Audi to adopt self-learning AI technology for next-gen autonomous driving systems



While self-driving cars are the inevitable future of transportation, the journey towards autonomous driving varies from manufacturers to manufacturers. Audi has recently displayed their vision for autonomous driving using AI technology at the NIPS conference in California, USA.

Audi’s latest flagship model, the A8 features Level Three autonomous driving where the driver can completely leave his/her attention off the road and let the car handle the rest. Using multiple sensors and radars, the Audi AI Traffic Jam Pilot allows autonomous driving at speeds of up to 60 km/h. However, the only limitation of the system is that the system is not operable above speeds of 60 km/h due to the inability to accurately detect its surroundings 100 percent of the time.

Audi AI traffic jam pilot in the new Audi A8

Now, Audi is attempting to change that with its subsidiary company Audi Electronics Venture (AEV) working on using Artificial Intelligence to capture the exact surroundings of the car. The new AI system uses conventional front cameras as sensors and delivers 15 images per seconds at a resolution of 1.3 megapixels to a central computer which then processes the information.

The neural network (or AI) will then use semantic segmenting to identify each of the objects in the images and classifies it to 13 object classes allowing it to differentiate other cars, trucks, houses, road markings, people and traffic signs. The neural network will be trained using “unsupervised learning” which allows it to learn from observations of scenarios, cutting down the time needed to process information.

Audi hopes that the AI will eventually be able to independently generate a model of the environment from camera data and drive the car itself without needing a preloaded or localised map data.