Paper detail

Finite-Memory Strategy Synthesis for Robust Multidimensional Mean-Payoff Objectives

Two-player games on graphs provide the mathematical foundation for the study of reactive systems. In the quantitative framework, an objective assigns a value to every play, and the goal of player 1 is to minimize the value of the objective. In this framework, there are two relevant synthesis problems to consider: the quantitative analysis problem is to compute the minimal (or infimum) value that player 1 can assure, and the boolean analysis problem asks whether player 1 can assure that the value of the objective is at most $ν$ (for a given threshold $ν$). Mean-payoff expression games are played on a multidimensional weighted graph. An atomic mean-payoff expression objective is the mean-payoff value (the long-run average weight) of a certain dimension, and the class of mean-payoff expressions is the closure of atomic mean-payoff expressions under the algebraic operations of $\MAX,\MIN$, numerical complement and $\SUM$. In this work, we study for the first time the strategy synthesis problems for games with robust quantitative objectives, namely, games with mean-payoff expression objectives. While in general, optimal strategies for these games require infinite-memory, in synthesis we are typically interested in the construction of a finite-state system. Hence, we consider games in which player 1 is restricted to finite-memory strategies, and our main contribution is as follows. We prove that for mean-payoff expressions, the quantitative analysis problem is computable, and the boolean analysis problem is inter-reducible with Hilbert's tenth problem over rationals --- a fundamental long-standing open problem in computer science and mathematics.