Uncategorized

Imagine a world where your car listens to what other cars are saying about road conditions and tells the cars behind you what’s happening up ahead.

This technology might sound a little far-fetched right now, but the National Highway Traffic Safety Administration wants to know your input on it. Manufacturers might be required to implement vehicle-to-vehicle (V2V) crash avoidance technology on new vehicles as soon as 2020.

The problem that V2V aims to solve is the No. 1 cause of accidents: driver error. I recently wrote about the many stats that indicate that we (humans) tend to be dangerous drivers. In 2014, more than 3,000 people were killed in distracted driving accidents alone, not to mention tens of thousands more in alcohol-related crashes. The premise behind V2V is that no matter what mechanical safety systems we use, drivers will still cause accidents if they fail to slam on the brakes quickly enough. Anti-lock brakes won’t help if someone is about to t-bone you and you can’t see them coming.

Practically speaking, with V2V, cars can listen to other cars to determine if there is a potential hazard and warn the driver to take immediate action. Of note, V2V communications is a separate technology than highly-automated vehicles (fully autonomous vehicles). The former notifies human drivers to take action and the latter wholly operates the vehicle.

NHTSA has published a four-page summary, along with 327 pages of research on V2V communications. It’s packed with plenty of analysis into the technology, its costs, its adoption, various studies, and the anticipated reduction in automotive-related injuries and fatalities.

There is great promise with V2V communications, but also several areas of concern. First is the efficacy of it, second is the lack of infrastructure, and third is the overall cost. Let’s take a look at each.

Efficacy

On a non-technical level, would a V2V system be effective? Drivers today already are taught to check their blind spot, keep their hands at ten and two on the steering wheel, and always signal before changing lanes. Despite legally proving that we understand the rules of the road (passing your driver’s test), we still tend to make errors intentionally or not, resulting in more than 900,000 collisions on U.S. roads in 2014 alone.

The V2V communications system will potentially create pathways to expand into autonomous driving capabilities, but the NHTSA proposal doesn’t say so explicitly. The stated outcome of this crash-avoidance technology is to notify the driver to take immediate action. Frankly, I don’t believe one more beep, tone, flashing light, or vibration in the steering wheel will compel drivers to make the correct, split-second decision. I’m proof.

I have a radar detector, and anyone who has owned one will tell you that they pick up numerous false-positives. These false-positive alerts are triggered by driving in front of a store with automated doors and certain vehicles with their proprietary collision-avoidance tech onboard. These false-positive alerts chip away one by one from the driver’s trust in the $200 device sitting on their dashboard. Eventually, the driver no longer responds favorably when a Ka-band (a popular RF wavelength used in traffic enforcement) alert triggers because of the dozens of false alerts they’ve experienced in the past. Additional research is necessary to prevent V2V-enabled vehicles from responding to false-positives that erode the trust of the operators. That is, how will people respond when the false alerts compel drivers to take evasive action that makes them vulnerable to more accidents?

I see V2V communications offering the most value to drivers who endure hazardous weather conditions such dust storms and blizzards. Regardless, the V2V proposal genuinely tests the assumption that drivers only need to be made aware before an accident occurs, but I’m not able to find convincing evidence that this technology would address the actual causes of human error in a meaningful way, because it would depend on the driver to take action.

In essence, you can’t always trust people to make the right decision even if they are warned something bad might happen.

Infrastructure

Currently, a proof-of-concept demonstrating the promise of the V2V technology doesn’t exist. We don’t yet understand the impact of dozens or hundreds of cars on the roadways all communicating with each other. Let’s dive deeper.

In the proposal, cars would need to communicate with each other over wireless signals, not unlike the wireless router in your house that delivers internet access all over your home. The problem with this medium is that it is prone to radio interference and attackers might want to compromise it for fun, profit or both. To support such a network of connected cars, we will need to equip cars with a cellular data connection. Our cell towers would all need to be upgraded to ensure there isn’t network congestion. And on top of that, we’d still have to develop and test a system for ensuring updates are delivered safely and securely for all vehicles.

