Learning gait patterns for the fuzzy synthesis of piped walk

The gait synthesis is one of the tasks that must be performed by the controller of a biped walking machine. A fuzzy logic controller (FLC) has been designed to perform this task. The initial rules have been obtained from biomechanical studies using the description of human limb motion during walking, given by different authors. A learning mechanism has been added to the FLC. A walk can be viewed as a sequence of steps, and each step is a cyclical and repeatable trajectory in the state space. We can relate this trajectory in the state space with a gait pattern that is described by a set of fuzzy rules (a gait description). Using this set of rules as a piece of knowledge, we define a pattern learning mechanism, based on an evolution system.<<ETX>>