Efficient Configuration-Constrained Tube MPC via Variables Restriction and Template Selection
Filippo Badalamenti, Sampath Kumar Mulagaleti, Mario Eduardo Villanueva · 2025
Read Efficient Configuration-Constrained Tube MPC via Variables Restriction and Template Selection by Filippo Badalamenti, Sampath Kumar Mulagaleti, Mario Eduardo Villanueva (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-14440.txt.
Introduction
Model predictive control steers a system by solving an optimization problem many times a second, and flexible versions of it can be too slow for real use. Filippo Badalamenti and co-authors tackle that cost for a tube-based variant that wraps the planned path in a safety region with adjustable shape. Their first fix deliberately limits the free choices in each solve, which cuts computing time sharply while keeping stability guarantees.
Their second fix is a routine that refines the shape templates by solving a series of small quadratic programs, balancing accuracy against bulk. Tests on a benchmark and on a ten-state system show strong robust performance with little computing overhead, keeping the flexible controller viable in real time.
Catalog record
| CREATOR | Filippo Badalamenti, Sampath Kumar Mulagaleti, Mario Eduardo Villanueva |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-14440
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Efficient Configuration-Constrained Tube MPC via Variables Restriction and Template Selection Thanks: F. Badalamenti (Corresponding Author, [email protected] ), S.K. Mulagaleti, M.E. Villanueva and A. Bemporad are with IMT School for Advanced Studies Lucca, Piazza San Francesco 19, 55100, Italy. B. Houska is with ShanghaiTech University, China. This work has received support from the European Research Council (ERC), Advanced Research Grant COMPACT (Grant Agreement No. 101141351) Filippo Badalamenti, Sampath Kumar Mulagaleti, Mario Eduardo Villanueva Abstract Configuration-Constrained Tube Model Predictive Control (CCTMPC) offers flexibility by using a polytopic parameterization of invariant sets and the optimization of an associated vertex control law. This flexibility, however, often demands computational trade-offs between set parameterization accuracy and optimization complexity. This paper proposes two innovations that help the user tackle this trade-off. First, a structured framework is proposed, which strategically limits optimization degrees of freedom, significantly reducing online computation time while retaining stability guarantees. This framework aligns with Homothetic Tube MPC (HTMPC) under maximal constraints. Second, a template refinement algorithm that iteratively solves quadratic programs is introduced to balance polytope complexity and conservatism. Simulation studies on an illustrative benchmark problem as well as a high-dimensional ten-state system demonstrate the approach’s efficiency, achieving robust performance with minimal computational overhead. The results validate a practical pathway to leveraging CCTMPC’s adaptability without sacrificing real-time viability. 1 Introduction Tube Model Predictive Control (TMPC) has proven to be a central tool in robust control due to its ability to enclose all realizations of the state of an uncertain system within a sequence of bounded sets [ 9 , 11 , 18 ] . Among the various set parameterizations, a variety of polytopic set parameterization approaches are particularly attractive, because they permit the online solution of a Quadratic Programming (QP) problem [ 8 ] . Examples for such polytopic schemes include Rigid [ 13 , 15 ] , Homothetic [ 16 ] , and Elastic [ 17 , 3 ] Tube MPC. While these methods typically enforce set invariance through linear feedback laws, previous works have also explored the use of a freely parameterized control law within the Homothetic framework [ 12 ] . Departing from conventional parameterization methods, Configuration-Constrained Tube MPC (CCTMPC) [ 20 ] introduces a novel framework centered on selecting polytopic sets with fixed facet orientations that share a predefined geometric structure. This approach optimizes facet displacements while explicitly enumerating the vertices of the polytopes and their corresponding control inputs, inherently supporting systems with multiplicative uncertainty. Despite its merits, CCTMPC presents dual challenges. First, its formulation inherently involves a larger number of optimization variables—encompassing both control inputs and set geometry—compared to methods that predefine either component. This complexity inevitably grows with system dimensionality. Second, the control scheme’s efficacy hinges critically on the template design, particularly in high-dimensional spaces. Here, achieving a balance between template expressiveness and computational tractability becomes nontrivial, as overly simplistic templates risk conservatism, while intricate ones incur prohibitive online costs. Contribution: This paper advances the state-of-the-art in CCTMPC through two novel methods that enable computational efficiency and template design flexibility. Here, we first introduce a novel family of Tube MPC controllers that systematically reduce computational overhead. By strategically constraining the number of optimization variables, our approach transitions from the full flexibility of CCTMPC to a highly efficient scheme reminiscent of Homothetic