To make a fully functional self-driving car means replicating the human driving process inside a machine. This includes the control, perception, reasoning and situational awareness of a human driver. Nobody has managed to do this so far, at least not to replicate enough of the essential human capabilities for safe driving in complex environments. Automated driving is quite easy provided there are no surprises and the boundaries of safe motion are known. Those boundaries should be predictable within an adequate time-horizon, which could be as much as 10 seconds for high speed driving. The challenges are different from robot navigation and control at low speed; a mobile robot can stop if an obstacle is detected. But a highway vehicle cannot simply stop whenever it detects an uncertain situation. And as speeds become higher the challenges become greater. In this talk, a number of scenarios for automated motion control are presented. Some can be implemented today, but some rely on futuristic AI capabilities.