Architecting Safety-Critical Systems for AI-Enabled Software-Defined Vehicles: A Compute, Timing, and Isolation Framework with a Brake System Case Study
Sep 2026· SAE technical paper series· 0 citations· 10 references
TL;DR
A safety-oriented architectural framework that constrains how AI-enabled functions may be integrated into fail-operational by-wire systems and clarifies trade-offs among safety isolation, latency, and AI performance while preserving braking safety guarantees under continuous software evolution is proposed.
Abstract
Software-defined vehicle (SDV) platforms are reshaping safety-critical system design by consolidating braking and other motion-control functions on centralized heterogeneous edge compute that also executes physical-AI workloads. This consolidation breaks traditional assumptions of fixed ECUs and simple timing envelopes, complicating assurance of determinism, isolation, and fail-operational behaviour for ASIL-D brake functions. Building on a decentralized brake-by- wire (BbW) architecture with dual controllers, redundant low-voltage power grids, and smart electromechanical brake corner actuators, this paper proposes a systems-level framework for architecting safety-critical functions in AI-enabled SDVs along three dimensions: compute, timing, and isolation. The framework classifies conventional and AI-based functions and maps them to heterogeneous compute classes; defines architectural patterns that combine safety islands, power-domain redundancy, and hardware partitioning to support freedom from interference; and formalizes timing domains and contracts that bound latency, jitter, and failover dynamics across sensors, centralized controllers, and decentralized actuators. The contribution is not a new AI algorithm, but a safety-oriented architectural framework that constrains how AI-enabled functions may be integrated into fail-operational by-wire systems. A BbW case study with edge-resident AI observers and anomaly detectors shows how the framework complements System Analysis Tool (SAT)– based failure modelling and clarifies trade-offs among safety isolation, latency, and AI performance while preserving braking safety guarantees under continuous software evolution.
FAA+RTS is a novel solution as it integrates previously-isolated methods for dealing with multiple constraints into a single framework, presenting a single overview of all possible trade-offs to the application designer, resulting in a robust and reliable system.
L. Miedema, D. Sapra, P. Novobilsky et al.· 0 citations
Discussions of the software-defined vehicle (SDV) paradigm have developed almost entirely around automotive assumptions, where middleware fragmentation and connectivity availability get treated as the primary obstacles to adoption. For off-highway equipment construction, mining, and agricultural machinery, a different...
Shyam Laheri Chunduri· International journal of com...· 0 citations
The results prove the platform provides a highly flexible, extensible, and economically accessible open-source alternative to proprietary motorsport instrumentation.
G. Kourtis, Paris Kitsos· Electronics· 0 citations
The time-triggered paradigm has proven highly suitable for closed safety-critical distributed real-time systems with predetermined configurations and stringent timeliness and dependability requirements. The Time-Triggered Architecture (TTA) embodies this paradigm by combining a globally synchronized time base, offline-...
Mohammad Ibrahim Alkoudsi, Marcus Völp, Gerhard Fohler· IEEE Access· 0 citations
A four-layer framework that supports traceable, bidirectional interaction between a system’s architecture and its physical implementation and contributes to MBSE-driven digital twin research by providing a structured process for integrating architecture, translation, execution, monitoring, and synchronization in a trac...
M. Islam, Aditya Akundi, Sergio A. Luna Fong· Systems· 0 citations
Modern intelligent cockpits consolidate In-Vehicle Infotainment (IVI), navigation, and control-related services on shared heterogeneous hardware. This consolidation creates a resource-management problem because cockpit workloads have different controllability, safety, and performance requirements. Existing open-source...
Yi-Cheng Gu, Bing Deng, Yun Wang et al.· ACM Transactions on Architec...· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduNov 20, 2026
Senior Research Scientist Jim Glass discusses how his MIT-IBM Computing Research Lab collaboration connects perception and language, and addresses the shortcoming of current models.