A long time ago, not that far from where I'm living now - I studied hard to become an EE.
I succeeded, and in the process made some fun stuff.
One of the semester projects we did was a GPS controlled model car.
We were 3-4 in a group, and were given a former RF controlled model car with an Atmel minimum system with display+keypad. The servos had been hacked a bit, so the 80C52 system controlling the car could read out the direction and drive the internal DC motor in both directions. A power motor controller was built, so the car could drive back/forwards at varying speeds.
A GPS receiver and other stuff like keypad and display was also needed to record waypoints, execute program, clear, etc.
This was +12 years ago, and I am aware that the tech back then wasn't as fast as now.
However,
however,
yep - however, it was a great way to learn about hardware limits - and work around them.
It's like implementing a fully functional RIAA amplifier using ONLY a single 741 op-amp and passives. It isn't easy, you have to know a lot about poles/0-points, practical limitations, etc., and PSpice it to get the curves just right.
And yes, we got it working. A single 12 MHz Atmel CPU doing all the work, including the NMEA-XYZ_map plane decoding in real time.
What great fun it was to see the little car sprint around the parking lot, searching - searching - searching - and then racing off to the next spot. Ah, thanks schuuul
So, what I'm suggesting is going a little old-school. Teaching that a multi-gigahertz system isn't needed to control something as simple as an autonomous moving car.