Efficiently navigating intersections with autonomous vehicles

What is it about?

With several prototypes for autonomous vehicles hitting the market, it will not be long before we encounter autonomous vehicles in our day-to-day lives. Our current traffic management systems are designed for human-driven vehicles, which begs the question: Can traffic management systems be improved to be more efficient when dealing with autonomous vehicles? It turns out that autonomous vehicles with their advanced technological capabilities can pave the way for newer traffic management strategies, especially at bottlenecks such as road intersections. Vehicle platooning is one such strategy that exploits the communication capabilities of (autonomous) vehicles to keep vehicles in a `linkless train'. Imagine a train with numerous carriages. The carriages are linked to each other and the train locomotive typically pulls the whole train forward along the tracks. Platooning achieves the same idea, but with cars and trucks (for example) instead of carriages, and there is no physical link between two consecutive vehicles. Rather, vehicles communicate with each other to synchronise their movements and move as a unit (also called a platoon) at a high speed. Research studies have shown that platooning near intersections can provide benefits such as reducing travel times and fuel emissions compared to traditional traffic lights with a static cycle. This is partly because platooning can handle dynamic changes in traffic, and vehicle platooning reduces the aerodynamic drag experienced by all vehicles except the leading vehicle (=locomotive, according to our train analogy) in the platoon. Moreover, with platooning in place, vehicles are now able to cross the intersection at high speeds, leading to more vehicles being able to cross the intersection in the same span of time. No more waiting for the vehicle in front of you to notice that the light has turned green and then accelerate! In this paper, we consider a platooning-based system to manage traffic around intersections, by using a combination of two algorithms -- a platoon-forming algorithm and a speed-profiling algorithm. The traffic can be composed of various kinds of autonomous vehicles. The platoon-forming algorithm organises previously `scattered' vehicles into platoons and generates a customised `green' time for each vehicle; this is the time when it can safely enter the intersection area. This green time is generated well before the vehicle reaches the intersection, based on information about other vehicles arriving at the intersection. Since we would like vehicles to cross the intersection at a high speed, it is no longer possible for them to come to a stop and wait just before the intersection area. The speed-profiling algorithm thus produces an optimal path for each vehicle so that it can approach the intersection at the correct `green time' while maintaining a high speed.

Featured Image

Photo by Firdouss Ross on Unsplash

Photo by Firdouss Ross on Unsplash

Why is it important?

A common assumption made when developing new frameworks for autonomous vehicle environments is that of traffic homogeneity -- assuming that all traffic has identical characteristics. For example, all vehicles in the world are cars of the same make and model. This simplifies the problem at hand, and can even lead to elegant solutions; however, this approach drives the problem further away from reality. With this work, we make away with the traffic homogeneity assumption and show that the resulting framework is not complex to analyse. In fact, a focal point of our work is a variant of the speed-profiling algorithm which generates closed-form expressions for vehicle trajectories instead of relying on solving optimisation problems numerically. Closed-form expressions are valuable in many ways: they allow for significantly faster implementation, they lead to more precision during the trajectory-generation process, and they lead to more insights that would be otherwise difficult to gain by numerical optimisation alone.

Perspectives