Synthia – A Simulation of Driverless Car Training

It’s an nice cartoon simulation, but I want to see the real video. In the simulation I noted:
-Ambient light. Cool, that’s a real problem in only a few parts of the world, like New Zealand. As are long periods of sun-strike as a consequence of the low zenith of the sun, especially in winter, but also all year round which has caused problems for other technologies, like supermarket scanners (ask me about it)
-It didn’t seem to slow down for speed humps. Very bad for car suspension as taxi drivers at airports keep telling me.
-It didn’t seem to slow down for pedestrian crossings. As we all know, a large percentage of people walk with 2 eyes on their smartphone and relying on peripheral vision for way-finding. The law says they have right of way on a pedestrian crossing that doesn’t have signals.
-It appeared to be good in snow and other weather conditions, but this appears to be based on locations where it already has great information about existing street furniture. The problem there is things can change daily. Road maintenance for example can be unplanned, or may proceed at different times to what is planned. In many cases that information is not shared with car navigation and other data sources and it is also not specific. It may say that there is a project happening on Monday to Friday between the hours of 09:00 and 16:00, but not that there will be a digger and a big hole in the middle of the road surrounded by cones as there are outside my driveway right now. In bad weather such as snow, the systems might not see holes surrounded by cones or covered with cloth.
-People are unpredictable, look at the number of crashes where cars (driven by intelligent people) hit parked vehicles, crash into trees and buildings where there are no other vehicles involved, or people just walk into the path of a car that can’t stop in time.

In 50 years or so when almost all cars (other than classics) have V2V communications, normal driving will be more predictable for autonomous vehicles, but a lot of people are predictably irrational.

If there were zones that were only allowed to be used by vehicles with compatible communications technology, that could work, but you would still have to bar pedestrians from those locations.

This is also so easy in a perfect world, but humans are not perfect and they train AI’s. These simulations are important and we will enjoy many fringe technologies that will come from this testing that can go into people driven cars and make them safer as interim steps. That will save money and lives.

It’s interesting to see how well people can walk on bust Manhattan streets while looking at their mobiles, but we have all seen people walk into poles. 2 steps or a sudden slip onto a road with an average speed of 30mph.

If you think driverless cars will become commonplace in the next 10 years, I think you are being overly optimistic. I have spent a lifetime asking “when the benefits are so obvious, why aren’t we all doing it?” The answer is because we are people and we are not highly motivated to change and especially not when the change is expensive or inconvenient. Of course without idealists nothing would change and today its no longer the minority of us who want change.



About Luigi Cappel

Writer for hire, marketing consultant specialising in Location Based Services. Futurist and Public Speaker Auckland, New Zealand
This entry was posted in Artificial intelligence, Autonomous cars, car accidents, car crash, Crash, distracted pedestrians, Driverless car, People and tagged . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s