Convex Optimization for Parameter Synthesis in MDPs
IEEE Transactions on Automatic Control, vol. 67, pp. 6333–6348
Abstract
Probabilistic model-checking aims to prove whether a Markov decision process (MDP) satisfies a temporal logic specification. The underlying methods rely on an often unrealistic assumption that the MDP is precisely known. Consequently, parametric MDPs (pMDPs) extend MDPs with transition probabilities that are functions over unspecified parameters. The parameter synthesis problem is to compute an instantiation of these unspecified parameters such that the resulting MDP satisfies the temporal logic specification. We formulate the parameter synthesis problem as a quadratically constrained quadratic program, which is nonconvex and is NP-hard to solve in general. We develop two approaches that iteratively obtain locally optimal solutions. The first approach exploits the so-called convex–concave procedure (CCP), and the second approach utilizes a sequential convex programming (SCP) method. The techniques improve the runtime and scalability by multiple orders of magnitude compared to black-box CCP and SCP by merging ideas from convex optimization and probabilistic model-checking. We demonstrate the approaches on a satellite collision avoidance problem with hundreds of thousands of states and tens of thousands of parameters and their scalability on a wide range of commonly used benchmarks.
Authors 5
-
The University of Texas at Austin
Affiliation as printed
Department of Aerospace Engineering and Engineering Mechanics, University of Texas at Austin, Austin, TX, USA
University of Texas at Austin
-
Radboud University Nijmegen · The University of Texas at Austin
Affiliation as printed
Institute for Computing and Information Science, Radboud University Nijmegen, Nijmegen, The Netherlands
University of Texas at Austin
-
Affiliation as printed
Institute for Computing and Information Science, Radboud University Nijmegen, Nijmegen, The Netherlands
Radboud University Nijmegen
-
RWTH Aachen University · University of California, Berkeley
Affiliation as printed
Departement of Computer Science, RWTH Aachen University, Aachen, Germany
University of California–Berkeley
-
Ufuk Topcu Aachen
The University of Texas at Austin · RWTH Aachen University
Affiliation as printed
Department of Aerospace Engineering and Engineering Mechanics, University of Texas at Austin, Austin, TX, USA
RWTH Aachen University
Cited by 21 stored of 21
No patents citing this paper on Lens.org (checked 2026-10-06).