Self-driving cars negotiating their way through the intersection

In the video above we see three cars heading for an intersection, at a speed of 50 km/h, and passing the intersection almost simultaneously. The self-driving cars are initially on a collision course, but negotiate with each other and adjust their speeds to avoid collisions and maximize the traffic flow.


Intersections are particularly challenging and dangerous in today’s traffic system. A large number of cars need to share the same crossing and, to avoid accidents, traffic lights, signs and rules are necessary. Unfortunately, this causes the traffic flow to slow down and queues to arise, and fuel consumption to increase. With communicating self-driving vehicles there is a great potential for improvement and many problems could be alleviated.

The demonstration shown in the video is conducted at the AstaZero test track by a group of researchers from Chalmers. When the self-driving vehicles approach the intersection, they communicate with the other vehicles and a joint decision is taken on how to cross the intersection in an optimal way.

"Today's traffic system is both accident-prone and inefficient. Often the vehicles are forced to halt unnecessary. With our method, we can automatically find the most energy efficient and safe solution. For example, a smaller vehicle would accelerate through the intersection in order to avoid slowing down a truck”, says Robert Hult, researcher at the Electrical Engineering Department.

The decision-making process is very fast and is repeated as the cars approach the intersection to adapt to unforeseen events. The demonstration is developed to promote efficient traffic flow and the method is scalable and can be applied to more than three cars.

What does it feel like sitting in a self-driving car passing another car with such small margins?
"The passenger's experience must of course be taken into account, and although the comfort was very good, the crossing felt a little tight at first. After a couple of times, however, it became quite natural", says Mario Zanon, researcher at the Department of Electrical Engineering.

The research involves several research groups at Chalmers, and is partly performed within the research project Copplar. Copplar stands for Campus Shuttle cooperative perception and planning platform, which aims to develop a self-driving vehicle for city traffic, regardless of weather conditions, which can be run between Chalmers two campuses Johanneberg and Lindholmen.
 

The results have been published:
Primal decomposition of the optimal coordination of vehicles at traffic intersections
Robert Hult; Mario Zanon; Sébastien Gros; Paolo Falcone, Chalmers University of Technology

An Asynchronous Algorithm for Optimal Coordination at Traffic Intersections
Mario Zanon; Sébastien Gros; Paolo Falcone; Henk Wymeersch, Chalmers University of Technology


The research has been funded by VR, the Swedish Research Council, FFI, Strategic Vehicle Research and Innovation, Vinnova grant no 2015-03075, the research project Copplar, and Chalmers Area of Advance Transport. The demonstration was carried out with support from Volvo Cars, Fengco Real Time Control and Leica Geosystems.

Contact:
Department of Electrical Engineering, Division of Systems and Control
Robert Hult, PhD student, robert.hult@chalmers.se
Mario Zanon, Postdoc, mario.zanon@chalmers.se
Sébastien Gros, Associate Professor, grosse@chalmers.se
Paolo Falcone, Associate Professor, paolo.falcone@chalmers.se

Published: Tue 16 May 2017. Modified: Thu 29 Jun 2017