This document discusses different approaches to decision making for self-driving cars, including state machines, reinforcement learning, and relational learning. It describes 5 levels of vehicle automation from no automation to full automation. For decision making, level 1 uses state machines with preprogrammed rules, level 2 adds reinforcement learning to accumulate knowledge of rewarding actions, and level 3 uses relational learning and cognitive prediction. Examples are given of using these approaches for tasks like controlling car lights and navigation.