这是早期成功应用RL到robotcis的文章。
Goal
Challenge
自动直升机是一个具有挑战性的控制问题,具有high-dimensional, asymmetric, noisy, nonlinear, non-minimum phase dynamics的特点。
Contribution
1.Learning a Helicopter Model from Flight Data
Collect data from a human pilot flying the desired maneuvers with the helicopter. Learn a model from the data