There is no logic behind self driving cars. If you want self driven vehicles, take the train, bus, monorail, bullet train, etc.

The current design concepts of self-driving cars (without human participation and control) are a demonstrably bad idea, given that the core of their navigation on the roads is GPS based on transmission from satellites through open airwaves to pre-programmed computers on the vehicles.

Drivers with override capability would be required to address at least three high-risk points of failure in the system as currently designed.

- 1 - The satellites are unprotected from debris in space or attack by some potential enemy. Without the satellites being in their intended positions and fully functional, they can give the GPS inaccurate information.

This misinformation already happens occasionally on the GPS navigation programs I use (Garmin and GoogleMaps.) Sometimes my GPS tells me to go straight at T-intersections, turn where there are no intersections, etc.. Often it thinks I am on an entirely different road than I am on.

Some of this is because of drift in the clocking, and some of this is due to temporary misalignment of a satellite by infinitesimal fractions of degrees.

Consider what will happen when the self-driving vehicle trusts that there is a road, and turns onto it, when in fact there is no road. A human driver is required to address this risk.

- 2 - The transmissions from the satellites are through open air, which means it is possible to transmit false information from anywhere. Provided the transmission includes the satellite's signature, the car's navigation has no way to know the information is bogus. This makes the GPS hackable by people with nefarious intent.

Just as problematically, the transmissions can be subject to interference by electrical storms. That is not a big deal when I am in control of the car at 60 MPH and the GPS has to reset when the transmission is interrupted for a few seconds, but a blind 2 seconds at 60 MPG on an irregularly curved road is a big deal in an autonomous vehicle.

A human driver could act in the event of inaccurate map information or transmission interruption.

- 3 - The software of these computers is subject to the same issues that all software is, including requirements for updates, malware, etc..

This is in addition to the accuracy and currency of the maps that are uploaded. One of the problems I have had with the GPS has to do with problems with out of date maps and blocked roads due to road construction and construction based on things like utility repair. Even if the uploads are done regularly, we are still depending on people to be competent in uploading correct information and accurate revisions. What happens if a terrorist gets a job with a state Department of Transportation, and intentionally uploads false information?

A fleet of self-driving cars are a terrorist's dream. You do not even need 2% of cars to be autonomous in order to cause major economic havoc. What if even 1000 cars (out of the 250 million in the US) were to simply accelerate and at turn left all at the same time?

Given that functions like steering, acceleration, and brakes are already computerized, the driver would be unable to counteract the risk of a malware takeover of such functions.

This risk is already here. Cars systems are already hackable due to the combination of computerized systems with communication systems like OnStar.

(https://www.caranddriver.com/features/a15124906/can-your-car-be-hacked-feature/) +++

Seriously, the risks are too great, and damned near impossible to mitigate reliably given our current technology and design models. Humans having control is required as a failsafe, at the very least.

If we get away from a satellite-dependent GPS model, there is a fair chance the concept of driverless autonomous vehicles will suck less. We will see.