Autonomous Cars Are a Long Way Off

Want to get an extended night’s sleep while your autonomous car drives you from home to work? Think again!

Last week I attended Automotive World’s three day conference. The second day was dedicated to Autonomous Cars. I learned that autonomous cars are so far out that it’s absolutely ridiculous what the media comes up with about these futuristic machines.

First off, the word “autonomous” is dangerous because cars may never fully independent of a driver. A better term is “automatic”. Partially automatic cars will be part of our future. There will always be some driver interaction that is required. Why is that?

Human’s are incredibly effective at driving. There is only 1 death for every 100 million miles driven. That is hard for a computer to beat. It will take billions miles driven by a computer in order for it to encounter the majority of road conditions and situations where it needs to make a critical decision. Humans have to analyze that data and make adjustments which is a lengthy process. Additionally, what do you do in the case where a crash is unavoidable? Say you have a split second to decide between crashing into an old lady crossing the street or a teenager in car, one will die, who do you choose? It’s philosophical decision that no one can make. A computer wouldn’t be any better at it. There are other scenarios we don’t know how to handle yet either such as changing destination mid-route, speeding, and alcohol. Should a drunk passenger be allowed to “operate” an autonomous car (if we eventually get there)? If your autonomous car decides that it’s better to run a red light than be rear-ended, can it make that decision? Would you get a ticket?

I mentioned this in my review of Connected Cars and the same applies to autonomous cars – how do we keep them updated? Modern day “connected cars” aren’t being updated. What happens when an older autonomous car is driving next to a newer autonomous car? There are no protocols that define how autonomous cars talk to each other. These need to be defined.

Security is a huge concern. How do you know that your autonomous car isn’t running malicious software or being controlled remotely? We’re starting to see this pop up in certain cars already. The problem is only getting worse as manufacturers don’t have a handle on software security. In general, the type of developers working on cars have limited experience developing with security, or with being “connected”, and are exposing vulnerabilities an elementary school student could find.

Even if we had autonomous cars today, you couldn’t afford it. No one will pay $90K for a $20K car with $70K worth of autonomous equipment in it. In order for it to become reasonably priced you have to consider our country’s economics… The average national income is $50K/yr which is $25/hr. If your commute is 20 minutes, twice a day, five days a week, fifty weeks a year (two weeks of vacation!) that means your driving time is worth just over $4k per year. If you keep your car for 5 years that’s $20K. And who’s to say you’ll be as productive in a car than you would be at your desk? The additional price of autonomous cars will have to come down to no more than $15K over non-autonomous cars before consumers will start considering them. Insurance cost won’t change much. The additional cost of the car will offset any insurance break you get from an autonomous car being “safer”.

Back to sleeping in your car. Did you know it takes 8 seconds for a distracted driver to become fully aware of the situation they are in when they chose to regain focus on driving? (eg. your on your cell phone and hang up) If you are sleeping and need to take over control of your car, how much time is needed for you to become fully aware in that case? It’s starting to sound like being autonomous is not a good idea at all, isn’t it?

Two ideas that I feel will help automated drivers are an increased usage of voice control and physical feedback (such as vibrating seats). These will help automated drivers stay aware of what’s going on around them even though they may be distracted.

Automated driving will come in increments, slowly. A good place to start is in parking lots and on the highway. The situations encountered in these scenarios severely limit the amount of work needed to be done and makes the concept realizable in the future. Once we see these things at work, our trust level will rise, and we will begin to see more and more automated vehicles. We’re talking 5 years out though… for the upper class who can afford it. Another 5 years after that it will become mainstream. Another 5 years after that most cars will have some level of automation. Not fully autonomous granted, but, partially automated.


Rush Frisby

Rush is a software architect who loves solving problems. You'll usually find him writing about security concepts and sharing solutions to problems he's faced.