Research On Attitude Observation And Heading Control Of Wave Glider

The "Black Pearl" Wave Glider(WG)is a zero-emission,pollution-free autonomous long-range ocean unmanned observation platform.The float and the submerged glider are connected by the umbilical cable to form the unique double-body structure of the WG.The core content of the marine unmanned observation platform control is to achieve accurate heading control.However,the high-precision heading control is particularly difficult due to the complex and changeable marine environment.In this paper,researches on the above-mentioned problems are carried out.Firstly,a multisensor data fusion framework for the WG attitude observation is constructed,and an improved covariance matching adaptive Unscented Kalman Filter(UKF)algorithm is proposed.Secondly,a fuzzy adaptive PID heading control algorithm based on Q?learning reinforcement learning is proposed and the WG closed-loop heading control system is built.Finally,it is verified by simulation and ocean experiments.The main research contents of the paper are as follows:Firstly,the two-body attitude coordinate system and the six-degree-of-freedom dynamic model of the "Black Pearl" WG are established.According to the attitude coordinate system,the accelerometer and the magnetometer are used for preliminary calculate attitude of the WG.In the calculation process,and the inclination complementary fusion algorithm is used to effectively make up for the shortcomings of the magnetometer,which has strict requirements on the installation plane.At the same time,the gyroscope sensor is used to calculate the attitude,and the two calculation results are analyzed.Secondly,a multi-sensor two-layer data fusion framework is constructed for the defects of the gyroscope attitude calculation error accumulation and the complementary solution of magnetometer and accelerometer with a large amount of noise.The fusion algorithm of the top-level fusion framework is studied in this paper.The unscented Kalman filter(UKF)algorithm is introduced to realize the optimal attitude estimation.However,the classical UKF algorithm has the defects of lack of prior knowledge of the noise,which leads to declining the accuracy estimation or even diverged.The covariance matching technology is introduced to realize the estimation and adjustment of the online noise,and the accurate attitude estimation of the WG can be realized.Thirdly,aiming at the problem of low precision of classical PID course control in "Black Pearl" WG,an adaptive fuzzy PID heading control algorithm based on the Q?learning reinforcement learning is proposed,which can realize the adaptive process and online adjustment of the classical PID control parameters,and be applied to timevarying nonlinear systems.According to the heading control algorithm,a closed-loop control system of the WG is constructed to realize the high-precision heading control.Finally,the multi-sensor attitude observation of the WG was verified by simulation experiment and ocean experiment,and the course control algorithm was simulated.The results show that the proposed multi-sensor attitude observation framework and the heading control algorithm based on reinforcement learning can achieve more precise attitude observation and heading control.