Is it more irritant than feeling trapped in a traffic jam that you don''t want to go straight to another red light?
According to a statement from the university on Tuesday, scientists at NowAston University have developed a new artificial intelligence system that might stop long queues at traffic lights.
Deep reinforcement learning
The system is so effective because it utilizes deep reinforcement learning, which makes sense that it actually adapts its operations when it is not functioning well and continues to improve when it progresses.
The concept of traffic control is based on a modern photo-realistic vehicle called Traffic 3D, which is characterized by an unexpected reward.
Adaptable to real-world settings
This simulator has been developed to handle several traffic and weather scenarios and was therefore able to quickly adapt to real traffic intersections, making it effective in many real-world situations.
Weve compared this program to learning behaviors, which it hasn''t explicitly identified before. The computer will ultimately determine what that link is as long as there is a causal link, according to Dr. George Vogiatzis, a senior lecturer in Computer Science at Aston University.
The study was published in Aston University''s Library Services.
Abstract of the study
Despite being fully trained in simulation, our traffic control agent evaluates the possibility of individualized signaling in real-time. Due to economic and safety limitations associated training such agents in the real world. This paper proposes a fully autonomous, vision-based DRL agent, which is capable of reducing the real-world spectrum. Our agent analyzes and develops robust behavior in response to complex traffic situations and provides a clear pathway to previously unobservable real intersections.