[Paper Review] Strategy Synthesis for Multi-dimensional Quantitative Objectives

Krishnendu Chatterjee, Mickaël Randour+1 more|arXiv (Cornell University)|Jan 24, 2012

Formal Methods in Verification38 references4 citations

TL;DR

This paper presents tight exponential bounds on finite-memory strategy requirements for multi-dimensional quantitative games with parity objectives, introduces an optimal symbolic and incremental algorithm for strategy synthesis, and provides a precise characterization of when finite memory can be traded for randomness. The key contribution is a significant improvement over prior triple-exponential bounds for multi-energy games, establishing matching upper and lower bounds for memory usage in both mean-payoff and energy games with parity conditions.

ABSTRACT

Multi-dimensional mean-payoff and energy games provide the mathematical foundation for the quantitative study of reactive systems, and play a central role in the emerging quantitative theory of verification and synthesis. In this work, we study the strategy synthesis problem for games with such multi-dimensional objectives along with a parity condition, a canonical way to express $ω$-regular conditions. While in general, the winning strategies in such games may require infinite memory, for synthesis the most relevant problem is the construction of a finite-memory winning strategy (if one exists). Our main contributions are as follows. First, we show a tight exponential bound (matching upper and lower bounds) on the memory required for finite-memory winning strategies in both multi-dimensional mean-payoff and energy games along with parity objectives. This significantly improves the triple exponential upper bound for multi energy games (without parity) that could be derived from results in literature for games on VASS (vector addition systems with states). Second, we present an optimal symbolic and incremental algorithm to compute a finite-memory winning strategy (if one exists) in such games. Finally, we give a complete characterization of when finite memory of strategies can be traded off for randomness. In particular, we show that for one-dimension mean-payoff parity games, randomized memoryless strategies are as powerful as their pure finite-memory counterparts.

Motivation & Objective

• To establish tight exponential bounds on the memory required for finite-memory winning strategies in multi-dimensional quantitative games with parity objectives.
• To develop an optimal symbolic and incremental algorithm for computing finite-memory winning strategies in such games.
• To characterize the trade-off between finite memory and randomness in strategies for different classes of games.
• To improve upon the prior triple-exponential upper bound for multi-energy games by deriving matching lower bounds.

Proposed method

• The authors prove matching upper and lower bounds on memory requirements using game-theoretic analysis and reductions to subgames.
• They introduce a symbolic and incremental algorithm that computes finite-memory winning strategies efficiently, leveraging on-the-fly computation and state-space pruning.
• The approach uses induction on priority levels in parity games to recursively construct winning strategies in subgames.
• The characterization of memory vs. randomness relies on probabilistic reasoning and analysis of achievable payoff vectors under different strategy types.
• They employ game decomposition techniques, isolating winning regions and analyzing subgames with reduced priority sets.
• Theoretical results are validated through a case study on a one-player game with two-dimensional mean-payoff, demonstrating strict superiority of randomized finite-memory strategies.

Experimental results

Research questions

• RQ1 What is the exact memory complexity of finite-memory winning strategies in multi-dimensional mean-payoff and energy games with parity objectives?
• RQ2 Can an optimal symbolic and incremental algorithm be designed to synthesize such strategies efficiently?
• RQ3 In which classes of games can randomized memoryless strategies achieve the same winning power as pure finite-memory strategies?
• RQ4 Is there a strict hierarchy in expressive power between pure finite-memory, randomized memoryless, and randomized finite-memory strategies?
• RQ5 What is the precise trade-off between memory and randomness in terms of winning capabilities across different game types?

Key findings

• The paper establishes a tight exponential bound (matching upper and lower bounds) on the memory required for finite-memory winning strategies in both multi-dimensional mean-payoff and energy games with parity objectives.
• The proposed symbolic and incremental algorithm achieves optimal complexity and has been successfully integrated into the Acacia+ synthesis tool, demonstrating practical efficiency.
• For one-dimensional mean-payoff parity games, randomized memoryless strategies are as powerful as pure finite-memory strategies, enabling a complete trade-off between memory and randomness.
• In two-player multi-mean-payoff games, randomized memoryless strategies are strictly less powerful than pure finite-memory strategies, indicating that memory remains essential in such settings.
• Randomized finite-memory strategies are strictly more powerful than both randomized memoryless and pure finite-memory strategies in multi-mean-payoff games, even in the one-player case, due to their ability to achieve exact linear combinations of payoff vectors.
• A counterexample is provided to show that randomized finite-memory strategies can achieve a payoff vector (0,0) that is unattainable by any pure finite-memory or randomized memoryless strategy, proving strict superiority.