This paper addresses the problem of control and commanding robots in their work spaces. The problem is approached by deriving the image-space and object-space spatial relations. The relations are represented as fuzzy sets to capture the ambiguity inherent to the linguistic terms for the relations. Distances and object sizes and shades have been explored as the fuzzy qualifiers in conditioning the spatial relation reasoning. By using a simple syntax, simple natural language sentences are mapped to fuzzy queries for robot commanding. Implementation and experiment have verified the feasibility and demonstrated potentials of the approach.