News JVTech Sciences: nuclear power at the service of self-driving cars

Facilities that have it in their brains

CERN facilitieslocated near Geneva, on the border between France and Switzerland, are home to some of the most complex and sophisticated machines ever created by man. Its particle accelerators, including the LHC, which is the largest on the planet, are complex, but the detectors that collect collision data are even more so. From the outside, it might seem like once these machines are ready to perform experiments, all that’s left to do is child’s play – but nothing could be further from the truth than this information. Indeed, the processing of the enormous volume of information generated by particle collisions require a titanic computational effort.

Fortunately, scientists from CERN have come up with a system that works wonderfully. Basically what they did was develop a machine learning model able to analyze the data provided by the detectors very quickly to identify those that really have value.

A valuable machine learning model for autonomous driving

At first glance, it might seem that deriving knowledge from particle collisions in CERN detectors has nothing to do with autonomous driving, but in fact, it’s almost the complete opposite. CERN scientists have realized that it is possible to use a machine learning model very similar to that used in the field of particle physics to process and infer knowledge in real time from the information collected by the sensors of an autonomous car.

Photo credit: Zenseact

According to the CERN researchers, the most important contribution their technology can make in this field is its ability to radically refine artificial vision. What they plan to do is nothing less than make self-driving cars able to analyze the information collected by their sensors much faster and more accurately, in order to be able to react immediately to changes in the conditions of the environment in which the vehicle is moving. CERN technicians are so confident in their technology that they work with Zenseact on a research project to develop it so that it can be integrated into future self-driving cars. The latter company is dedicated precisely to the development of the technologies required for this type of vehicle.

CERN has confirmed that the hardware it uses in its self-driving car research project is essentially the same hardware it has been using for many years in its particle physics research: FPGA (Field-Programmable Gate Array). We won’t go into details so as not to complicate this article too much, but it is at least interesting to know that these are extremely versatile programmable logic circuits. CERN researchers decided to use them because they are able to execute complex decision algorithms in a few microseconds, and therefore deliver their results with minimal latency. So understand here that, on paper, these features fit like a glove to the processing hardware required for self-driving cars.