It is difficult to control the motion of the assembling parts precisely with current virtual reality I/O devices during virtual assembly. In this paper, a semantics recognition method is presented to capture the user抯 intention. A motion guidance method based on semantics recognition and constraint recognition is provided to help the user move or rotate the parts in the virtual environment freely and precisely. The semantics recognition is implemented according to the matching of spatial position, assembly relationship on assembly task-level rather than lower-level geometry constraints.The methods are implemenred in the development of VIRDAS(virtual reality design and assembly system)and a virtual assembly example of automotive engine component is also given in this paper.