By employing multi-agent deep reinforcement learning, the model prioritises time-sensitive messages, reducing delays and enhancing reliability, paving the way for safer roads and smarter cities in India.
Indian researchers have developed a model that aims to address a potential problem that will arise in future, when there are too many autonomous vehicles on the move, at the same place, at the same time.
The model aims to solve a major challenge in Vehicular Ad-Hoc Networks (VANETs), where nearby vehicles communicate directly with each other. When many vehicles send messages at the same time, it can cause network congestion, leading to delays or lost messages, which can make the system unreliable. This model is designed to address that problem.
Researchers at NIT Rourkela have suggested a solution to this issue involving the use of Artificial Intelligence (AI). Their model utilises a system known as multi-agent deep reinforcement learning which enables each vehicle to stagger the time of its messages depending on the actions of other vehicles.
Instead of communications competing, the vehicles system learns to sequence and give priority to messages that are time sensitive. This adaptive adjustment reduces the chances of delays and helps ensure that important alerts are transmitted reliably, the researchers say. They have also secured a patent for their model.
“In 2023, India reported around 480,000 road accidents and around 172,000 deaths, many of which could be prevented using modern technologies. Our work is a step towards building safer roads and smarter cities. We envision a near future where autonomous vehicles become a reality in India, and this patent is a small step in that direction, driving the spirit of Innovate in India and Make in India,” Dr. Arun Kumar Assistant Professor, NIT Rourkela said.
In addition to Dr Kumar, Prof. Bibhudatta Sahoo, and Dr. Lopamudra Hota, Department of Computer Science & Engineering, NIT Rourkela are part of this research project.
“The patent represents a practical step towards preparing India's road system for vehicle-to-vehicle communication. By addressing potential congestion in VANETs and providing a model for adoptive, coordinated communication, the findings lay the groundwork for safer and a more efficient traffic management. We invite researchers across institutions to join our efforts in making future autonomous vehicles and collaborate with our research lab at NIT Rourkela,” Prof. Bibhudatta Sahoo, said.