Browsing by Author "Visinsky, Monica Lynn"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Dynamic fault detection and intelligent fault tolerance for robotics
    (1994) Visinsky, Monica Lynn; Cavallaro, Joseph R.
    As robotics has begun to spread from the accessible arenas of laboratories and industry into more dangerous and remote environments such as space or hazardous waste remediation tanks, the need for autonomous and intelligent robotic fault detection and fault tolerance has increased. Our original fault detection and tolerance algorithms enabled the robot to detect and react to internal failures in the sensors or motors. The current research extends this original work to provide a more flexible, dynamic, and intelligent fault tolerance system. This dissertation first presents the three-layer framework concept we developed to divide the fault tolerance problem into modular layers. The lowest layer is the servo layer which contains the plant or robot and robot controller. The next layer is the interface which monitors the servo layer for failures and responds with instinctive tolerance actions. The interface also informs the higher level supervisor of failure events so the supervisor can perform long term analyses of the failures, keep track of the changing status of the robot system, and develop more sophisticated fault tolerance strategies. The dissertation next focuses on the specific improvements made to the interface layer fault detection routines in the form of a new algorithm which uses regressor-based dynamics to produce efficient, model-based detection thresholds. A previous method providing the foundation for the new work as well as several versions of the new algorithm were tested on a pendulum, a two-link planar robot, and a non-planar 3DOF robot consisting of the first three joints of the PUMA 600. Finally, the dissertation discusses the encapsulation of the supervisor functions into an expert system. The expert system was implemented using a simulation of a planar four-link manipulator and a full PUMA 600. Future work can expand on the robotic fault tolerance system by incorporating other long term failure analysis algorithms into the expert system supervisor functions and by further testing the dynamic thresholds.
  • Thumbnail Image
    Item
    Fault detection and fault tolerance methods for robotics
    (1992) Visinsky, Monica Lynn; Cavallaro, Joseph R.; Walker, Ian D.
    Fault tolerance is increasingly important in modern autonomous or industrial robots. The ability to detect and tolerate failures allows robots to effectively cope with internal failures and continue performing designated tasks without the need for immediate human intervention. To support these fault tolerant capabilities, methods of detecting and isolating failures must be perfected. This thesis presents new fault detection algorithms which detect failures in robot components using analytical redundancy relations. The robot components critical to fault detection are revealed using an extended fault tree analysis. The thesis validates the algorithms using a simulated robot failure testbed. An intelligent fault tolerance framework is proposed in which a fault tree database and the detection algorithms work together to detect and tolerate sensor or motor failures in a robot system. Future work will expand the detection and tolerance routines and embed the framework into a more flexible expert system package.