A survey of direct learning control

Learning control, such as iterative learning control (ILC), has been developed for decades, but is mainly used to solve repetitive problems. Direct learning control (DLC) is proposed for non-repetitive problems in practice based on pre-stored information. According to different varying factors, direct learning control can be divided into time-scale DLC, magnitude-scale DLC and dual-scale DLC. DLC has its unique advantages and can be applied to control tasks with high accuracy requirement. In this paper, we have sorted out the emergence and development of DLC, summarize the DLC scheme for linear and nonlinear systems, review the recent advances in DLC, and finally put forward some further development directions.