Artificial intelligence (AI) has shown that it is capable of performing everyday tasks that were previously only possible with the human mind. With AI now a heavy load truck moves from one point to another without being driven by a personto. With this technology you can also win a Formula 1 race.
This is not where AI came to replace drivers: Its main function is to deliver key data that is transcendental to make better decisions. Precisely, this information is what allows training the machine learning models (automatic learning) that understand the data, which learn how to react according to the situation.
What this system does are forecasts. That is to say, machine learning models are trained and for this more than 30 sensors are used (temperature, speed, fuel level, brake) distributed in various parts of the vehicle that yield more than 20 types of statistics in real time and that are hosted in the cloud.
In cargo trucks, for example, These sensors plus several special cameras provide information on how the roads are, the changes in traffic lights, the animals that can be crossed on the road, among others. Identify all these situations. This training lasts at least two years and is used for the models to carry out the simulations, as if they had a driver: they are the eyes, the accelerator, the brake. All.
This is an example of the company TuSimple which uses cloud computing services to develop its autonomous trucks.
“These technologies have mathematical algorithms that the person who is going to apply it does not necessarily need to know, but based on the data they can understand it and generate rules based on what happens in real life”, said Américo de Paula, Head of Solutions Architecture of MCO at Amazon Web Services (AWS).
Continuing with the case of the truck on the highway, these machine learning models identify the signals (colors) of the traffic lights: if it recognizes that it is red, it stops; in green, go ahead. In other words, the car guides itself according to the information provided by the model.
“AI manages to interpret data that comes from real life through sensors and these data are applied to the mathematical model that guarantees that a truck or a Formula 1 car is handled in an automated way”.
AI in Formula 1 racing
In Formula 1 cars, for pre-training speed and temperature sensors are installed which are the ones that deliver the data that reveals how likely it is that one vehicle will pass another.
Machine learning models at this point make a car more competitive. Through AWS, explained Américothese data are transmitted, processed and analyzed in time to transform it into knowledge. They show how close drivers are to braking at the apex of a corner, as well as different braking styles, which can give the driver an advantage when executing.
“Simulations are made in wind tunnels and all this data is taken to make an analysis that guarantees that the pilots can make transfers more easily on the track. It also generates a more entertaining experience for the fans, because they see the statistics in real time”, added de Paula.
In practice, What the teams do is use the data that comes from the sensors to try to predict if one of the cars will have a broken engine or a flat tire, situations of this nature to try to predict what the strategy for the next few laps would be.
The AI in this context makes it possible to know which drivers are pushing their car to the limit of its capabilities according to the conditions of the race: this includes traction, braking and cornering. These AWS developments were applied in the past Austrian Award.