By using end-to-end machine learning, Wayve has developed an autonomous vehicle system that has human-like driving skills by using “only cameras and a sat-nav,” according to the company. The system forgoes hand-coded rules, sensor suites, and high-definition maps, yet still demonstrates complex navigation on unfamiliar urban roads.

Below you can see Wayve’s data-centric system on the public streets of Cambridge, U.K.

The system mimics a human’s driving skill by utilizing imitation and reinforcement learning, and traverses roadways through computer vision.

“Our model learns both lateral and longitudinal control (steering and acceleration) of the vehicle with end-to-end deep learning. We propagate uncertainty throughout the model. This allows us to learn features from the input data which are most relevant for control, making computation very efficient,” according to Wavye.

The AI can operate on systems similar to today’s laptops, according to the company, and uses less than 10 percent of the power, sensor, and compute requirement of current methods.

For safety reasons, trained human drivers are behind the wheel during all tests.

Click below to watch another example of the Wayve’s autonomous vehicle driving through U.K.’s narrow streets. During this test, it had never seen these roads before, and was armed with just 20 hours of training data. It learned all its behavior through data patterns, meaning the system wasn’t previously told to slow down for right-of-way rules at intersections, or to drive on the left side of the road.