As the development of self-driving cars continues, a number of key questions still remain unanswered. For starters, we still need to find out how self-driving vehicles will interact with normal, human-driven cars. This is due to the fact that what might be sensible to a tired human driver may not always make sense to a programmed computer chip. A research team working at the EPFL (École Polytechnique Fédérale de Lausanne) may have worked out a solution to this conundrum. They have developed an algorithm that enables autonomous cars to move around, platoon, and merge with a lot more freedom.
Anybody who has been following the development of self-driving cars is sure to have come across the term 'platooning'. This refers to a situation where autonomous cars that are linked through a wireless connection are moving at the same speed in close formation on a public highway. Because of the wireless connection, the cars in a platoon can travel while being much more nearer to each than human-driven cars, and in this way reduce traffic and enhance aerodynamic efficiency.
Platooning has been tested in Europe and other places. In one instance, a fleet of self-driven trucks drove themselves from various factories to Rotterdam without any input from humans.
EPFL says that the main drawback with platoons is that they move as a single large block. This may work well where three or four motor vehicles are involved, but managing large convoys can be problematic. In addition, there is no communication between the autonomous vehicles and the traffic near them, and as a result, human drivers are not aware of what is going on.
The Lausanne team advocates for a cooperative and evenly distributed system. Rather than having one leader directing the actions of the whole fleet, the system recommends every vehicle to change its position and speed depending on what the other connected vehicles around them are doing. Since all the vehicles are linked together, each one of them is aware of what the sensors on the vehicles around them can see. The result is something that is close to having 360-degree vision.
In case a vehicle wishes to join the convoy, the vehicles are reshuffled by an algorithm that utilizes the information that it gets from the platoon. This might entail making one vehicle to slow down in order to create a gap, an action that is performed by human drivers every day on the highway. The researchers began to test this system by performing simulations, then they up-scaled to miniatures before finally using real vehicles.
There are a number of potential problems, mainly arising from the fact that the algorithm is founded on the idea of car-to-car communication. Even though connected vehicle technology is widely viewed as an important element of successful autonomous driving, there are very few vehicles that are equipped with the necessary wireless capabilities.
Even vehicles with human drivers can gain from using the software. A modified display can show them when the self-driving vehicles around them intend to merge, accelerate, slow down and change lanes. The researchers at EPFL are hoping that this proof of concept is sufficient to persuade manufacturers that it is viable to develop software that can modernize older cars and enable them to interact with connected vehicles.
Should they succeed, human drivers would be notified when another car intends to change lanes or merge using their current dashboard screens. This would help their movement to be smoothly synchronized with that of the platoon.
Alcherio Martinoli, who heads the laboratory that is responsible for developing algorithms and intelligent systems at EPFL says they have been working on this kind of distributed control algorithm for about ten years. He adds that the objective is to get a way that can enable agents that are not especially clever, be they cars or robots, to cooperate and behave like a complex group.
Image credit: 2025ad.com