Evaluation of Ambiguity in Fuzzy Algorithm Represented with Fuzzy Petri Net

Abstract

In order to represent a "human-like" behavior-decision ability in an autonomous mobile robot it is necessary to describe behavioral sequences while taking into account the recognition of ambiguous states. Fuzzy algorithms capable of expressing sequence flow including a mixture of crisp and fuzzy processing provide an efficient behavior-decision algorithm. We have already proposed the Modified Fuzzy Algorithm (MFA) with tuning functions where branch is selected according to a controlled threshold. Results from computer simulations were also shown to prove the effectiveness of this method for giving intelligent mobile robots the autonomous macro behavior decision abilities like humans. In this paper we extend our previous work by using Fuzzy Petri Nets (FPN) to express a fuzzy algorithm. Ambiguous state transition is evaluated by marking change of the fuzzy truth tokens. The state analysis mechanism supplied by the Petri net makes possible the objective and global evaluation of the ambiguity in multi-firing of sequential rules. This new method assures that we can express the fuzzy algorithm in which explosions of fuzziness do not occur. For illustration purposes we present simulation results obtained using this method for the behavior-decision fuzzy algorithm for an autonomous mobile robot.