基于RBF神经网络自适应滑模的机器鱼运动控制

Motion control of robotic fish based on RBF neural network adaptive sliding mode

针对水中扰流、波浪等外界环境扰动下机器鱼空间运动控制问题,以灵活的3自由度胸鳍推进仿生机器鱼为研究对象,在设定胸鳍最优运动轨迹的基础上,将控制输入简化为拍翼角相位差,并建立仿生机器鱼在扰动条件下的水动力学模型。由于流体运动的复杂性以及机器鱼自身变形对动力学参数的影响,提出了RBF神经网络自适应滑模的机器鱼控制方法,采用RBF神经网络对动力学模型中的不确定项和水流干扰项进行在线逼近,并结合自适应滑模控制器对控制参数进行实时调节,通过李亚普诺夫稳定性定理证明所设计控制器具有渐近稳定性。实验结果表明机器鱼在螺旋下潜运动时,仅在初始阶段具有较大的误差,约8 s后机器鱼的运动轨迹和期望轨迹基本吻合,验证了所述方法的有效性。

To address the problem of controlling the spatial motion of robotic fish under the disturbance of external environment such as water disturbance and waves,this paper uses a flexible 3-degree-of-freedom pectoral fin propulsion biomimetic robotic fish as the research objects.Based on setting the optimal motion trajectory of the pectoral fin,the control input is simplified to the phase difference of the flapping wing angle,and a hydrodynamic model of the biomimetic robotic fish under the disturbance conditions is established.Due to the complexity of fluid motion and the influence of the robotic fish deformation on the kinetic parameters,the RBF neural network adaptive sliding mode control method of the robotic fish is proposed.The RBF neural network is used for online approximation of the uncertainty terms and the water flow disturbance terms in the kinetic model,and the adaptive sliding mode controller is combined to adjust the control parameters in real time.Then,the asymptotic stability of the designed controller is proved by Liapunov’s stability theorem.The experimental results show that the robotic fish have a large error only in the initial stage during the spiral downward motion,and both their motion trajectory and the desired trajectory basically match after about 8 s,which verifies the effectiveness of the method described in this paper.