基于Koopman双线性的软体执行器模型预测控制研究

Research on Model Predictive Control of Soft Actuator Based on Koopman Bilinear Model

为了实现对软体执行器的精确控制,使用了一种基于Koopman双线性的建模及其模型预测控制(MPC)的方法。该方法在获得明确的面向控制的模型的同时避免了传统建模方法物理上的简化假设。为证明Koopman双线模型对于软体执行器的控制方面的优势,建立了软体执行器Koopman双线性模型、Koopman线性模型和状态空间线性模型3种模型并在轨迹跟踪任务中,将3种模型和MPC算法进行结合。仿真和实验结果一致表明,Koopman双线性模型的MPC控制器策略跟踪性能优于同一系统Koopman线性模型的MPC控制器和线性状态空间模型的MPC控制器。

To achieve accurate control of the soft actuator, a Koopman bilinear-based modeling and its Model Predictive Control(MPC) approach is used.The method avoids the physically simplifying assumptions of traditional modeling methods while obtaining an explicit control-oriented model.To demonstrate the advantages of the Koopman bilinear model for the control aspects of the soft actuator, three models of the soft actuator Koopman bilinear, Koopman linear and state space linear models are developed and combined with the MPC algorithm in a trajectory tracking task.Simulation and experimental results consistently show that the MPC controller with Koopman bilinear model outperforms the MPC controller with Koopman linear model and MPC controller with linear state space model for the same system.