Functional-safety design and verification strategies for GPU-based systems in the AI era

Overview

Connects GPU timing, memory, synchronization and resource-interference failures to vehicle hazards. Explains independent CPU monitoring, isolation, safe-state transitions, fault injection and reassessment after model or driver changes as safety-case evidence.