Fundamentals of economic model predictive control
James B. Rawlings et al., CDC 2012
- progress: 10%
- The first paper introduce economic model predictive control
difference between standard MPC and economic MPC:
- Take economic objective directly as the objective function of the control system. (What is an economic objective?)
- Standard MPC: cost function
for all admissible where is the stage cost, is set of admissible states and is the set of admissible inputs. The stage cost need not bear any functional relationship to the operating cost of the plant and, in fact, is usually chosen to be quadartic. The weights to the quadratic objective can be used to prioritize target tracking on some states more than others or change the size of the control action for a given tracking error. - Economic MPC: the operating cost of the plant is used directly as the stage cost in the MPC objective function.
The cost of operating the plant need not be zero at the optimal steady-state operating conditions. The standard MPC cost function cannot hold in general. More importantly, it may happen thatfor some feasible pair that is not a steady state.(setction III)
Deep Learning Tubes for Tube MPC
David D. Fan et al., RSS2020
- Video
- 存在的问题：传统的find robust bounds的方法是给
中的扰动设定上下界，如 ，但是过于保守，未考虑时间和空间的变化怼扰动的影响。解决方案是扩展 使之time and state-varying，如一种常用的方法是 - 其中一种解决方案是使用带参数的概率（如高斯分布）对系统进行建模（用概率跳过对
的直接建模）
- 其中一种解决方案是使用带参数的概率（如高斯分布）对系统进行建模（用概率跳过对
- 主要工作：通过deep learning直接从数据中学习出Tube MPC的disturbances的constraints
Robust model predictive control of constrained linear systems with bounded disturbances
D.Q. Mayne et al., Automatica 2005
- 系统特点：constrained, linear, discrete-time systems in the presence of bounded disturbances 如何处理扰动有上下界
- 实现代码示例：Github
- Robust model predictive control using tubes