2 ms·

Automated driving is the single biggest opportunity in robotics in the next decade.

axiom 19y ago
I'm not sure I agree with that. I think the potential is definitely there, but the regulatory burden is just too massive. The potential impact of robotics on manufacturing and transportation is absolutely staggering, and could probably be measured in the trillions of dollars (speaking of just the next 10-15 years.) However, I just don't see it happening there. The markets are just too heavily regulated and there are too many competing political interests. It's next to impossible for a new company to break into these markets, even with unlimited funding. If the robotics revolution(tm) is going to happen anywhere, it's going to come from either the consumer market or a tiny niche market that nobody's ever heard of, and then spread from there.
- ivankirigin 19y ago
  Car keys. They give the decision making power to the driver, and are what would shield the car maker from a huge regulatory burden. Any autonomous action will be dictated by the driver and the driver will be given explicit instructions to continue paying attention to the road. The same is true for today's cruise control. Also, the consumer market is for low end processing, not advanced autonomy. You won't get a revolution there.
  - axiom 19y ago
    I'm not claiming that autonomous driving won't happen or won't be a massive industry in the near future (30-40 years.) My feeling is just that it won't happen there first. Here's a couple of reasons: 1. It's a very complicated problem compared to others that could have a similarly large impact. 2. The cost of a mistake made by the robot is quite likely to kill someone. This isn't just a problem as far as lawsuits go, but it's a problem because as soon as the first accident occurs you're going to get a barrage of regulations (and quite possibly bans) on autonomous driving. 3. As soon as you can build a robot that can clean your floors, walk your dog, and bring you a beer when you want it, you're going to sell a million units in a year. Just look at the success of a fairly simple device like the Roomba. So my thinking is that autonomous dirving will be one of the later things happen as compared to other adoptions of robotics. Edit: this just reminded me of something. Not too long ago the University of Toronto solar car team was driving their solar car down the 401 highway. The driver lost control and ended up getting killed. Within a month all solar cars were banned from driving on public roads in Ontario.
    - ivankirigin 19y ago
      My original comment is about the business opportunity. Robots that can operate independently in a particular environment will be deployed in the next decade or two. The problem is that lots of environments aren't standard, e.g. a particular manufacturing plant might require a niche robot. Freeway driving is very standard. Hard, but achievable. I might be wrong on the time frame, but auto bots make up a very big single application. That it is why the opportunity is huge.
      - axiom 19y ago
        I completely agree with you there.
rms 19y ago
It's been possible for a while, see CMU's pioneering Navlab, the final Navlab car drove across the country. But it's still much to difficult to actually implement something like this, because of the regulatory burden and the burden of perfection. Now no one is funding self-driving cars outside of the DARPA competition which is for something very different.
- ivankirigin 19y ago
  I'm pretty familiar with automated driving, and see a clear path for incrementatl deployment over a few years. It wouldn't start with full automation. It starts with gps/ins/lidar positioning, camera & lidar based car and obstacle detection, camera based lane finding, etc. and built up from there. As I mention in another comment, the regulatory burden is exaggerated considering the human driver still in the vehicle for the early stages of incremental deployment.
  - rms 19y ago
    OK, you're right, Navlab was doing too much at once for that technology to ever be deployed, but it was still important work for the incremental improvements we are seeing, like the auto-parallel parking functions... with everything computerized in a car anyways, it really shouldn't cost too much to implement things like this even on entry level cars.

axiom 19y ago

I'm not sure I agree with that. I think the potential is definitely there, but the regulatory burden is just too massive. The potential impact of robotics on manufacturing and transportation is absolutely staggering, and could probably be measured in the trillions of dollars (speaking of just the next 10-15 years.) However, I just don't see it happening there. The markets are just too heavily regulated and there are too many competing political interests. It's next to impossible for a new company to break into these markets, even with unlimited funding. If the robotics revolution(tm) is going to happen anywhere, it's going to come from either the consumer market or a tiny niche market that nobody's ever heard of, and then spread from there.

ivankirigin 19y ago
Car keys. They give the decision making power to the driver, and are what would shield the car maker from a huge regulatory burden. Any autonomous action will be dictated by the driver and the driver will be given explicit instructions to continue paying attention to the road. The same is true for today's cruise control. Also, the consumer market is for low end processing, not advanced autonomy. You won't get a revolution there.
- axiom 19y ago
  I'm not claiming that autonomous driving won't happen or won't be a massive industry in the near future (30-40 years.) My feeling is just that it won't happen there first. Here's a couple of reasons: 1. It's a very complicated problem compared to others that could have a similarly large impact. 2. The cost of a mistake made by the robot is quite likely to kill someone. This isn't just a problem as far as lawsuits go, but it's a problem because as soon as the first accident occurs you're going to get a barrage of regulations (and quite possibly bans) on autonomous driving. 3. As soon as you can build a robot that can clean your floors, walk your dog, and bring you a beer when you want it, you're going to sell a million units in a year. Just look at the success of a fairly simple device like the Roomba. So my thinking is that autonomous dirving will be one of the later things happen as compared to other adoptions of robotics. Edit: this just reminded me of something. Not too long ago the University of Toronto solar car team was driving their solar car down the 401 highway. The driver lost control and ended up getting killed. Within a month all solar cars were banned from driving on public roads in Ontario.
  - ivankirigin 19y ago
    My original comment is about the business opportunity. Robots that can operate independently in a particular environment will be deployed in the next decade or two. The problem is that lots of environments aren't standard, e.g. a particular manufacturing plant might require a niche robot. Freeway driving is very standard. Hard, but achievable. I might be wrong on the time frame, but auto bots make up a very big single application. That it is why the opportunity is huge.
    - axiom 19y ago
      I completely agree with you there.

rms 19y ago

It's been possible for a while, see CMU's pioneering Navlab, the final Navlab car drove across the country. But it's still much to difficult to actually implement something like this, because of the regulatory burden and the burden of perfection. Now no one is funding self-driving cars outside of the DARPA competition which is for something very different.

ivankirigin 19y ago
I'm pretty familiar with automated driving, and see a clear path for incrementatl deployment over a few years. It wouldn't start with full automation. It starts with gps/ins/lidar positioning, camera & lidar based car and obstacle detection, camera based lane finding, etc. and built up from there. As I mention in another comment, the regulatory burden is exaggerated considering the human driver still in the vehicle for the early stages of incremental deployment.
- rms 19y ago
  OK, you're right, Navlab was doing too much at once for that technology to ever be deployed, but it was still important work for the incremental improvements we are seeing, like the auto-parallel parking functions... with everything computerized in a car anyways, it really shouldn't cost too much to implement things like this even on entry level cars.