The main goal of this research is to investigate new methods intended to increase the degree of driving automation of autonomous vehicles operating in realistic conditions, in both on-road and o-road environments. Safe driving in an urban, extra-urban, or, in general, uncertain environments requires the continuous combination of three important complementary functions: (i) interpretation of the surrounding environment (sense); (ii) evaluation of safe-driving plans taking into account the risk of collision or unsafe driving according to the characteristics of the scene, including any moving participants (plan); (iii) action, considering the outputs of the two previous tasks and involving an on-line reactive navigator able to avoid any choices leading to probabilistically inevitable collisions in the near future (act).