Artificial intelligence is already an old lady for automotive engineers. For around ten years, it has contributed to the democratization of driving aids. Since the little irritating beep which announces the proximity of a post along the sidewalk, which appeared in the 1990s, parking assistance has been converted to automatic slot, then to the scanned place in a specific place where the car leaves park alone.
Better: Valeo is working with BMW on an “autonomous valet” that will pick you up after a week of travel at your airport exit gate. No more shuttles to cross Roissy and reach the long-term car park! This function will appear from 2027 on certain series from the Bavarian manufacturer. “Equipment that makes driving easier and more comfortable is experiencing real appeal, especially as it becomes more user-friendly,” notes Antoine Lafay, director of autonomous vehicle research and development at Valeo.
Big star of the Las Vegas Electronics Show last January, ChatGPT is now in the passenger compartments of General Motors, Mercedes, DS, Volkswagen and soon Peugeot. Unlike traditional aids, the learning machine easily understands natural language. The passenger complains loudly about the heat? She suggests opening the window or turning on the air conditioning. On request, she finds the grocery store still open at this late hour, draws up the shopping list for the planned recipe, tells a story to the children or that of the monument seen along the way.
“But self-learning in complete freedom stops where security begins,” warns Antoine Lafay. The algorithms feed the car with vast sets of data on driving patterns, traffic conditions or new uses such as electric scooters.
Artificial neural networks
The information from the sensors, processed in real time, triggers independent decisions on acceleration, braking, changing lanes, etc. “Because we trained the system,” explains Antoine Lafay. Artificial neural networks are much more efficient than older methods in terms of accuracy. However, the elements collected continue to be validated before re-entering the automobile like phone updates.
These developments reinforce the effectiveness of perception systems. Reading the road signs triggers the display of the speed limit pictogram on the dashboard; reversing cameras immobilize the vehicle when they detect a child in a blind spot… Wisely! Since January 1, 2024, all new vehicles are required to have an active safety panel: emergency braking, overspeed prevention, monitoring of driver distraction due to fatigue or checking their phone.
Vigilance remains crucial in level 2 of autonomy, the one that already equips many models: automatic lane change, branching off towards the motorway ramp corresponding to the programmed route, etc. At this stage, the hands can let go of the steering wheel, but eyes must not leave the asphalt.
From stage 3, driving is entirely delegated in well-defined geographical, environmental and meteorological conditions: using certain roads, within a precise speed limit, depending on specific traffic conditions. No question of doing a Paris-Amiens therefore. “We are changing the paradigm. We are moving to levels close to aeronautics,” enthuses Antoine Lafay. There, image capture and electromagnetic waves are no longer enough. Lidar, with its laser beam, is becoming essential. It continuously scans the environment and provides a very precise measurement, at the speed of light. The BMW 7 Series, Mercedes S-Class and Honda Legend are equipped with this high degree of self-management and approved in Germany, the United States and Japan.
Hence the race for the “car designed around software”, SDV in the jargon of the trade. This involves moving towards a centralized IT architecture and reducing the complexity of processors. “Current models have more than 100 million lines of code, while an F35 fighter jet only has 13 million,” explains Payam Shokrzadeh, software supervisor at Continental. With the advantages of remotely modifying engine power, battery management, identifying worn parts and installing new applications from ultra-secure online servers.
Accelerated toll passes
Artificial intelligence strengthens connectivity between cars and infrastructure and contributes to the flow of traffic. If a signal indicates to the adaptive cruise control that the light is about to turn red, the engine decelerates automatically. In the same vein, Valeo is collaborating with the Paris-Rhin-Rhône highways to improve toll passage. “Dialogue with the facilities makes it possible to anticipate which line to take in the event of a closure or when a user blocks a barrier because they cannot find their credit card,” relates Antoine Lafay.
Above all, the exchange of messages helps to reduce mortality. In the event of work, they trigger an offset on the furthest axis and thus protect the workers. They announce the imminent crossing of a masked pedestrian by a delivery van. Intelligent roundabouts, equipped with lidars, warn the user of a hidden obstacle.
“The next step will concern trajectory planning, prediction and decision-making for fully autonomous driving, on motorways for example,” announces Antoine Lafay. And even remote driving! A bit like the remote-controlled cars of our childhood. An operator will pilot the vehicle like a drone, from its factory assembly line to the parking lot, then transport it to the truck intended to deliver the dealerships. Rental companies will also appreciate no longer requiring their jockey to return on foot from the washing stations to the customer reception office. No more need for staff to bring it back. Saving valuable time for commercial activities.
No, the autonomous vehicle is not a gadget.
And here are the 4.0 tractors!
Trektor
© / Sitia
“Today, the inside of a tractor looks like the cockpit of an airplane.” The formula comes from Clément Aubry, director of research and development at Sitia, an SME from Nantes which designed Trektor, the first autonomous and hybrid multi-purpose vehicle. So, unsurprisingly, farmers, whatever their age, can easily control it from… a phone or a tablet. They select the plot to cultivate, the desired program (trimming the vines, weeding in market gardening, loosening the soil, etc.) and refine them according to last-minute parameters, such as the weather.
“We work with living things. The environment is constantly changing and there is no road marked by white lines. This results in choices of sensors and algorithms different from those of a car, better able to to predict variations in the terrain and manage the diversity of tasks”, specifies the doctor in robotics.
The machine also sends information on real-time consumption and maintenance needs. Above all, it partly compensates for the lack of labor – increasing in the fields – and frees up time for logistics and… administrative tasks. Waiting for the shock of simplification.
An article from the special “Automobile” report of L’Express, published in the weekly on April 18.
.