Real world artificial intelligence is still less developed than a toddler, especially when considering the automotive industry – but progress is made faster than you can say AI.
The development of the sector – alongside many others – is crucial for automakers because they want to put out on the streets as fast as they can truly autonomous cars. This concept is just a piece of the incredible autonomous puzzle, dubbed the Audi Q2 “deep learning project” and capable of locating a parking spot entirely on its own – through the good old trial and error. The Q2 has a couple of mono cameras – one facing the front, the other facing the rear – as well as ten ultrasonic sensors that will scan the environment. The on-board computer will gather and convert the data from the sensors and cameras into control signals for the steering and powertrain.
This allows the car to self-drive around the owner’s location until it can find an available parking spot, slide in – all without being monitored by anyone inside. The system uses “deep reinforcement learning” – meaning it will use trial and error to better itself. Software and algorithms will register the successful attempts and refine the parking strategy – just like a human it starts with the easiest parking maneuvers and in time grows into an expert. The Q2 seen here is for now just an 1:8 scale model built as a demonstrator and will participate in the Conference and Workshop on Neural Information Processing Systems, hosted n Barcelona, Spain.