Stanford Autonomous Helicopter Project has developed new reinforcement learning techniques to stabilize and control their helicopters. The developers have been able to teach their heli to perform just about every aerobatic figure imaginable. However, the heli has no on-board intelligence. Instead, data is fed from onboard GPS and IMU with serial outputs directly into Xbee Pro modules. The signals are received by a PC on the ground, which enables operator control via the training port of a conventional R/C transmitter.