基于异步灵敏度的分布式非线性模型预测控制

Numerical experiments with multistep model-predictive control approaches and sensitivity updates for the tracking control of cars

The paper discusses multistep nonlinear model-predictive control (NMPC) schemes for the tracking of a car model along a given reference track. In particular we will compare the numerical performance and robustness of classic single step NMPC, multistep NMPC without re-optimization, multistep NMPC with re-optimization, and multistep NMPC with sensitivity updates instead of a full re-optimization.

Performance bounds for NMPC combined with Sensitivity Updates

In this paper we present a stability proof of model predictive control without stabilizing terminal constraints of cost which are subject to unknown but measurable disturbances. To this end, a relaxed Lyapunov argument on the nominal system and Lipschitz conditions on the open loop change of the optimal value function and the stage costs are employed. Based on the special case of sensitivity analysis, we show that Lipschitz assumptions are satisfied if a sensitivity update can be performed along the closed loop solution. To illustrate our approach we present a halfcar example and show performa…

Performance of Sensitivity based NMPC Updates in Automotive Applications

In this work we consider a half car model which is subject to unknown but measurable disturbances. To control this system, we impose a combination of model predictive control without stabilizing terminal constraints or cost to generate a nominal solution and sensitivity updates to handle the disturbances. For this approach, stability of the resulting closed loop can be guaranteed using a relaxed Lyapunov argument on the nominal system and Lipschitz conditions on the open loop change of the optimal value function and the stage costs. For the considered example, the proposed approach is realtime…

ASAP-MPC：一种用于非线性模型预测控制在线运动规划的异步更新方案

ASAP-MPC: An Asynchronous Update Scheme for Online Motion Planning with Nonlinear Model Predictive Control

针对NMPC运动规划求解时间难以满足实时性的问题，本文提出ASAP-MPC异步更新框架，结合线性状态反馈实现轨迹无缝融合，并在四旋翼和卡车-拖车实验中验证其嵌入式适用性。

PANDA: 一种基于近端平均拟牛顿与自适应线搜索算法的无矩阵可微非线性模型预测控制求解器

PANDA: A Matrix-Free Differentiable NMPC Solver via Proximal Averaged Quasi-Newton with Adaptive Linesearch Algorithm

研究针对可微NMPC的计算与内存成本问题，提出PANDA无矩阵求解器，通过前向的近端梯度迭代加拟牛顿加速、反向的隐式求导及Krylov子空间迭代等方法，在非凸拖车NMPC任务上实现更快计算、更低内存开销且保持模仿学习性能。

基于ESDIRK的随机微分代数方程非线性模型预测控制

ESDIRK-based nonlinear model predictive control for stochastic differential-algebraic equations

提出基于ESDIRK的NMPC算法，结合CD-EKF和OCP求解指标1随机DAEs系统，通过多重打靶法离散化并利用内部数值微分计算灵敏度，在碱性电解槽堆栈模型中实现温度时变设定点跟踪。