Seems like the argument being made is that autonomy isn't possible until safety with an actual driver is dramatically improved. Once there are fewer accidents then autonomy could be considered and implemented. Sounds like a good strategy to me, but proving out 'just' crash avoidance features sounds like a very, long process. I don't expect to see autonomy for a very, very long time...
All good points in this article. I especially likes the point about redundancy, since that seems to be missed by those who think self-driving cars can never work. If one type of communication fails, or if one type of sensing fails, others will take over, and the car will operate in a degraded mode.
Honestly, unless people are talking about different things when they say "self driving," I can't begin to understand why there would be any debate about this. I mean, unless "self driving" is only driver assistance, e.g. to provide a little extra safety, but the driver is still driving. In that case, maybe.
When one designs a control system, what you do is you ask yourself, how would I do this manually? And not JUST the easy way, when everything works. But it has to include contingencies. An example here would be, when you drive in a snow-covered road, how do you manage to navigate without seeing the road markings? That needs to be worked out. Or even, how do I react if suddenly I lose visibility, e.g. in hard rain or snow?
And then you write algorithms that duplicate the human behavior, only hopefully without the panic, slow reflexes, or incompetence.
The only way I would accept a notion that V2I isn't imperative is if someone can make a convincing argument that maps, traffic lights, pavement markings, and road signs, are useless. That all you need to drive a car is to see the cars around you. Make that argument, or tell me that self-driving only really means "driver assistance," and maybe I'll be convinced.
Not sure why this side-debate gets so much attention either, actually. Use of cellular makes a lot of good sense, especially for long range V2I or V2V. Like, download the map. Or, talk to your buddies back home, or out on the road 50 miles ahead.
LTE doesn't make as much sense to show you the lanes or road edge, where you are traveling this instant. For that, you want a more direct and short-range link, with information that is valid right now. If you think this info is useless, then ask yourself how you drive on snow-covered roads (or any other situation where road markings aren't visible). What cues do you use? Provide those cues somehow, if not with an RF link then with some other scheme, and you're good to go. Or, how do you navigate through a construction zone?
It will take a mix of link types and/or on-board sensors to get there. LTE is only one piece of the puzzle, seems to me.
The article states that we need autonomy on the cars because of lesser accidents, traffic jams and better fuel economy? That last point draws my attention because I've always heard that having a stick shift is better for fuel economy than the automatic shift. Is this statement still truth? If so, what is the reason behind it?
I simply thought us humans would have even a better feeling than computers on how to step on the pedal and when to switch gears. Isn't it the gut feeling that which makes a driver better than others? Oh... just had a vision! There will come the day in which a autonomous vehicle be raced against a human driven one! First it was chess. then Jeopardy... then Nascar?
Human Errors are today considered the major cause of Accidents. Automation and/or Driverless Vehicles are of discussions today to overcome the human errors, the studies are comming up with the fact feagures stating that Automation will be adding to the safety, but once the Automation and/or Driverless Vehicles will start taking the critical decision "Machine Errors/Failures" will become the major cause of Accidents. And these will not be under anyone's control then after.