Learning From Accidents

Resembler is a recursive declarative traffic incident platform designed to re-map traffic accidents and incidents to similar traffic situations. The platform provides a statistical distribution of potential accident types.

 

The platform is highly structured so that the output can interface to the transversal domains working on traffic safety from ADS software, Active Safety, ITS city safety to systematic verification methodologies based on real accidents.

 

It utilises the increased availability of accident footage, the availability of higher resolution maps (HD Maps and Satellite Images of road topologies) and the deep learning computational power and algorithms.

 

Resembler allows time deduction (step back) of the incident situation, thus as more information related to the incident (time, surrounding traffic, topological situation, environment, etc.) becomes available, the statistical distribution of the accident type changes. Because the deduction is based on “actual” real accidents and incidents, the platform removes ambiguity in the creation of accident types.