Ensuring functional safety in autonomous systems is not merely a technical challenge - it is also an economic one.
Achieving high diagnostic coverage often incurs significant cost, and the pursuit of cost effectiveness introduces a fundamental trade-off: whether to detect faults locally at the subcomponent level or globally at the system level.
Local fault detection is appealing because it allows precise diagnostics without requiring an understanding of the overall system function.
However, this approach struggles to detect interaction faults that arise from unexpected behaviors between subcomponents.
In contrast, system-level monitoring inherently captures such interaction faults by observing deviations in overall system behavior, but it is harder to argue that it provides the high diagnostic coverage typically required by safety standards.
This tension highlights what we call the "1 Dollar Problem": how to achieve sufficient fault detection without overspending on diagnostic infrastructure – that is, deciding where to spend the dollar.
This presentation explores the competing aspects influencing costeffective safety design - such as verification cost and effort at different abstraction levels, reusability of safety mechanisms, and the impact of different safety strategies on system availability - and how these factors shape engineering trade-offs in autonomous systems.