Abstract
A behaviour-based control has been considered as the best approach to control autonomous robots. The robots have been expected to assist people in a people living environment such as houses or offices. Such situations require natural interaction between people and the robots. One way to facilitate this is by deploying a natural language interface (NLI) for human-robot interaction. The major obstacle in developing the NLI is natural language is always ambiguous. The ambiguity problem occurs when a word in a human instruction may have more than one meaning. Up to date, there is no existing NLI processor which can well resolve the problem. This paper presents a framework and an approach for resolving the problem. The approach is developed by utilizing fuzzy sets and possibility theory on linguistic and context knowledge.