If you don’t think your vehicle could be used as a proxy for other attacks, think again. Security cameras, digital video recorders (DVRs), and cloud-based thermostats were utilized to attack core services that operate the internet. On one hand, your vehicle might be able to alert you when you approach another too closely, on the other, your vehicle could be just another node in a botnet when they maintain high-speed data connectivity. It’s not a matter of “if,” but “when” these vulnerabilities all come together to introduce more security risks than benefits.

As a country, we would have to be willing to invest an inordinate amount of resources to ensure vehicles maintain reliable connectivity, while still making room for additional expansions in wireless infrastructure. Ultimately, every aspect of the V2V communications interface would need to be tested and proven in real-world conditions to ensure maximum security and reliability—and when issues are identified, deliver updates and patches to fix them regularly.

Cost

The estimates are rough, but NHTSA predicts between $1.1 and $6.4 billion per year to implement this initiative and gradually decrease to an annual $1.1 billion expenditure by 2058. Over the course of 40 years, the program would cost taxpayers and manufacturers an estimated $80 billion.

That $80 billion could be a relative bargain considering that damages involving all light-vehicle crashes totals about $274 billion annually. After 40 years of investment, the NHTSA proposal predicts we’d only reduce three percent of the automotive fatalities seen in 2014. The cost is a pretty big pill to swallow over the next four decades especially given the modest improvements in vehicle safety.

Here’s the better solution

We’re at a point where 33 automotive leaders and a handful of tech startups have made significant investments in autonomous driving. I’d contend that even though a driverless car is more difficult than adding a V2V warning system, this leap in technology will be what’s needed to dramatically improve automotive safety. Ridesharing has increased in popularity, and once autonomous vehicles become readily available, vehicle sales may decrease when people no longer desire vehicle ownership — especially when they realize the cost of ownership has increased.

In essence, I believe the pursuit for driving to go completely autonomous by entrepreneurs will outpace the speed (and cost) of automotive technology enhancements required by the U.S. government.

It may be possible that V2V-inspired technology may be developed voluntarily by manufacturers because of consumer demand. I would love for my brakes to react quicker than I do during my daily commute, but I’d love it even more if I didn’t have a commute. We all should work toward advancing meaningful developments in automotive safety and mobility, instead of merely ‘checking the box’ for one more beep or light that drivers will tune out. I don’t have much confidence that these standards would be adopted anytime soon.

In the meantime, I’m holding out for the federal autonomous driving standards to be finalized so we can take a giant leap toward a safer automotive world.

Image credit: NHTSA

Originally published on the Local Motors blog.

Would you trust being driven in a self-driving vehicle?

One of the concerns that some of you have expressed on our Facebook page [1, 2, 3] about self-driving vehicles such as Olli revolve around vehicle safety and more specifically, ethical dilemmas when an autonomous vehicle has to make a split-second decision to maintain overall vehicle safety. I want to preface that my opinions here are designed to get us thinking critically about some of these concerns so that we can improve safety and survivability for all scenarios of autonomous driving.

Researchers from MIT have set out to learn about people’s preferences for how they would expect an autonomous vehicle to behave in a catastrophic situation. The Moral Machineasks the user to choose between two different scenarios faced by an autonomously operated vehicle. The complex outcome of these decisions and why the user has made them is daunting even to fantasize about when it comes to choosing between life and death for passengers or pedestrians. After judging 13 scenarios, you’re presented with your results and how they deviated from the average by other participants.

I don’t deny that this survey is a neat tool to measure rationalizations for who should be the beneficiary of crash-avoidance maneuvers. This survey functions as a powerful tool to spark a great conversation, which is one of the project’s stated purposes. However, there are two reasons I don’t subscribe to the presumption that every outcome results in one or more fatalities. First, I don’t think autonomous vehicles need to operate faster than their onboard technology can reasonably keep up with. Second, I don’t believe every potentially hazardous scenario has to result in one or more fatalities.

