ScenicProver: A Framework for Compositional Probabilistic Verification of Learning-Enabled Systems

2025-11-04

AI Summary

Comprehensive probabilistic verification of learning-driven cyber-physical systems (CPS) remains challenging due to black-box components and complex, realistic operational environments. Method: This paper proposes a compositional probabilistic verification framework built on the Scenic language for component-based modeling. It integrates assume-guarantee contracts with a composable evidence fusion operator, and incorporates Lean 4 formal verification, an extended Linear Temporal Logic (LTL) semantics, test-driven evidence generation, and external assumption importation. Contribution/Results: The framework enables hierarchical, traceable, and trustworthy assurance from components to full-system level. Evaluated on an autonomous emergency braking system, targeted testing under uncertainty—under identical computational budgets—significantly strengthens probabilistic guarantees, demonstrating both efficacy and practicality.

Technology Category

• Reasoning under Uncertainty: Probabilistic Programming
• Machine Learning: Calibration & Uncertainty Quantification
• Multiagent Systems: Multiagent Systems under Uncertainty

Application Category

• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization

Abstract

Full verification of learning-enabled cyber-physical systems (CPS) has long been intractable due to challenges including black-box components and complex real-world environments. Existing tools either provide formal guarantees for limited types of systems or test the system as a monolith, but no general framework exists for compositional analysis of learning-enabled CPS using varied verification techniques over complex real-world environments. This paper introduces ScenicProver, a verification framework that aims to fill this gap. Built upon the Scenic probabilistic programming language, the framework supports: (1) compositional system description with clear component interfaces, ranging from interpretable code to black boxes; (2) assume-guarantee contracts over those components using an extension of Linear Temporal Logic containing arbitrary Scenic expressions; (3) evidence generation through testing, formal proofs via Lean 4 integration, and importing external assumptions; (4) systematic combination of generated evidence using contract operators; and (5) automatic generation of assurance cases tracking the provenance of system-level guarantees. We demonstrate the framework's effectiveness through a case study on an autonomous vehicle's automatic emergency braking system with sensor fusion. By leveraging manufacturer guarantees for radar and laser sensors and focusing testing efforts on uncertain conditions, our approach enables stronger probabilistic guarantees than monolithic testing with the same computational budget.

Problem

• Addresses compositional verification challenges in learning-enabled cyber-physical systems
• Overcomes limitations of monolithic testing through probabilistic assume-guarantee contracts
• Enables stronger system-level guarantees by combining multiple verification techniques

Innovation

• Compositional system description with clear component interfaces
• Assume-guarantee contracts using extended Linear Temporal Logic
• Systematic evidence combination through contract operators