User:Schwalb edward/sandbox

Bayesian Hazard Analysis Risk Assessment (HARA) leverages quantitative Bayesian approach to probabilistic analysis to estimate risk associated with hazards.

When is Bayesian HARA needed?
Safety is central to numerous existing industries. Aviation, as an example, has a long and rich successful history of safety administration. Starting from humble beginning, today aviation is safely transporting hundreds of lives on commercial jets; the expectation is that a catastrophic failure occurs less than once in a billion flight hours. A key ingredient to such safety are the airworthiness safety checklists and traffic management rules, which require maintaining physical separation of miles between aircraft [ref], allowing more than 30 seconds of reaction time to imminent collision hazards. Having the large airspace to enable such large separation and long reaction times allows maintaining safety using purely qualitative safety measures, such as manual checklists and human-in-the-loop traffic controls.

When separation is small and reaction times are in the order of seconds, a quantitative approach such as Bayesian HARA is required. As an example, in contrast to safe flight, safely driving freeways and city streets requires complex perception and sub-second reaction times to avoid collisions with other road users or infrastructure. With the advent of autonomous driving features, designing safe vehicles which avoid accidents with sub-second reaction times requires more than qualitative approaches and human-in-the-loop controls.

Operational Safety vs Design for Safety
Operational safety is concerned with avoiding harm to users of already developed products. As an example, airworthiness checklists executed by pilots ensure that the aircraft is functioning as designed. In contrast, design for safety is focused on building the functionality which ensures safety. As an example, designing an aircraft to be safe is concerned with specification and build of the component and their features that ensure safety, and which are verified by pilots as part of their airworthiness checklists. Whereas the control surface mechanism is design to be fail-safe, the pilots must confirm that the control surface mechanism is functioning as specified.

Bayesian HARA can be used for both design for safety and for operational safety. When used for design, Bayesian HARA is used to discover hazards, estimate their probability distributions and associated risks. When used for operational safety, Bayesian HARA is used for estimating the relative importance of scenarios.

The Multi-Agent Safety Problem
TBD

The Basic Model
TBD

Relationship to existing operational methodologies
TBD

Relationship to existing design methodologies
TBD

Example #1: Backing up scenario
TBD

Example #2: Driving thru city streets
TBD

Example #3: Following vehicle making maneuver
TBD