Assuring Learning-Enabled Cyber-Physical Systems with Foundation Models: A Neuro-Symbolic Agenda for Testing, Verification, and Runtime Assurance
Authors: Xi Zheng, Ivan Ruchkin, Hong Jin Kang, Tevfik Bultan, Vijay Ganesh, Taylor T. Johnson, Ziyang Li, Lei Ma, Aloysius K. Mok, Daniel Neider, Ruzica Piskac, Nalini Venkatasubramanian, Simon Thompson, Sanjoy Baruah
Organizations: Macquarie University, Australia · University of Florida, USA · University of Sydney, Australia · University of California, Santa Barbara, USA · Georgia Institute of Technology, USA · Vanderbilt University, USA · Johns Hopkins University, USA · The University of Tokyo, Japan · The University of Texas at Austin, USA · TU Dortmund University, Germany · Yale University, USA · University of California, Irvine, USA · TIER IV, Japan · Washington University in St. Louis, USA
Learning-enabled cyber-physical systems (CPS), including autonomous vehicles and UAVs, increasingly incorporate foundation models, yet fragmented assurance practices provide insufficient support for system-level guarantees. We argue that restoring engineering control across stochastic neural, symbolic, and physical boundaries requires a shift from model-centric evaluation to system-level assurance. We introduce the Foundation Model Assurance Stack, connecting three directions: (i) neuro-symbolic architectures that constrain foundation models through explicit symbolic world models, reasoning layers, and interfaces; (ii) semantically grounded, conformance-based testing that combines formal specifications and symbolic world models with foundation-model-assisted specification mining, scenario generation, and test synthesis; and (iii) compositional verification and runtime assurance that target formally tractable components and interfaces through modular contracts, equivalence checking, verified synthesis, uncertainty-aware reasoning, and runtime monitoring and enforcement. Symbolic abstractions establish assurance boundaries between stochastic learning and formal reasoning, connecting testing, verification, and runtime evidence across the system lifecycle. Foundation models serve as constrained assistants whose outputs are independently checked wherever correctness can be formally specified. We identify research challenges in specification engineering, abstraction fidelity, evidence composition, and lifecycle certification, and outline a staged agenda toward trustworthy, deployable, and certifiable learning-enabled CPS.
Figures & tables
Figure 1. Foundation Model Assurance Stack for AI-Enabled CPS. The stack organizes assurance into three complementary pillars: foundation models as assistive components in CPS architectures, testing and validation of AI-enabled systems, and verification and runtime assurance. These pillars are connected by cross-cutting neuro-symbolic integration, including symbolic world models, DSLs and structured representations, contract specifications, and neural–symbolic interfaces. Architecture provides world models, specifications, and interfaces to testing, and contracts, abstractions, and constraints to verification. Testing provides empirical evidence through conformance-based testing, structured scenario generation, DSL-guided test synthesis, LLM-assisted specification mining, and adequacy assessment. Verification and runtime assurance provide formal guarantees through contract verification, interface verification, transformation verification, runtime monitoring, and safety enforcement. At runtime, a Simplex-style resilience layer combines a learning-enabled controller, runtime monitor, safe controller, and shielding or switching mechanism to support operational safety under uncertainty.
Cyber-Physical Systems (CPS) study networked computation and control interacting with physical processes, from factories and power grids to robots and vehicles.