Robust Multidimensional Mean-Payoff Games are Undecidable
This result shows that a central problem in quantitative synthesis and verification is fundamentally unsolvable, which is a foundational negative result for the field.
The paper tackles the problem of determining the winner in robust multidimensional mean-payoff games with arbitrary strategies for both players, and proves that this problem is undecidable.
Mean-payoff games play a central role in quantitative synthesis and verification. In a single-dimensional game a weight is assigned to every transition and the objective of the protagonist is to assure a non-negative limit-average weight. In the multidimensional setting, a weight vector is assigned to every transition and the objective of the protagonist is to satisfy a boolean condition over the limit-average weight of each dimension, e.g., $\LimAvg(x_1) \leq 0 \vee \LimAvg(x_2)\geq 0 \wedge \LimAvg(x_3) \geq 0$. We recently proved that when one of the players is restricted to finite-memory strategies then the decidability of determining the winner is inter-reducible with Hilbert's Tenth problem over rationals (a fundamental long-standing open problem). In this work we allow arbitrary (infinite-memory) strategies for both players and we show that the problem is undecidable.