In The Future Autonomous Cars May Also Be Used For Traffic Management On Roads

Autonomous vehicles are anticipated to solve many on road problems that the world faces today. One of them is the ever increasing issue of traffic congestion. It is believed that once the entire vehicle network on the road is autonomous and connected, traffic congestion would be a thing of the past. And this is not at all a sci-fi concept that has just been conceived out of nowhere, without any evidences. Time and again, it has been concluded through the experiments conducted by several auto-makers that autonomy in driving will prove to be better than any human behind the wheel.

But that is not the case right now. That won¡¯t even be the case 10 years down the line. So what can we expect until the above mentioned situation is achieved, until the roads comprise of all kinds of vehicles, including autonomous, semi-autonomous as well as human driven vehicles?

Researchers suggest that autonomous driving can still act as a solution for traffic management. More like a herder for the entire traffic.

Transportation researchers at UC Berkeley credit a machine learning algorithm for this, named ¡®Flow¡¯. The recently developed tool uses deep reinforcement learning, allowing it to find the apt solution as it keeps facing different situations and learning from them. The situations included bottlenecks traffic, gridlock traffic or on-ramp merges.

Berkeley/ YouTube

Flow was used on a single autonomous vehicle in a hoard of other vehicles in the above mentioned traffic scenarios. The system gained input from smart vehicles and infrastructure around it. It then interpreted the data to come up with an optimum solution and drive accordingly.

In an example cited in his demonstration, Alexandre Bayen, director of UC Berkeley's Institute of Transportation Studies, showed how humans driving in a simple circle could not maintain a constant speed and hence led to traffic halt points within the circle, obviously resulting in energy inefficiencies. The experiment was originally conducted back in 2008 in Japan.

Berkeley/ YouTube

However, in 2018, an autonomous vehicle was introduced to the same experiment. Once a single autonomous vehicle was ingested into this circle, it maintained its speed in such a manner that all the other vehicles could smoothly move through the course.

Berkeley/ YouTube

Similar results were observed in other traffic scenarios as well. As can be seen in the video below, Flow managed to create a snake or come up with a randomised solution as per the scenario for better traffic management. Surprisingly, the results were more efficient than with just human drivers, every time. The effect can be perceived like the autonomous vehicle guiding the entire flow of traffic on its own.

So does it mean that autonomous vehicles can be a traffic management solution in a mixed autonomy on the roads. Probably. The lack of assurity is mostly because conditions on the roads are largely different from the smooth flow that has been depicted in the demonstrations. For instance, considering the driving practices in India, an autonomous vehicle holding up the traffic will only blared on and nothing else. Also, given the bumper to bumper traffic in most scenarios, the autonomous car may not even move in an effort to avoid any collision.