Google car accident: What does this mean for autonomous vehicles?

On Valentine’s Day this year, in Mountainview, CA (Where Google’s Global HQ resides), Google’s autonomous test car in that area had a minor fender-bender when it struck a bus at 2 mph. You can see the DMV report here.

So what? There was no one inside to say, “A bee flew in my window”, “I spilled hot coffee on my lap”, “I was texting”, “The sun was in my eyes”—-Robots don’t have these problems!

Yes, the accident was partially due to software programming decision-making and No, this is not the end of autonomous vehicles. As engineers, we do our best to make a robust product, but we know that nothing is perfect and in this case the software needs to be refined and refined and refined again like any other design ever made before or that will be made in the future.

Just check out the Google self-driving car reports. Their autonomous vehicle is far safer than any car with an imperfect human driver—yes, we are imperfect as well as software is imperfect. We text, we put makeup on in the mirror, we answer our cell phones while driving, we get distracted by a myriad of different things that distract a human driver but will not affect a robot! (I facetiously call it a robot for lack of any better term in my mind)

So move on Google, correct the problem as best you can and get us to the point where autonomous vehicles will be far safer than we are now on the roads. (Robots don’t get Road Rage, don’t drink and drive and definitely don’t speed or ignore road signs and signals like we humans do either)

See how the Google car works here. See Figure 1 for a good block diagram on a typical autonomous vehicle.

Figure 1

A typical Autonomous vehicle functional diagram (Image courtesy of Texas Instruments in Reference 1)

A typical Autonomous vehicle functional diagram (Image courtesy of Texas Instruments in Reference 1)

It was back in the 1920s that all of this started when Francis P. Houdina, an EE in the US Army set up a 1926 Chandler with a mounted receiving antenna for remote control via another car following it close behind with a transmitter. The signals drove some small electric motors to control the movement and direction of the car.

And now we are on a journey toward safer driving eventually and, of course, there are still many refinements needed in software and improved hardware, but probably the most difficult challenge will be in security. Hackers and Black Hats will want to get into these systems just because they can. They want to show their peers how creative and talented they are. For this reason, we must be vigilant and install strong security measures in this type of vehicle.

The final hurdle will be the formal government and safety standards bodies which could be the longest delay, but if we have secure, mature software, this will inevitably speed up this process.

What are your thoughts on this subject?


1 Scalable-electronics-driving-autonomous-vehicle-technologies

2 An Open Approach to Autonomous Vehicles, S. Kato, E. Takeuchi, Y. Ishiguro, K. Takeda, Nagoya University; T. Hamada, Nagasaki University, IEEE Computer Society, 2015

More articles on this topic




8 comments on “Google car accident: What does this mean for autonomous vehicles?

  1. didymus7
    March 9, 2016

    I think that we will have no choice in the matter.  At some point, the US Govt, with its poor grasp of technology will be bamboozled by a company wanting a customer with deep pockets who will convince some senators that this technology will solve every highway and traffic problem.  Then we will be forced to use an autonomous system that will closely resemble municiple bus service by spending 2 to 4x more time in commuting.  Of course, congress will continue to have their own cars…..

  2. johnspeth
    March 9, 2016

    You stated Google's autonomous vehicle is far safer than any car with an imperfect human driver.  That's quite a claim with only a million test miles logged under probable favorable conditions.  Where's the data to support that claim?

  3. jimfordbroadcom
    March 9, 2016

    You know, the way humans drive around here (Southern California), I say bring on the robot vehicles!  Can't be any worse.

  4. Steve Taranovich
    March 9, 2016

    @johnspeth, Thanks for your question.

    Of course only a million miles under good conditions is not enough to support my claim—I agree. I still stand by my claim because, As I say in the article:

    “A bee flew in my window”, “I spilled hot coffee on my lap”, “I was texting”, “The sun was in my eyes”—-Robots don't have these problems!

    We text, we put makeup on in the mirror, we answer our cell phones while driving, we get distracted by a myriad of different things that distract a human driver but will not affect a robot!

    Just add up all the accidents that happened from just the above things human beings experience as distractions, knowing that the list goes on with far more distractions than these, and you will see that an autonomous vehicle will be avoiding these types of accidents.

    That's not to say that we are there yet to begin riding in autonomous vehicles because we have more work to do.

    Also, I am an advocate for having a fast and convenient means for the human driver to over ride this “advanced auto-pilot” in an instant should they decide to do so. I still think that for most of us, a human being's reasoning will be far superior than a robot, but the distractions issue would make a machine better than a human—and that would prevent untold numbers of accidents

  5. tpcipri
    March 9, 2016

    It is a matter of choice:

    For a good many people an autonomous car may be a good choice as long as it is their choice to make. It will have all the benefits of mass transportation with less delays. So what if the commute takes a few minutes longer if you can be productive in that time and you are able to relax while doing it. Personally, I'm a driver and driving is something I like doing and until I can't safely drive a car I will… well continue riding a motorcycle. 

