Adaptive cruise control will not yet resolve congestion, a public experiment by the Faculty of Electrical Engineering showed - News service

Publication date:

2019/12/18

In December, the scientific team of Associate Professor Zdeněk Hurák from the Department of Control Engineering at the Faculty of Electrical Engineering, together with volunteers, tested what would happen if a multi-car column was created on the road. The experiment showed that the random sample of cars on Czech roads using the ACC will behave qualitatively the same as in the case of human drivers - ie unjustified (phantom) traffic jams will occur.

Under the leadership of doc. Zdeněk Hurák, a team from the Faculty of Electrical Engineering conducted a unique experiment with cars equipped with adaptive cruise control (ACC) at the airport in Mnichovo Hradiště on Friday, 6 December. Owners and drivers of twelve such vehicles responded to public call, making it an experiment with public participation. Scientists assembled a series of a total of 12 vehicles with the ACC on, driving behind the lead car on a two-kilometer track. The lead car’s task was to start with a smooth ride at 60 km / h, then a slight and brief braking to slow to 50 km / h and then accelerate again to the original cruising speed. The other cars had an adaptive cruise control set to 80 km / h. Scientists then recorded the reactions of these cars in the column, changing their speed.

Drivers in the rear of the column could observe that their car braked up to 30 km / h and then accelerated up to 80 km / h (the value set on their cruise control) before returning to its original cruising speed of 60 km / h. The observed phenomenon is known as string instability, well known from human car driving. In the case of adaptive cruise control, where acceleration and braking are decided by the on-board computer only by measuring its own speed and distance to the previous vehicle, the international expert community acknowledges that chain instability should be suppressed. However, the experiment showed that a random sample of cars on Czech roads exhibits qualitatively the same behavior as for human drivers. The hope that cars with adaptive cruise control (in today's version) become more present on the Czech motorways will be preventing an unjustified (phantom) congestion, seems to be false.

The subject of the experiment was not to determine whether the observed contribution to chain instability in vehicles equipped with adaptive cruise control is stronger or weaker than when driving. Adaptive cruise control is generally praised with reliability, accuracy, and speed of machine decision making, but has drawbacks in inability to predict, even by recognizing the brake light (s) of the previous car (or several cars) and inability to read and evaluate alerts on impending constriction or roadwork.

Similarly, the aim of the experiment was not to compare individual cars with each other. Scientists did not conduct thorough (and time-consuming) experiments with individual vehicles. Only on the basis of such measurements would it be possible to reliably quantify the contribution rate of specific adaptive cruise controllers to the chain instability of the whole. However, it is not excluded that it will perform different for cars of different makes and types.

Significant improvements in the capabilities of these assistance systems with regard to their impact on traffic flow can only be expected after other features have been incorporated, such as the ability to detect the front brake light on one or more vehicles or even using vehicle-to-vehicle wireless communication (V2V) to receive information about what is going on ahead. In the international professional community, such systems, known as cooperative adaptive cruise control (CACC), are subject to very intensive research, and the first prototypes have been introduced in experimental operation.

Doc. Hurák's team wants to use the data measured in the experiment to develop more accurate models describing the dynamics of cars equipped with adaptive cruise control. These will be published in the professional press, however, the data from the experiments are already available to the public at: https://gitlab.fel.cvut.cz/aa4cc/acc/lkmh