Workshop: Surrogate Measures of Safety—when can we start trusting them?
Surrogate Measures of Safety (SMoS) promise to deliver safety diagnoses without relying on real accident data. In this workshop, we will focus on validity of SMoS. Is it ever possible to prove the validity? Can it be proven incrementally, or it is a paramount question that requires yes/no answer once and for all? How much validation is required before we can start trusting and using an SMoS technique?
The SMoS idea can be traced back to the 1950s, and within site-based observation tradition, much theoretical work has been done to define and operationalise the concept of SMoS validity. Though often limited in scale, some practical validations have also been performed.
In-vehicle instruments could generate terabytes of data long before such extensive data collection became feasible for site-based observers. While significant advancements in the practical usage of such data were made, the underlying theoretical grounds might not have received as much attention.
The presented materials come primarily from site-based observations; many of the theoretical constructs are equally applicable for the in-vehicle setup, too. We will also talk a lot about the most recent data analysis techniques, such as Extreme Value Theory, that might finally provide the long-awaited break-through for the surrogate measures of safety.
Workshop programme (preliminary)
- ‘SMoS and validity—problem formulation, current status, future directions’
Aliaksei Laureshyn, Lund University, Sweden
- ‘Extreme Value Theory—theory and application for dummies’
Zhankun Chen, Lund University, Sweden
- ‘Advancing EVT while keeping it operational’
Lai Zheng, Harbin Institute of Technology, China
- ‘The challenge of showing safety of an automated driving system’
Fredrik Sandblom, Zenseact, Sweden
- ‘Making SMoS a tool for practitioners’
Andrew Tarko, Purdue University, IN, USA (online)
The workshop will also include discussion sessions, a coffee-break (courtesy of SAFER), and a joint lunch (courtesy of SUperSAFE ERC project).