Imitation learning (IL), proposed in 2010, is a method for acquiring strategies by learning from expert demonstrations. Unlike traditional reinforcement learning, IL does not require an explicit reward function. Instead, it directly learns what to do based on what the experts do.

Imitation learning mainly includes several methods such as behavioral cloning, inverse reinforcement learning, DAgger algorithm, and generative adversarial imitation learning, and is widely used in many fields such as autonomous driving and robot control.