Autonomous Systems

Hacking a car can be safely done from your living room. And depending how systems are designed, especially fleet type trucking or ups, hacking could involve a whole fleet of trucks, maybe to be used in some grand terror attack, for example.

IDK, just throwing out spit balls here bros.

Abefroman said:

I would think there would have to be many safeguards put into place that would disable any autonomous system for consumer use if certain criteria aren't met. People don't take care of their shit anymore. Routine maintenance is usually put off to save a buck or just straight up ignored. I would think testing of all the electronics, cameras and other things used in the car would have to be on some sort of schedule and if ignored the system would be disabled. At least I hope the fuck so.

Click to expand...

Another great point.

a_skeleton_03 said:

You can already do that now though.

Click to expand...

Having full control over the vehicle is long ways from just disabling it or shutting it down.

Sentagur said:

So far allot of talk about active detection in the autonomous cars. How feasible would it be to retrofit cars on the road with some short range transponders that would help the cars detect each other or maybe even talk to each other to pass along information about traffic conditions?
Lets say there are no privacy concerns and all meatbags maintain their cars to be in perfect running order. Would a system like that be possible as an assist to active detection at least?


Edit: Hell, why just work on modernizing car? Add high-tech autonomous express lanes to the highways and if you want to participate you need to have the beacon active and your car software uptodate.
Make the cars talk to the highway too.

Click to expand...

Very feasible and vehicle to vehicle and vehicle to infrastructure comms are being actively worked on:
Vehicle-to-vehicle - Wikipedia
http://www.nhtsa.gov/staticfiles/ru...-of-V2V-Technology-for-Application-812014.pdf

As cool as V2x is, it's not a generalized solution and future-looking autonomy is going to have to work on that dirt road in the back of your hunting ground, or anywhere really.

Besides that, most vehicles won't know where they are with a huge degree of accuracy and you'll absolutely want some perception sensors telling you where everything is relative to yourself.

I mean, if you're a navigation system and you have two sources of input, one from v2v saying, "Hi I'm a car somewhere between 10 and 20 meters from you, at somewhere between 75 and 80 degrees." and one from perception saying, "There's an object of this shape with center of mass at exactly 13.54 meters from you at exactly 78.54 degrees", you're going to want that perception input.

chaos said:

Tuco, mind if I ask why you seen quite measured in your predictions for autonomous driving? At a conference I talked to a guy from Tesla and another guy from Uber and they were talking it up like this shit was imminent. both were talking about autonomous trucking fleets within 5 years and standard in every consumer car in 10.

Click to expand...

What Tesla has in market now, supercruise, is imminent. It works well on the highway in good conditions. Delphi, Bosch et all have a system that's going into new cars and I think it'll be a standard feature in luxury cars in 5-10 years.

But general purpose autonomy like the google car is further than 10 years away because:
1. There's still a lot of cases where even the google car fails.
2. The sensor and processing payload for it is fairly expensive.

I can go into more detail (but it'd probably be repeating what I've already said in this thread), but the bottom line is there are a tremendous number of technical challenges to getting full autonomy, and once they are achieved there is a lot of commoditization and testing that needs to happen before it's a standard feature in consumer vehicles.