Motorists are always in a hurry to cross as soon as a traffic light turns green, but this congestion can lead to serious accidents and sometimes lead to traffic violations with heavy fines.
Waiting at a traffic light is tedious, time consuming and fuel-intensive, it also harms the environment. To end this problem, policymakers and motorists are looking for more relevant and reliable alternatives, albeit very small. A team of researchers at the Massachusetts Institute of Technology (MIT) thinks they have found a solution to the problem.
They tried to find many ways to make sure drivers didn’t have to wait at the junction for the signal to turn green. Instead of waiting so long for the signal to turn green, what would happen if they did just when the signal turned green? Although it is difficult for a human driver to calculate the exact moment when the signal turns green and arrives at that exact moment, an autonomous vehicle, that is, a vehicle that does not require a driver and which operates with artificial intelligence, can achieve this. Very consistently
A vehicle that works with artificial intelligence can be controlled accordingly, its speed can be adjusted so that the car reaches the traffic light at the moment it turns green and does not have to wait for the light to turn green.
The team has already published a study, in the electronic archive of Preprints arXiv.org, where it validates a machine learning method, which will learn how to direct a fleet of autonomous driving vehicles to streamline traffic flow. The research team was led by Bindula Jayawardene, a graduate student, and the team reported that this machine learning method uses less fuel and helps reduce emissions, but it can increase the speed of an automobile, thus reducing our travel time.
The downside is that researchers want this method to reduce fuel consumption while reducing travel time, but to reduce emissions their car doesn’t need to come to a complete stop or just slow down. So it can be difficult for a learning agent to fulfill these two inconsistent objectives.
To address this problem, the researchers proposed an alternative solution called “reward shaping”, which provides the learning agent with information that he or she cannot learn on their own. They penalize the system every time the automobile stops to teach him not to repeat this mistake in the future.
Using a single intersection traffic simulation model, they test when their control algorithm is fully ready. The system does not immediately stop the vehicle as it approaches the intersection, as it does in intermittent traffic.
Compared to self-propelled cars, many of them have gone to a single stage of green signal and this method has effectively reduced greater fuel savings and emissions.
Therefore, if every automobile on the road runs autonomously without human intervention and has an integrated system, fuel consumption can be reduced by 18% and carbon dioxide emissions by 25%. In addition, it can improve the average travel speed of a car by 20%. Finally, if only 2% of cars drive autonomously, they can help reduce fuel consumption and emissions by at least 50%.
What do you think?
Volkswagen executives say Level 5 self-driving cars can never happen, but Level 4 looks achievable
Tesla has unveiled its new supercomputer (5th most powerful in the world) for AI training for autonomous driving, a machine currently being used and developed by the company
Traffic jams will be a distant memory for AI traffic lights created by Aston University researchers