V2V Tech Gives Fleets a Third Eye

V2V Tech Gives Fleets a Third Eye

Vehicle-to-vehicle (V2V) technology has been lauded for its potential to help connected cars inform each other about their presence, send warnings of oncoming traffic at intersections and provide a host of other features designed to curb accidents.

It all sounds like fantasy today but Krystian Gebis, co-founder and CEO of Autobon AI, thinks that V2V communication could go even further by offering an extra set of eyes for commercial trucks. “It’s pretty critical that if I’m in a truck and there’s another pulled over in the emergency lane, and there’s a third in front of me, I can’t see that well because there’s a box trailer blocking my view,” said Gebis. “How is that issue going to be combated? Even though there may be good sensors on the truck, you’re still blocked. Solving it is going to boil down to connectivity and passing that information from all these vehicles to one another.”

Gebis presented an interesting use case for V2V where a truck, or indeed any automobile, could benefit significantly from the cameras, radar and LiDAR of other vehicles. If and when this becomes possible, it might provide a workaround for detecting things that aren’t connected. Imagine a dog bolting into traffic – if just one vehicle sees the animal, then all of them would become aware of its presence.

“Once a truck can see 30 seconds ahead, estimate where the vehicle positions are more than 1,000 feet away, I think that, especially for the highway situation, is where we need to get to,” Gebis added.

Driving in unison

Highways are the source of most traffic jams but Gebis believes that a mix of autonomous and connected technologies could one day eliminate this hassle. When combined with sensor data and information sharing, he envisions a fleet of vehicles that can move in unison.

“It’s all going to be able to happen at the same time,” said Gebis. “What you can imagine is, if there is the potential for a traffic jam, there may be a situation in the future where all these vehicles press on the gas at exactly the same millisecond. The cars start moving forward and you get rid of the traffic jam but, right now, you can’t do that because humans are all uncoordinated. They’re not connected.”

Defining artificial intelligence

AI is frequently promoted as one of the keys to making autonomous cars and other automated technologies come to life. It will, theoretically, allow robots to think for themselves and handle a plethora of complex situations but are we really there yet, or is it just hyperbole? Gebis has his doubts. He said that he does not believe anyone really knows what AI is or will become for automobiles.

“I don’t think the question has been answered yet,” said Gebis. “I think everything in the autonomous driving space is in more of a category of machine learning. Where a machine is learning a mathematical representation of various values that it sees and these values are coming from the sensors. The cameras are in color and the machine is learning various color codes in certain patterns and associating those features to then do a certain action or maneuver.”

Billions of miles

Test tracks and other confined locations have proven to be helpful to carmakers and tech start-ups, especially as they strive to perfect various autonomous vehicle features and concepts. However, as effective as they may be in the beginning, test tracks alone will never be enough to advance AV development.

“Testing is beneficial initially,” said Gebis. “Of course, it’s a safe environment but collecting real-world data is really where you are going to learn how humans actually drive, where they are good or bad. You can extract that good driving for the vehicle itself and extract the bad driving to understand how others perform, which provides the system with more insight as to what to expect.”

Gebis really drove that point home. If the industry expects to reach deployment, he said that a more scalable model is in order and to achieve that, real road experiences are absolutely necessary.

“It’s the trust level that drivers have with the vehicle,” he said. “Can I let go of the wheel and trust it with my life? Millions if not billions of miles need to be collected. There are two approaches for that. One is you can deploy all these vehicles. If you don’t have that ability you will have to simulate billions of miles of driving within a single day or week or so forth. I think what it boils down to is comparing an automated system’s mileage to an average human and beating that at least 10X.”

Leave a comment

Your email address will not be published. Required fields are marked *