Just heard that California is about to start working on formal regulations for robot cars to travel their roads which is the second state to regulate these autonomous machines, the first was Nevada. At the moment the legislation signed into law requires CA to draft regulations for these vehicles by January 1, 2015.
I suppose being in the IT industry this shouldn’t be a surprise to me or anyone else. Google has been running autonomously driven vehicles for over 300K miles now.
But it always seems a bit jarring when something like this goes from testing to production, seems almost Jetson like. I remember seeing a video of something like this from Bell Labs/GM Labs or somebody like that when they were talking about the future way back in the 60s of last century. Gosh only 50 years later and its almost here.
DARPA Grand Challenges spurred it on
Of course it all started probably in the late 70s when AI was just firing up. But robot cars seemed to really take off when DARPA, back in 2004 wanted to push the technology to develop a autonomous vehicle for the DOD. They funded a and created the DARPA Grand Challenge.
In 2004 the requirements were to drive over 150 miles (240 km) in and around the Mojave desert in southwestern USA. In that first year, none of the vehicles managed to finish the distance. Over the next few years, the course got more difficult, the prize money increased, and the vehicles got a lot smarter.
In 2005 DARPA grand challenge once again a rural setting, 5 vehicles finished the course 1 from Stanford, 2 from Carnegie Mellon (CMU), 1 from Oshkosh Trucking, and the other 1 from Gray’s Insurance Company. At first I thought an insurance company, then it hit me maybe there’s a connection to auto insurance.
DARPA’s next challenge for 2007 was for an urban driving environment but this time DARPA providing research funding to a select group as well as larger prize to any winners. Six teams were able to finish the urban challenge, 1 each from CMU, Stanford, Virginia Tech, MIT, University of Pennsylvania & Lehigh University and Cornell University. That was the last DARPA challenge for autonomous vehicles, seems they had what they wanted.
Google’s streetview helped
Sometime around 2010, Google started working withg self-driving cars to provide some of the streetview shots they needed. Shortly thereafter they had logged ~140K miles with them. Fast forward a couple of years and Google’s Sergey Brin was claiming that people will be driving in robotic cars in 5 years. To get their self-driving cars up and running they hired the leaders of both the CMU and Stanford teams as well as somebody who worked on the first autonomous motorcycle which ran in the Urban Challenge.
For all of the 300K miles they currently have logged, the cars were manned by a safety driver and a software engineer in the car, just for safety reasons. Also, local police were notified that the car would be in their area. Before the autonomous car took off another car, this one driven by a human, was sent out to map out the route in detail including all traffic signs, signals, lane markers, etc. This was then up(?) loaded to the self-driving car which followed the same exact route.
I couldn’t find and detailed hardware list but Google’s blog post on the start of the project indicated computers (maybe 2 for HA), multiple cameras, infrared sensors, laser rangefinders, radar, and probably multiple servos (gear shift, steering, accelerator and brake pedals), all fitted to Toyota Prius cars. Although the servos may no longer be as necessary as many new cars, use drive by wire for some of these function.
I could imagine quite a few ways to monetize self-driving, robotic cars:
- License the service to the major auto and truck manufacturers around the world, with the additional hardware either supplied as a car/truck option (probably at first) or provided on all cars/trucks (probably a ways down the line).
- Cars/trucks would need computer screens for the driving console as well as probably for entertainment. Possibly advertisements on these screens could be used to offset some of the licensing/hardware costs.
- Insurance companies may wish to subsidize the cost of the system. Especially, if the cars could reduce accidents, it would then have a positive ROI, just for accident reduction alone, let alone saving lives.
- In the car internet would need to be more available (see below). This would no doubt be based on 4G or whatever the next cellular technology comes along. Maybe the mobile phone companies would want to help subsidize this service, like they do for phones, if you had to sign a contract for a couple of years. I am thinking the detailed maps required for self-driving might require a more bandwidth than Google Maps does today, which could help chew up those bandwidth limits.
- With all these sensors, it’s quite possible that self-driving cars, when being driven by humans, could be used to map new routes. If you elected to provide these sorts of services then maybe one could also get something of a kickback.
I assume the robotic cars need Internet access but nothing I read says for sure. Maybe they could get by without Internet access if they just used manual driving mode for those sections of travel which lacked Internet Perhaps, the cars could download the route before it went into self-driving mode and that way if you kept to the plan you would be ok.
Other uses of robotic cars
Of course with all these Internet enabled cars, tollways and city centers could readily establish new congestion based pricing. Police could potentially override a car and cause it to pull over, automatically without the driver being able to stop it. Traffic data would be much more available, more detailed, and more real time than it is already. All these additional services could help to offset the cost of the HW and licensing of the self-driving service.
The original reason for the DARPA grand challenge was to provide a way to get troops and/or equipment from one place to another without soldiers having to drive the whole way there. Today, this is still a dream but if self-driving cars become a reality in 5 years or so, I would think the DOD could have something deployed before then.
If the self-driving car maps require more detailed information than today’s GPS maps, there’s probably a storage angle here both in car and at some centralized data center(s) located around a country. If the cars could be also used to map new routes, perhaps even a skosh more storage would be required in car.
Just imagine driving cross country and being able to sleep most of the way, all by yourself with your self-driving car. Now if they could only make a port-a-potty that would fit inside a sedan I would be all set to go…, literally 🙂