Fault-Tolerant Gait Planning of Multi-Legged Robots

A fault-tolerant gait of multi-legged systems is defined as a gait which can maintain the gait stability and continue its walking against the occurrence of a leg failure (Yang & Kim, 1998). The notion of the fault-tolerant gait comes from the fact that legged robots with static walking have inherent fault tolerance capability against a failure in a leg, since a failed leg for itself does not cause fatal breakdown or instability to walking motions (Nagy et al., 1994). This means that for a given type of failure, the problem of finding fault-tolerant gaits can be formulated with which legged robots can continue their walking after an occurrence of a failure, maintaining static stability. As a novel field of gait study, fault-tolerant gaits are worth researching since the feature of leg failure can be involved into the frame of gait study and its adverse effect on gait planning can be analyzed with performance criteria such as stability margin, stride length, etc. Fault-tolerant gaits are classified by the kind of leg failure to be tolerated and the point of time that fault tolerance is carried out, that is, before or after a failure occurs. Several algorithms of fault-tolerant gaits developed in the past can be sorted out by these categories (Chu & Pang, 2002; Lee & Hirose, 2002, Nagy et al, 1994; Yang, 2002). Among them, Yang (Yang, 2002) proposed fault-tolerant gaits for post-failure walking of quadruped robots with a locked joint failure, in which a joint of a leg is locked in a known place (Lewis & Maciejewski, 1997). As one of common failures that can be frequently observed in dynamics of robot manipulators, the locked joint failure reduces the number of degrees of freedom of the robot manipulator by one and consequently its workspace to a certain limit. To establish the fault-tolerance scheme, the reduction of the workspace of a failed leg should be interpreted and reflected based on gait study. The objective of this article is to develop the fault-tolerant gait algorithm for a locked joint failure when the model of legged robots is hexapod. Compared with the previous results on quadruped robots (Yang, 2002, Yang, 2003), the following aspects will be considered in this article for the motion of hexapod robots. First, quadruped robots can have only one gait pattern in static walking, (4å3å4å3···), i.e., one leg is lifted off and swung while other three legs are in the support phase. But hexapod robots can have variable gait pattern, i.e., tripod, quadruped and pentaped gaits. In this article, we show that fault tolerance can be realized for walking with any gait and, especially, periodic gaits can be generated using tripod and quadruped gaits for straight-line walking on even terrain. We also present faulttolerant gaits with non-zero crab angle, or a walking motion with the direction of