    Positive Benefits:

    Unlike the drug commercials there will be positive side-effects to the development of autonomous cars that will make driving safer long before these cars become a reality. Right now we are seeing upscale cars with driver parking assist, car paging and most importantly road warnings. What if cars didn't start when the driver is unlicensed, incapacitated by drugs or alcohol or isn't insured? What if a radio signal from traffic lights prevented cars from speeding up when the light turns yellow, or knowing the traffic conditions in 360 degrees prevents the driver from slamming on his brakes when it is unsafe to do so? What if cars became a live network on the road and warned each other of dangers beyond the drivers vision and, for that matter, on-board sensors range. What if traffic lights and signs sent out notifications as vehicles approached warning the drivers that they will be required to stop? What if the car detected the driver was erratic and alerted him or her to the fact? Those are all passive actions that would improve highway safety significantly. 

    There is a lot more that can be said  about this subject and I'm sure a lot more will be said. The key point being made is cars can be made a lot safer without being self-driving. Instead of forcing self-driving cars on the general public, let's reap the benefits from the technology self-driving cars provide and, if someone prefers to be driven, they can still buy the much more expensive self-driving car and enjoy its benefits! 


  6. tpcipri
    March 9, 2016

    I drive in the same areas as the Google cars. You cannot underestimate how hazardous it is to drive in this area. The “Expressways” have car pool lanes on the right with traffic moving at up to 50 mph while the other lanes are at a dead stop. All entering and exiting traffic has to cross through the car pool lanes. Consequently there are cross-traffic spots where it is completely unsafe to use the car pool lanes and in a five mile commute it is not uncommon to see two or three multi-car accidents in a single week. The most perilous of these is north-bound Lawrence Expressway  which crosses over a railroad track and comes up to a traffic light. Cars in the carpool lane come over the crest of the bridge going 50 mph (or faster) only to find that the lane is backed up with stopped traffic and there is no time to stop. On the next block is the exchange with Central Expressway where, sorry about the expression, idiots pull into the 50mph car pool lane from a dead stop. Knowing this I slow down as I approach these blind hazards and can't tell you how many times I probably prevented high speed collisions. Unfortunately, not enough people drive like I do so the accidents still happen frequently. Central Expressway has merge lanes that are marked as carpool lanes but they come and go which is a hazard unto itself but I infrequently travel that route so can't say if it is as bad as Lawrence Exp in terms of preventable accidents. Eliminating these foolish carpool lanes would relieve a lot of congestion and accidents but cars that are networked together with real time data about what lies ahead in this forward looking blind spot would prevent many of these accidents since distracted drivers are unlikely to see warning signs and lights.

    You also have to consider that right alongside Google there are Microsoft, Oracle, NASA, Linked-In and just down the road Lockheed, Apple and Cisco and the entire city of San Jose – California's third largest city and is less than 10 miles from Google. At this point it is probably pointless to say that traffic congestion is extremely dense between 6AM to 10AM and 3PM to 8PM for commuters and between 11AM to 2PM for lunch time traffic. Downtown Mt. View has just about every type of restraunt and is very popular. This means that there are at best 4 hours between 6AM to 10PM when traffic isn't heavily congested. We may have fair weather but definately not fair traffic.

  7. Steve Taranovich
    March 9, 2016

    @tpcipri—-Don't get me wrong, I am DEFINITELY not an advocate of forcing autonomous vehicles upon people. I look at it like driving an Electric Vehicle vs. a Gasoline or Diesel-powered vehicle choice.

  8. araasch
    March 9, 2016

    Will autonomous vehicles be able to instantly decide that a tumbleweed does not pose a danger but a cardboard box in the roadway may because it  could contain objects?  If the vehicle in back of the autonomous vehicle is being human driven, an accident may needlessly occur with the tumbleweed causing the autonomous vehicle to break unnecessarily.

    Black ice, standing water, loose rocks, fallen tree limbs, snow/slush between lanes are all things we encounter on the road.  How will the software deal with these hazards?

    I take little comfort in the attitude that the human is the backup for the automation, as the human will be lulled to sleep or otherwise deep in thought just when you need 100% performance from the human.  Good luck with that.

    Aircraft manufacturers have looked at aircraft accident causes and tried to eliminate those caused by pilot error through automation.  This has caused accidents when pilots did not understand that the automation was configured incorrectly, or disengaged.  It has also lead to the overall flying skill of pilots to decrease as hands-on flying is done less and less.

    In my opinion, it would be much better to focus on virtual reality as a way to avoid a great deal of car traffic through telecommunting instead of trying to automate the driving of cars.  A car that is not being driven gets an amazing MPG rating and causes no traffic jams, or wear and tear on roads and has a significantly smaller carbon footprint.


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.