This paper describes a systematic approach to using robots to automate making difficult measurements in real-world environments. The approach is illustrated by a detailed description of the architecture's actual application to inspection of commercial aircraft for skin cracks and skin joint delamination. Several robot designs are evaluated; a small mobile robot with suction cup feet is found to be best in the practical context of commercial aircraft maintenance and inspection. Approaches are discussed for primary eddy-current flaw detection sensor signal understanding and for secondary vision-based guidance image understanding. The prospects for vision-based inspection are discussed.