Attitude control of underwater glider combined reinforcement learning with active disturbance rejection control

摘要

Buoyancy-driven underwater gliders are highly efficient winged underwater vehicles driven by modifying the net buoyancy and internal shape. Many advantages, such as wide cruise range, less power consumption, low noise, and no pollution, make the underwater glider an important platform for marine environment observation and ocean resource exploration. For the wide cruise range, attitude control of underwater glider becomes the core technology. In this paper, the underwater glider named OUC-III has been developed for marine observation. To control the attitude of glider, the kinematic and dynamic models of it have been calculated by mathematical analysis. Furthermore, a novel control algorithm is proposed to control the attitude of glider. The algorithm is combined reinforcement learning with Active Disturbance Rejection Control (ADRC) and compared with classical ADRC by simulation based on the dynamic model of OUC-III. The simulation experimental results indicate that the proposed algorithm compensates well for the ocean current disturbances on OUC-III attitude control mission and it obtains high-precision and high-adaptive control ability.