Low-speed autonomous vehicles are the place to start

It’s not necessary for driverless vehicles to go 0-60 in 2.2 seconds or to travel more than 75 MPH. Autonomous driving technology is rather nascent; it could perhaps keep up with a driving environment moving at 35 MPH or less. As the technology is developed and becomes proven, manufacturers can certify and ratchet up the speeds as needed. The industry has already seen success in the form of adopting automatic emergency brakingstandards by 2022.

When a vehicle moves at a slower speed, it’s able to make better decisions over a longer period. In situations where an impact is inevitable, the vehicle can take evasive action sooner to protect both passengers and third parties alike. Think about how many objects you can remember when you’re cruising down the highway at 75 MPH versus an urban environment at 35 MPH. Consider the fact that autonomous vehicles are designed to maintain full 360° visibility, unlike human’s near-180° field of view.

As with any new technology, self-driving vehicles will have defects as they’re being developed. We can’t dismiss potential risks with driverless vehicles. However, we can engineer them to be safer, smarter, and more consistent than human drivers. (Not to mention that with additive manufacturing, we can make cost-effective iterative improvements for future vehicles, too.) And while some legitimate security concerns may linger, security researchers are eager to test and propose updates to strengthen vehicle security to prevent unwanted vulnerabilities.

Human drivers are prone to error, sometimes highly so

There’s a common perception that human drivers operating vehicles are innately better (or more ethically sound) than autonomous vehicles. Hard data begs to differ.

In 2014, NHTSA reported that there were 3,179 people killed in the United States from vehicle crashes involving distracted drivers, representing 10 percent of all the motor vehicle fatalities seen that year. Out of those fatalities, 520 non-occupants such as pedestrians and bicyclists that have been killed as a result of distracted driving. It’s not only the deaths that count; NHTSA estimated 431,000 people were injured due to distracted driving.

These figures make a strong case that human drivers are not infallible in their driving habits. We can create legislation to mandate the use of hands-free cell phones while driving, and public service announcements to discourage phone usage while driving, but it isn’t making a dent in our behaviors. I didn’t even touch on intoxicated driving, either.

We have a problem and we ought to confront these with practical solutions that save lives.

Developing autonomous vehicle technology is ethical

For us to hold autonomous vehicles to an ethical standard, we have to understand that ethics are a set of principles defined by a moral code. Are we being intellectually and morally honest by impeding the progress of developing autonomous vehicles? I don’t think we should be blind to the risks, but we have a very real problem to solve. We owe it to ourselves to improve the odds of survival for more than 30,000 people every year in the United States.

Innovation in driverless cars is legitimately disruptive, and that goes against what we’re used to in transportation. We’ve become cozy with incremental progress in today’s cars. Add an airbag here, add another airbag there, and add blindspot alerts and we have only seen marginal safety improvements to show for it. We have the opportunity before us to remove the largest causal factor in automotive crashes: us. This is a scary thought for some because it would eventually relegate human-controlled driving to closed-course tracks and become a hobby, not a rite of passage into adulthood.

One ethical aspect of the industry that needs improvement is how we describe and set expectations in the eyes of consumers. Additionally, we need to do a better job of educating consumers about autonomous vehicle technology so they are informed about how the vehicle operates and what their responsibilities are. I’m confident that once consumers understand the immediate and long-term benefits of self-driving cars, they’ll give them the green light.

If we avoided developing new transportation solutions for fear of the potential risks, we’d still get around riding on the back of exactly one horsepower. The moral cost for not positioning autonomous vehicles for success is something that not too many of us think about — but maybe you have now.

Do you support or oppose the idea of having autonomous vehicles on the streets? I’d love to learn why you think so in the comments below.