will be presented, along with an overview of the University of Wyoming’s robotics research.
- Speaker
- Prof. John E. McInroy (University of Wyoming)
- Venue
- FN, G011