The self-driving car has become a hot topic over the last several years. Many companies, including Google, believe this technology could do wonders for world transportation.
Self-driving cars won’t just be convenient; they’ll also be less expensive, more fuel efficient and safer. They might even turn long, boring commutes into an opportunity to relax, read a book or call in to a meeting.
But tomorrow’s transportation is not just about the self-driving car. The future will see networks of cars working together to keep passengers safe and deliver them to their destinations efficiently.
For that to happen, though, cars need a way to talk to each other.
Ready To Talk?
Wireless communication between autonomous vehicles has always been a topic of interest to researchers developing the car of tomorrow. Demonstrations like Google’s self-driving car, which doesn’t even include a steering wheel, are impressive – but they’re also lone projects built on a limited scale.
The problem facing researchers is no longer how to build an autonomous vehicle, as that’s already been accomplished. Instead, the problem is how to make an autonomous vehicle safe and reliable on today’s roads. Self-driving cars operating alone might provide their owners with convenience, but they won’t fully realize the efficiency, safety and cost benefits the autonomous vehicle can provide.
Those improvements can only be unlocked via an autonomous car network. No such network has been built, so opinions of what it might look like vary, but researchers are working to flesh out the idea.
The Mobility Transformation Center at MIT, for example, is pushing to make Ann Arbor (the school’s home town) a leader in automated motoring. Larry Burns, an engineering professor at the school, has turned to the animal kingdom for inspiration, pointing out that:
“Bees swarm. Geese flock. And they’re not running into each other.”
A swarm of bugs may seem an odd comparison to automated cars, but it’s indicative of the tight tolerances a network of autonomous cars could enable. A typical human driver, if not distracted, requires 215 milliseconds to react. That means a car moving at 100 kilometers per hour will travel about six meters (almost twenty feet) before the driver can even respond. Safe drivers often leave several car lengths between them and the vehicle in front of them because of this delay.
Radio waves, however, are almost instantaneous (at the distances automated cars operate), which means automated cars can theoretically operate safely with only a few feet between them. Suddenly the image of a swarm makes more sense; a network of autonomous cars would not look like today’s traffic but instead like a constant flow of vehicles moving organically, leaving spaces of a meter (and sometimes far less) between each car. At a glance, the movement might appear random, but it would actually be highly coordinated; you’d witness a channel of cars moving left, merging into gaps just centimeters larger than the cars themselves, if there’s an exit half a mile up the road.
But to simply say this will be made possible by radio waves is akin to stating “a wizard did it!” There are many different concepts of how a network of automated cars might work, and they generally work in two main categories.
The most obvious way to enable networks of automated vehicles is to have them speak to each other directly. From a technical perspective, this is relatively simple, and in fact leapfrogs from current collision avoidance technologies. Many luxury automobiles now include automated cruise control and low-speed automated breaking systems that operate using a variety of sensors. Add in a radio, and a standard through which vehicles can share data via radio, and presto! You’ve got a basic wireless network.
This has an appeal because it’s immediately usable and can operate with vehicles that are not automated. The National Highway Traffic and Safety Administration, the top regulatory body overseeing the roads in America, has already recommended the implementation of vehicle-to-vehicle (V2V) communication to prevent collisions. A report written by four NTSB researchers found that:
“…excluding drivers impaired by alcohol or drowsiness, these systems [V2V] deal with 81 percent of all-vehicle crashes involving unimpaired drivers.”
This means V2V systems could prevent the majority of automotive collisions if all vehicles implemented them.
A popular theoretical implementation of V2V is the “platoon” system. This idea, which has been around since at least 1993, involves groups of automated vehicles that come together to form a long, tightly-spaced line. This keeps the automated cars away from those which aren’t automated and provides aerodynamic benefits which reduce fuel consumption (with the exception of the lead car).
In this system virtually any type of wireless communications could work, as each vehicle in the platoon would only have to communicate with the one in front of it. Any number of modern wireless technologies (Volvo demonstrated a platoon using 802.11p WiFi) could operate reliably, as the short range of communication limits interference and reception issues. Even a momentary lapse in communication would not be disastrous, as each automated car need only match speed with the one before it. Erik Coelingh, an engineer with Volvo, told Phys.org that, “We [Volvo] believe platooning can be safer than normal driving today,” and elaborated that the automotive manufacturer is closely examining the most efficient – and safest – way to implement the idea.
V2V systems like platooning are a relatively simple way to implement autonomous vehicles, but the idea isn’t perfect. All V2V systems lack centralized hardware in charge of overall transportation. Platoons, for example, are efficient for the cars involved, but they don’t respond dynamically to traffic and can’t communicate with roadway infrastructure. If a platoon encounters heavy traffic it’ll simply slow down and follow the route determined by the lead car. There’s no way for V2V networks to “see” a traffic jam and calculate an alternative route, or predict the timing of the next three stoplights and adjust speed accordingly. The full potential efficiency of the automated vehicle can’t be realized with a larger and more complex system.
That efficiency can be enabled only if there’s a way to let autonomous cars interact not only with each other, but also with the environment, enabling the “swarm of bees” mentioned earlier. To do this, each car needs to be able to hook into a network that spans not only its immediate vicinity but a much wider area, perhaps as large as the entire city the vehicle is operating in. This kind of network is called vehicle-to-infrastructure, and it’s far more complex.
A German company is currently conducting a three-month trial of a V2I system called simTD which lets connected cars communicate with infrastructure elements. For example, a car with this system can speak with an upcoming traffic light and adjust its velocity to time its arrival with the light’s change. In doing so it decreases idle time, which improves fuel efficiency. The system can also warn a car and its occupants to upcoming road hazards by receiving data when another car skids or experiences loss of traction.
Even this rudimentary implementation of V2I enables safety and efficiency benefits, but the downside is complexity. A combination of WiFi, UMTS and GRPS (the latter two are cellular data standards) are used to provide constant communication with both infrastructure and other vehicles.
SimTD also uses vehicle-to-vehicle transmissions as a daisy chain to enable infrastructure communication if none of a vehicle’s radios can receive a signal. That’s a great idea, but it means every car in the chain must use a compatible standard, and there’s also the question of how cellular communication will be handled by providers of that service.
And then there’s the infrastructure. SimTD has worked with vehicle manufacturers and the city of Frankfurt to conduct a field trial, but it was limited to only twenty traffic lights. Implementing the infrastructure required by V2I communication will be an expensive venture, and it’ll be particularly difficult (if not impossible) to implement in rural areas where there’s a lot of road and not much money to build the infrastructure needed.
The Combined Solution
All of this makes V2I sound difficult to implement, at best, but the good news is that it’s entirely compatible with V2V, and in fact is likely to include it in any real-world system. This means that cars which lack the ability to communicate with infrastructure could still operate in the network in a limited sense, and all cars could default to V2V communications if needed.
Indeed, it’s unlikely we’ll see an infrastructure solution spring up alone anywhere in the world. Building such a network is both costly and time-consuming. It also requires mature technology, since changing the communication standard halfway through building infrastructure could ruin the entire project.
V2V platforms, by contrast, are already being deployed in limited numbers. Contrary to what you may have heard, they still have a long way to go before they’ll be cruising the highways in large numbers, but they do exist and can be developed quickly by independent teams.
These two approaches to autonomous cars are compatible because they rely on the same communication technologies. In fact, communications is not the most pressing issue faced by autonomous vehicles; simTD has already demonstrated existing WiFi and cellular can work well. The problem facing researchers is not solving how they’ll communicate, but instead deciding how they should behave once they do.
Image Credit: Wikimedia/SreeBot