Uncertainty management in intelligent task planning of mobile robots

摘要

Abstract This paper focuses on uncertainty associated with intelligent task planning. The proposed framework combines the theory of fuzzy reasoning and decision analysis in order to deal with the uncertainties and redundant degrees of freedom in the generation of an optimal task plan. The method applies particularly to, but is not limited to, robotic task control in a dynamic environment. First, the paper discusses the various sources of uncertainty associated with task planning. These uncertainties lead to multiple candidate task sequences with varying degrees of confidence. Redundant degrees of freedom of the machine also lead to multiple candidate task plans and thus is considered as one type of uncertainty. In order to find the optima] plan, our model first deduces the likelihood of success of each candidate plan using fuzzy reasoning. Next, we use the decision analysis technique to determine which task sequence is optimal. The paper demonstrates this idea with two examples: (1) a robot that spots a golf ball out of several candidate targets and (2) an autonomous guided vehicle (AGV) that seeks the fastest route. The AGV example introduces multiple levels of success at the chance nodes.