But, it takes more than high-tech wings for a drone to safely do a deep stall. The team also had to build it a bird-like brain, one that could compensate for slight changes in speed, wind, angle and wing position. They achieved this through something called "Q-learning," a technique where an artificial intelligence learns an optimal course of action by raising its "Q," or satisfaction level. Like a kid in an arcade, it kept trying to beat its high score, until it worked out how to get from its starting point to its destination. After about 5,000 practice attempts, BMT says the drone pulled off a soft landing without a runway.