Skip to main content

discounted-mpc-plant-model-mismatch

Discounted MPC under plant-model mismatch - stability and suboptimality analysis for infinite-horizon optimal control with surrogate models. Activation: MPC, model predictive control, plant-model mismatch, robustness, stability, discounted control.

Ir para a instalação

Informações da origem

Repositório
hiyenwong/ai_collection
Última atividade na origem
4 de junho de 2026 às 13:32
Idioma detectado do SKILL.md
chinês
Estrelas
2
Forks
0

Opções de instalação

Por padrão, está selecionado o prompt que primeiro revisa a origem. Você pode mudar para um comando direto ou baixar uma cópia local.

Revise os arquivos de origem

Leia o SKILL.md e os arquivos complementares exibidos pelo SkillsMP antes de decidir se vai instalar.

Exibindo SKILL.md

SKILL.md
Instruções da origem · Visualização somente leitura
name
discounted-mpc-plant-model-mismatch
description
Discounted MPC under plant-model mismatch - stability and suboptimality analysis for infinite-horizon optimal control with surrogate models. Activation: MPC, model predictive control, plant-model mismatch, robustness, stability, discounted control.
# Discounted MPC under Plant-Model Mismatch ## Paper Information **Title:** Discounted MPC and infinite-horizon optimal control under plant-model mismatch: Stability and suboptimality **Authors:** Robert H. Moldenhauer, Karl Worthmann, Romain Postoyan, Dragan Nešić, Mathieu Granzotto **arXiv:** https://arxiv.org/abs/2604.08521v1 **Published:** 2026-04-09 **Category:** math.OC (Optimization and Control), eess.SY (Systems and Control) ## Core Problem 如何在使用**替代模型**(surrogate model)求解模型预测控制(MPC)和无限时域最优控制问题时,保证闭环系统的**稳定性**和**次优性**,即使替代模型与真实被控对象之间存在不匹配? ## Key Contributions ### 1. Unified Framework - 基于**二次型代价**的统一框架 - 同时分析有限时域和无限时域问题 - 涵盖折扣(discounted)和非折扣(undiscounted)场景 ### 2. Stability Guarantees **假设条件:** - 植物与模型之间的不匹配与状态和控制成比例 - 原点保持为平衡点 - 模型连续且代价可控 **结论:** - 在上述条件下,闭环系统**指数稳定** - 鲁棒性保证对时域长度一致,即更长的时域不要求更小的模型不匹配 ### 3. Suboptimality Bounds - 给出了闭环代价的次优性界 - 恢复了替代模型的最优代价 - 揭示了**时域长度、折扣因子和模型不匹配之间的权衡关系** ## Mathematical Framework ### Plant-Model Mismatch Model ``` 真实系统: x+ = f(x, u) + Δf(x, u) 替代模型: x+ = g(x, u) ``` 其中 `Δf(x,u)` 表示模型不匹配,假设满足: ``` ||Δf(x, u)|| ≤ σ_x||x|| + σ_u||u|| ``` ### Stability Analysis 关键工具: - Lyapunov 函数 - 输入到状态稳定性 (ISS) - 小增益定理 ### Suboptimality Analysis ``` V_N(x) ≤ V_∞*(x) + bound(ε, N, γ) ``` 其中: - `V_N` 是 N 步 MPC 代价 - `V_∞*` 是无限时域最优代价 - `ε` 是模型不匹配参数 - `γ` 是折扣因子 ## Key Insights ### 1. Horizon Robustness **重要发现:** 更长的预测时域不需要更小的模型不匹配来维持稳定性 - 这与直觉相反 - 提供了实际设计的灵活性 ### 2. Discount Factor Trade-off 折扣因子的影响: - **较小的折扣因子:** 更关注近期性能,但对模型误差更敏感 - **较大的折扣因子:** 更接近无限时域,需要更精确的模型 ### 3. Model Mismatch Bounds 模型不匹配的可容忍范围: - 与代价可控性相关 - 与系统稳定性裕度相关 - 可以显式计算 ## Practical Implications ### 1. Controller Design 使用替代模型时的设计指南: 1. 确保模型连续性和代价可控性 2. 验证模型不匹配在允许范围内 3. 根据性能要求选择合适的时域长度 4. 考虑折扣因子对鲁棒性的影响 ### 2. Model Reduction 该框架可以用于: - 降阶模型的设计 - 近似模型的选择 - 计算复杂度与性能的权衡 ### 3. Adaptive Control 为自适应 MPC 提供理论基础: - 在线模型更新的稳定性保证 - 模型改进的量化指标 ## Related Work - Classic MPC stability: Mayne et al. (2000) - Robust MPC: Bemporad & Morari (1999) - Model mismatch:近来更多关注,但缺乏系统性框架 ## Limitations & Future Directions ### Current Limitations 1. 假设原点为平衡点 (可扩展到跟踪问题) 2. 二次型代价 (可扩展到一般凸代价) 3. 比例型的模型不匹配 (可考虑更一般的形式) ### Future Research 1. 非线性系统的扩展 2. 约束处理 3. 分布式 MPC 中的应用 4. 学习型 MPC 中的应用 ## Key Equations ### Stability Condition ``` α1(||x||) ≤ V_N(x) ≤ α2(||x||) V_N(f(x, κ_N(x))) - V_N(x) ≤ -α3(||x||) + ε·σ(x) ``` ### Suboptimality Bound ``` J(x, κ_N) ≤ J*(x) · (1 + δ(ε, N)) ``` 其中 `δ(ε, N)` 是关于模型不匹配 ε 和时域 N 的函数。 ## Code Example (Conceptual) ```python class DiscountedMPC: def __init__(self, model, horizon, discount): self.model = model # surrogate model self.N = horizon self.gamma = discount def solve(self, x0): """Solve MPC with surrogate model""" # Optimize over control sequence u_opt = self._optimize(x0) # Apply first control return u_opt[0] def check_stability(self, mismatch_bound): """Verify stability condition""" # Check if mismatch is within allowable range return mismatch_bound < self._compute_tolerance() ``` ## References 1. Mayne, D. Q., et al. (2000). Constrained model predictive control: Stability and optimality. Automatica. 2. Rawlings, J. B., & Mayne, D. Q. (2017). Model predictive control: Theory and design. 3. Grüne, L., & Pannek, J. (2017). Nonlinear model predictive control. ## Summary 这篇论文为**模型不匹配下的 MPC** 提供了一个统一的理论框架,证明了即使使用不完美的替代模型,也能保证闭环系统的稳定性和可量化的次优性。关键贡献是揭示了时域长度、折扣因子和模型不匹配之间的权衡关系,为实际控制器设计提供了理论基础。 该工作是控制系统理论和系统工程的重要进展,特别是在以下场景中具有重要应用: - 复杂系统的降阶控制 - 自适应和学习的 MPC - 计算资源受限的实时控制 - 模型不确定性较大的系统 --- _Skill generated from arXiv:2604.08521v1 on 2026-04-10_
Ver no GitHub