3 ms·

Thanks to a number of wars and billions being pumped into the UAV market, the likes of Raytheon, Thales Optronics, FLIR Systems, ITT, BAE, IAI, and probably a h

• modeless 15y ago
  UAV control has basically nothing to do with self-driving cars. It's a completely different problem. The only piece that's shared is the GPS/IMU, and it's a small piece. And if you think "the state machine for how to behave on paved roads" is the hard part, you're delusional. "Environmental sensing accurate and reliable enough" isn't the problem either. Although fancy lasers and radars can help, we have had cameras, microphones, and accelerometers superior to human eyes, ears, and vestibular system for years now. The missing piece is the software, which is exactly the piece Google is working on.
  • dsl 15y ago
    Sorry for the confusing use of the term. UAV means unmanned autonomous vehicles. UAV in reference to aerial platforms is mostly the result of marketing/media. Even the early Daimler-Benz projects of the 90's had fully functional software stacks that had implemented all the required decision making logic of driving. While it is true that we have cameras and microphones that work better than our own eyes and ears, I don't think anyone would dispute the fact that the sensory processing isn't to human levels yet. The advancements that have been made are in different types of sensing devices that eliminate the need for heavy post-processing of data (image pattern recognition, etc) in software. I do give credit to Google for making it happen, both as an integrator, and in lobbying efforts. But there is a lot of standing on the shoulders of giants, in comparison to pure software fields where they have made huge industry disrupting shifts (like BigTable, GFS, etc).
    • modeless 15y ago
      I don't think anyone would dispute the fact that the sensory processing isn't to human levels yet And that's the exact problem Google is working on, and where all previous efforts have failed miserably, no matter how good their "rules of the road" state machines were in theory. Saying Daimler-Benz solved autonomous driving years ago is like saying SHRDLU solved natural language processing. The advancements that have been made are in different types of sensing devices that eliminate the need for heavy post-processing of data There are exactly two new kinds of sensing devices used on the Google cars AFAIK: small automotive radars and the Velodyne spinning LIDAR. The latter owes its existence not to giant defense contractors or car manufacturers but to a small loudspeaker company, of all things. Furthermore, it absolutely does not eliminate the need for heavy post-processing of data.
    • Retric 15y ago
      I think we could start putting self driving cars on the road today that would lower both the number of fatality's and accidents. We don't need more processing power, better sensors, or lower costs. What we need is slightly better software and the willingness to put it into production. After-all the 'worst case' in a car is basically solved 99.9% of the time by staying in the correct lane, obeying stop lights / signs, speed limits, and simply hitting the breaks if your going to hit something. Sure, you could improve on that, but get that to work reliably and your already doing better than human drivers, who get distracted, drunk, tired, impatient, angry, and just plain overwhelmed.
• ghshephard 15y ago
  Well, I guess "autonomous vehicles" were a solved problem the same way "portable music devices" or "smart phones" or "Tablet Computers" were a solved problem. But it takes _someone_ to push the concept over and move the category forward. In all three categories (particularly portable music devices) - I would argue that Apple didn't so much as make the technology possible, as it did make them _popular_. Likewise, with a autonomous vehicle - it isn't so much building a car that can drive itself - it's making a car that can drive itself _that I can buy_ that I get excited about. Of course, Google has yet to deliver on the part about me buying them - but can you identify any other company that has done so much to get them driving on roads _with other people_?
• marshray 15y ago
  Could your "solved problem" vehicles reliably tell the difference between a kid running into the street and a empty garbage bag blown in the wind?
  • dsl 15y ago
    If you designed the system, would you handle them any differently? A buddy of mine, when he was a kid, used to tie toilet paper across the road at windshield height then hide in the bushes. A car would come along at ~30 mph, see it, then come to a screeching halt. Some would get out and tear it down, some would stomp the gas. Point is, every human stopped for a single strand of toilet paper blocking the road.
    • marshray 15y ago
      Perhaps on a 30 mph residential street in light traffic. Screeching halts have a different effect at higher speeds in denser traffic. People would be less likely to slam on the brakes for a trash bag on the freeway. But either way it is a dangerous, controversial, split second, human risk calculation. I can't say how I would handle it in advance, much less how a robot car should be programmed to handle ambiguous inputs at highway speed.
• russell 15y ago
  I would posit that the secret sauce is the integration of Google maps and street views. The driverless car isnt just reacting to its sensor inputs, it has a pretty good idea of what is coming up including potential problems.
• hexagonc 15y ago
  "The difference between theory and practice is that, in theory, there is no difference but in practice, there is" -- numerous possible sources