1、Designing Fault Tolerant Manipulators:How Many Degrees-of-Freedom?Christiaan J. J. Paredis and Pradeep K. KhoslaDepartment of Electrical and Computer Engineering andThe Robotics Institute,Carnegie Mellon University,Pittsburgh, Pennsylvania 15213.AbstractOne of the most important parameters to consid
2、er when designing a manipulator is the number of degrees-of-freedom (DOFs). This article focuses on the question: How many DOFs are necessary and sufficient for fault tolerance and how should these DOFs be distributed along the length of the manipulator? A manipulator is fault tolerant if it can com
3、plete its task even when one of its joints fail sand is immobilized. The number of degrees-of-freedom needed for fault tolerance strongly depends on the knowledge available about the task. In this article, two approaches are explored. First, for the design of a General Purpose Fault Tolerant manipul
4、ator, it is assumed that neither the exact task trajectory, nor the redundancy resolution algorithm are known a priori and that the manipulator has no joint limits. In this case, two redundant DOFs are necessary and sufficient to sustain one joint failure as is demonstrate din two design templates f
5、or spatial fault tolerant manipulators. In a second approach, both the Cartesian task path and the redundancy resolution algorithm are assumed to be known. The design of such a Task Specific Fault Tolerant Manipulator requires only one degree-of-redundancy.1 IntroductionAs robots are being used in a
6、 growing range of applications, the issue of reliability becomes more and more important. Recently, with the Hubble telescope and the Mars Observer, NASA has experienced firsthand how devastating the consequences can be when a critical component fails during a multi-billion dollar mission. Space app
7、lications are particularly vulnerable to failure, because of the adverse environment (cosmic rays, solar particles etc.) and the demand for long term operation. In this context, NASA has started to incorporate fault tolerance in their robot designs (Wu et al. 1993). Reliability is also important in
8、medical robotics, because of the risk of the loss of human life. Although medical staff will probably always be on standby to take over in the case of a manipulator failure, the robot should at least be fail-safe, meaning it should fail into a safe configuration. A third domain of robot applications
9、 in which reliability is a major issue is the Environmental Restoration and Waste Management (ER&WM) program of the Department of Energy. Consider, for instance, the use of a manipulator in a nuclear environment where equipment has to be repaired or space has to be searched for radioactive contamina
10、tion. The manipulator system deployed in these kinds of critical tasks must be reliable, so that the successful completion of the task or the safe removal of the robot system is assured. In this article, we focus on fault tolerance as a technique to achieve reliability in manipulator systems. The tr
11、aditional approach to reliability has been that of fault intolerance, where the reliability of the system is assured by the use of high quality components. However, increasing system complexity and the necessity for long term operation have proven this approach inadequate. The system reliability can
12、 be further improved through redundancy. This design approach was already advocated in the early fifties by von Neumann in connection with the design of reliable computers: “The complete system must be organized in such a manner, that a malfunction of the whole automaton cannot be caused by the malf
13、unctioning of a single component, . , but only by the malfunctioning of a large number of them”(von Neumann 1956, p. 70). This is the basic principle of fault tolerance: add redundancy to compensate for possible failures of components. However, this does not mean that any kind of redundancy added to
14、 a system results in fault tolerance. The main goal of this article is therefore to shed some light on the redundancy requirements for fault tolerant manipulators. That is, how much redundancy is needed and how should this redundancy be distributed over the manipulator structure?The redundancy provi
15、sions needed for fault tolerance can be incorporated only at a price of increased complexity. This drawback can be overcome by a modular and structured design philosophy, as is advocated in (Schmitz, Khosla, and Kanade 1988; Fukuda et al. 1992; Sreevijayan 1992; Hui et al. 1993;Chen and Burdick 1995
16、). Modularity in hardware and software has the advantage of facilitating testing during the design phase and therefore reducing the chances for unanticipated faults. Modules also constitute natural boundaries to which faults can be confined. By including fault detection and recovery mechanisms in critical modules, the effect of local faults remains internal to the modules, totally transparent to the higher levels of the manipulator system. Such a modular design philosophy i
