Moving the Internet Closer to Autonomous Cars Could Make Them Safer

Robot cars are getting closer to reality with a new technology that could make fully autonomous driving vehicles safer and cheaper to implement. 

Cisco and Verizon recently demonstrated that mobile edge compute (MEC) technology can enable autonomous cars without expensive physical roadside units to extend radio signals. The idea behind MEC is that running applications closer to the cellular customer allows applications to perform better. Cities might be able to create less dangerous roads using the system. 

"With MEC, we can move the compute burden to the edge of the network, that is, nearer to the end-user and the vehicle and not in some faraway data center so that the total time it takes for data messages to be sent out and received back is much shorter," Dennis Ong, a senior manager of systems architecture at Verizon, told Lifewire in an email interview. "That enables autonomous driving features in vehicles that can be undertaken in less than one-tenth of a second—faster than humans can react in some cases, and fast enough to enable certain safety features."

Getting Robot Cars on the Road

Autonomous features in connected vehicles usually rely on roadside radios to extend the signals vehicles use for low-latency communication with each other and surrounding infrastructure. The recent test was meant to prove that cellular networks and special routers can meet the latency or delay standards necessary for autonomous driving applications.

One of the key uses for MEC technology is safety. Verizon's proof of concept with Cisco could help vehicles navigate intersections, for instance, assisting a loaded truck stop in time for a changing traffic signal, helping emergency vehicles preempt signals safely, or helping ensure robotaxis and unmanned delivery vehicles understand and obey traffic signals.

In another test, Nissan and Verizon demonstrated MEC technology that can notify drivers of pedestrians or other vehicles emerging behind visual barriers, for instance, during left turns with oncoming traffic, nearly in real-time.

MEC also makes things easier for auto engineers. The technology stores local maps of the roadways so that the vehicle doesn't have to waste processing power scanning the road and determining its geometry. 

"The road doesn't change, only the locations of vehicles change, so the only thing the vehicle should have to worry about is where it is relative to the fixed map and the moving vehicles," Tim Sylvester, the CEO of Integrated Roadways, a company that builds autonomous car infrastructure, said in an email interview. 

With MEC, onboard systems can be used to detect other vehicles and decide how to navigate safely. When MEC is supported by smart infrastructure, including in-road vehicle sensors, the self-driving car's job gets more straightforward. It doesn't even have to figure out where other vehicles are since the MEC can give the car the map and the other vehicle locations. 

"And with 'Where am I?' and 'Where are the other cars?' taken care of by network services, the self-driving cars' responsibilities are reduced to only having to figure out how to navigate safely," Sylvester said. "That's the real path to autonomy—MEC and smart infrastructure, so that autonomous cars are simple and inexpensive."

A Growing Need

MEC isn't a standard solution for existing autonomous vehicles because the technology isn't widely available yet. Designers for self-driving vehicles have relied mainly on the assumption that the car will be independent of support networks. 

“It’s a chicken-and-egg problem: cars can't use networks that aren't available, and it's hard to justify implementing a network that cars aren't using,” Sylvester said. “But once they're in place for those other uses, it becomes straightforward for the vehicles to adopt them, since all the vehicle has to do at that point is have a communications system that can receive the data and an onboard computer system that can use the data from the MEC and in-road sensors.”

Sylvester predicted that within the next decade the most heavily trafficked urban routes will be outfitted with smart infrastructure solutions like MEC and in-road sensors. At the same time, the automotive industry will continue to evolve with more advanced autonomous capabilities, Sid Krishnamurthi, head of product management at Recogni, which makes systems for self-driving cars, told Lifewire in an email.