Tube MPC (HTMPC). Moreover, we propose an iterative algorithm for designing configuration templates. This algorithm is initialized with a low-complexity template. It progressively enriches the template by solving QPs, systematically increasing the complexity until a user-defined target is achieved. Outline : The paper is structured as follows: • Section 2 formalizes the problem and establishes the necessary mathematical framework. • Section 3 derives the proposed family of reduced-complexity Tube MPC schemes, discussing their theoretical properties. • Section 4 details the adaptive template refinement procedures and NLP formulation. • Section 5 validates the approach through numerical simulations, including numerical results for a 10-state quadcopter model. • Section 6 summarizes the contributions and outlines directions for future research. Notation : The symbol \mathbb{I}_{n} denotes the identity matrix of order n . A block-diagonal matrix composed of matrices Q_{1},\dots,Q_{n} is denoted by \mathrm{blkd}(Q_{1},\dots,Q_{n}) . The operator \operatorname{convh} denotes the convex hull, and \left\lVert z\right\rVert_{Q}^{2}\coloneqq z^{\top}Qz . Given compact sets \mathcal{A}\subseteq\mathcal{B}\subset\mathbb{R}^{n} , the operator \mathrm{d}_{\mathcal{B}}(\mathcal{A}) measures the 2 -norm Hausdorff distance between the sets. 2 Tube Model Predictive Control This section briefly reviews the main idea of Tube MPC and introduces the notation used throughout this article. 2.1 Uncertain Linear Systems This paper is concerned with uncertain discrete-time systems of the form \displaystyle x_{k+1}=A_{k}x_{k}+B_{k}u_{k}+w_{k}, (1) where x_{k}\in\mathcal{X}\subseteq\mathbb{R}^{n_{x}} and u_{k}\in\mathcal{U}\subseteq\mathbb{R}^{n_{u}} denote the state and input at time k\in\mathbb{N} , while \mathcal{X} and \mathcal{U} are given, closed and convex, state and control constraint sets. Moreover, w_{k}\in\mathcal{W} is an additive disturbance, with \mathcal{W} assumed to be compact and convex. The matrices A_{k} and B_{k} are uncertain, but known to satisfy (A_{k},B_{k})\in\Delta\coloneqq\operatorname{convh}((A_{1},B_{1}),\kern-0.50003pt.\kern-0.09995pt.\kern-0.09995pt.\kern 0.09995pt,(A_{\mathrm{m}},B_{\mathrm{m}})). Here, the matrices (A_{i},B_{i}) are assumed to be given vertices of the matrix polytope \Delta that models the uncertainty set. 2.2 Robust Control Tubes Let \mathcal{F}(X) denote the set of sets that can be reached from a given set X\subseteq\mathbb{R}^{n_{x}} ; that is, \displaystyle\mathcal{F}(X):=\left\{X^{+}\subseteq\mathbb{R}^{n_{x}}\ \middle|\ \begin{aligned} &\forall\,x\in X,\ \exists\,u\in\mathcal{U}:\vskip 1.0pt\\ &\forall\,(A,B)\in\Delta,\;w\in\mathcal{W},\vskip 1.0pt\\ &Ax+Bu+w\in X^{+}\end{aligned}\right\}. We recall that a sequence of sets X_{0},X_{1},\ldots\subseteq\mathcal{X} is called a feasible Robust Control Tube (RCT) if \displaystyle\forall k\in\{0,\dots,N-1\},\quad X_{k+1}\in\mathcal{F}(X_{k}). (2) Moreover, a set X\subseteq\mathcal{X} is called a feasible Robust Control Invariant (RCI) set if (X,X) is an RCT. 2.3 Optimization of RCTs From a conceptual perspective, the main idea of Tube MPC is to optimize an RCT online. Here, the first element of the RCT is enforced to contain the current state measurement. For example, one can solve optimization problems of the form \displaystyle\begin{array}[]{cl}\displaystyle\min_{X_{0},\ldots,X_{N}}&L_{0}(X_{0})+\displaystyle\sum_{k=0}^{N-1}L(X_{k})+L_{N}(X_{N})\\ \text{s.t.}&\left\{\begin{array}[]{l}\forall k\in\{0,\dots,N-1\},\\ X_{k+1}\in\mathcal{F}(X_{k}),\\ x\in X_{0},\ X_{k}\subseteq\mathcal{X}.\end{array}\right.\end{array} Here, L_{0} denotes the initial cost, L the stage cost, L_{N} the terminal cost, and x the current state measurement. Choices for the cost triple (L_{0},L,L_{N}) that ensure closed-loop stability of such Tube MPC schemes are discussed in [ 19 ] . Alternate formulations may also introduce control-law penalties designed to regularize the control policies, which are hidden in our definition of \mathcal{F} ; for an in-depth discussion, see [ 8 ] . 3 Polytopic Tube MPC Schemes As the optimization variables of the generic Tube MPC controller ( 2.3 ) are set-valued, this problem is in general intractable. Therefore, this section focuses on polytopic set parameterizations. 3.1 Configuration-Constrained Polytopes Throughout this paper, we use the notation \displaystyle P(y)\coloneqq\{x\in\mathbb{R}^{n_{x}}\mid Fx\leq y\}, (7) to denote polyhedra with a given facet matrix F\in\mathbb{R}^{\mathsf{f}\times n_{x}} . Here, we assume that Fx\leq 0 implies x=0 , so that P(y) is bounded—and hence potentially empty—for each y\in\mathbb{R}^{\mathsf{f}} . In the following, we further refer to the cone \mathcal{E}:=\{y\in\mathbb{R}^{\mathsf{f}}\mid Ey\leq 0\} as a configuration cone for P if the face configuration of the parameterized polytope P(y) remains invariant for all y\in\mathcal{E} . This means that if \mathcal{E} is a configuration cone, we can find a set of matrices V:=\{V_{j}\mid j\in\{1,\cdots,\mathrm{v}\}\} such that \displaystyle P(y)=\mathrm{convh}(\{V_{j}y\mid j\in\{1,\cdots,\mathrm{v}\}\}) (8) for all y\in\mathcal{E} . As already mentioned in the introduction, constructing suitable configuration triples (F,E,V) with the above properties is a non-trivial task. One possible method for constructing such triples can nevertheless be found in [ 20