[Paper Review] A Model for Probabilistic Reasoning on Assume/Guarantee Contracts

Benoît Delahaye, Benoı̂t Caillaud|ArXiv.org|Nov 7, 2008

Software Reliability and Analysis Research10 references3 citations

TL;DR

This paper presents a compositional probabilistic extension of Assume/Guarantee contracts for modular reliability analysis in component-based systems. By modeling environmental inputs as probabilistic rather than non-deterministic, it introduces probabilistic satisfaction, refinement, and composition relations that preserve modularity, enabling top-down or bottom-up reliability verification with quantitative guarantees on system behavior under uncertainty.

ABSTRACT

In this paper, we present a probabilistic adaptation of an Assume/Guarantee contract formalism. For the sake of generality, we assume that the extended state machines used in the contracts and implementations define sets of runs on a given set of variables, that compose by intersection over the common variables. In order to enable probabilistic reasoning, we consider that the contracts dictate how certain input variables will behave, being either non-deterministic, or probabilistic; the introduction of probabilistic variables leading us to tune the notions of implementation, refinement and composition. As shown in the report, this probabilistic adaptation of the Assume/Guarantee contract theory preserves compositionality and therefore allows modular reliability analysis, either with a top-down or a bottom-up approach.

Motivation & Objective

• To address the lack of modular, scalable reliability analysis techniques in component-based design of complex embedded systems.
• To formalize reliability requirements using contracts that distinguish assumptions about the environment from guarantees about component behavior.
• To extend classical Assume/Guarantee reasoning with probabilistic semantics to support quantitative reliability analysis.
• To ensure that probabilistic refinement and composition operations preserve modularity for both top-down and bottom-up verification.
• To enable reasoning about system reliability by combining probabilistic environmental assumptions with behavioral implementations.

Proposed method

• The formalism models components and their environments using extended state machines that define sets of runs over shared variables, with composition via intersection over common variables.
• Probabilistic contracts are defined by specifying probabilistic assumptions on input variables, allowing both non-deterministic and probabilistic behaviors in the environment.
• A probabilistic satisfaction relation is defined such that an implementation satisfies a contract if its behavior meets the contract's probabilistic guarantees with a given probability.
• Refinement is formalized as a probabilistic relation where one contract is refined by another if the latter satisfies the former with a certain probability, preserving compositionality.
• Composition of contracts is defined via parallel composition, where the combined contract's satisfaction probability is derived from the product of individual satisfaction probabilities under compositional assumptions.
• The theory proves that probabilistic refinement and composition are compositional, enabling modular analysis independent of system size or structure.

Experimental results

Research questions

• RQ1 How can Assume/Guarantee contracts be adapted to support probabilistic reasoning about system reliability?
• RQ2 Can probabilistic satisfaction and refinement relations be defined such that compositionality is preserved in the presence of probabilistic environmental assumptions?
• RQ3 In what way does modeling environmental inputs as probabilistic rather than non-deterministic improve the modularity and scalability of reliability analysis?
• RQ4 How can the satisfaction probability of a composed system be derived from the satisfaction probabilities of its components under probabilistic assumptions?
• RQ5 What is the role of probabilistic refinement in enabling incremental verification of system reliability?

Key findings

• The proposed probabilistic Assume/Guarantee theory preserves compositionality, enabling modular reliability analysis regardless of system size or integration order.
• The satisfaction probability of a composed system is the product of the individual satisfaction probabilities of its components, i.e., α·β, when each component satisfies its contract with probability α and β respectively.
• The theory supports both top-down and bottom-up verification, allowing reliability analysis to be performed incrementally as components are developed.
• The probabilistic refinement relation allows reasoning about how a contract can be relaxed (e.g., from C to C') while maintaining a quantified level of satisfaction γ, such that M₁∥M₂ ⊨ₐᵦᵞ C'.
• The framework enables cases where individual components do not satisfy their contracts due to environmental uncertainty, but their composition does, due to probabilistic assumptions on the environment.
• The approach generalizes classical reliability formalisms like pCTL and Dynamic Fault Trees by linking satisfaction directly to behavioral semantics, not just failure probabilities.