Researchers have developed a new AI system that allows self-driving cars to exchange real-time driving insights without needing to connect directly with one another—a breakthrough that could improve both safety and efficiency on the road.
Generally, autonomous vehicles rely on central servers to store and process data. While effective, this setup increases the risk of large-scale data breaches and puts heavy strain on computing infrastructure.
Current systems also require vehicles to be within close range and grant permission before sharing information.
A social network for cars
The new method, called Cached Decentralized Federated Learning (Cached-DFL), changes that approach. Instead of depending on centralized networks or immediate data transfers, vehicles using Cached-DFL store what they learn in internal smart models.
They later share this data when they come into contact with other vehicles, without needing to establish a prior connection.
do we need to design future roads keeping in mind self-driving cars? and design cars keeping in mind they should be able to talk to each other. Need a global protocol for road and car design pic.twitter.com/UE9FUIKYFO
— Cyber Molecule (@cyber_molecule) February 26, 2025
“Think of it like creating a network of shared experiences for self-driving cars,” said Dr. Yong Liu, professor of engineering at NYU’s Tandon School of Engineering and lead researcher on the project.
“A car that has only driven in Manhattan could now learn about road conditions in Brooklyn from other vehicles, even if it never drives there itself.”
The system functions much like a social network for vehicles. Each car carries a type of profile, showing recent experiences such as traffic changes, road hazards, or signal patterns. This information remains anonymous and does not include personal driving habits.
Testing the system in a virtual city
To test the concept, researchers built a simulation of Manhattan and placed 100 virtual self-driving cars into the environment. The vehicles drove in semi-random patterns, and each carried 10 AI models that updated every two minutes.
Rather than transmitting data instantly, the cars cached it and shared it only when a direct vehicle-to-vehicle connection became available. This approach proved effective.
When a driver could not go to her car due to rain, she had Tesla come to her car with its "smart calling" feature.pic.twitter.com/W6H3MkB0rF
— The Figen (@TheFigen_) September 12, 2023
As long as vehicles were within 100 meters—about 328 feet—they could access each other’s stored updates, even if they were unfamiliar or from different systems. The frequent, short-range exchanges helped improve learning speed and decision-making accuracy.
Decentralization helps the system scale
“Scalability is one of the key advantages of decentralized FL,” said Dr. Jie Xu, associate professor of electrical and computer engineering at the University of Florida.
“Instead of every car communicating with a central server or all other cars, each vehicle only exchanges model updates with those it encounters.”
Expanding to more devices and real-world tests
Looking ahead, the team plans real-world testing and hopes to enable communication between different brands of autonomous vehicles. They are also working on expanding the system to connect with traffic lights, road signs, satellites and other smart devices.
This broader integration—known as vehicle-to-everything (V2X)—could enhance traffic coordination across cities.
Researchers believe the Cached-DFL approach could eventually support not just cars, but a wide range of smart machines, including drones, satellites and robots. By enabling these devices to learn from one another on the spot, the system could form the foundation for rapid, swarm-like intelligence in future networks.
“Decentralized federated learning offers a vital approach to collaborative learning without compromising user privacy,” said Javed Khan, president of software and advanced safety at Aptiv. “By caching models locally, we reduce reliance on central servers and enhance real-time decision-making, crucial for safety-critical applications like autonomous driving.”
See all the latest news from Greece and the world at Greekreporter.com. Contact our newsroom to report an update or send your story, photos and videos. Follow GR on Google News and subscribe here to our daily email!
