Autonomous software holds the promise of new operation possibilities, easier design and development and lower operating costs. However, as those system close control loops and arbitrate resources on board with specialized reasoning, the range of possible situations becomes very large and uncontrollable from the outside, making conventional scenario-based testing very inefficient. Analytic verification and validation (V&V) techniques, and model checking in particular, can provide significant help for designing autonomous systems in a more efficient and reliable manner, by providing a better coverage and allowing early error detection. This article discusses the general issue of V&V of autonomy software, with an emphasis towards model-based autonomy, model-checking techniques and concrete experiments at NASA. Pecheur, Charles Ames Research Center NASA/TM-2000-209602, A-00V0022, NAS 1.15:209602