A collection of inclement weather data enables the evaluation and training of generative and discriminative models for environment perception, but also the development of elaborate simulation methods. The project will enhance inclement weather data collection in controlled laboratory and outdoors conditions for testing, modelling and validation. The data will be collected from adverse weather facilities, outdoor measurement areas, and highway test sites. It will provide relevant input data, for sensor evaluation, simulation and validation of weather models and computer vision algorithms.
Driving testing campaigns for sensor and perception validation in real traffic scenarios and in the partners’ fog and rain chambers will be carried out. These tests will ensure that the prototypes will be functional in real driving conditions and validating the simulation model of the system under test. Such models allow for the development of sensors and perception through simulation as well as the generation of artificial environmental dataset. Starting from an accurately derived definition of the regions of interest, the test scenarios will be defined and the weather database will enhance the training database with artificially generated training data by simulation. The validation will proceed both by simulation and in real environment.
The partners have unique testing and development capabilities. As the main use cases base on an inclement weather performance, the main tests will be conducted at winter test center(s) in Sweden/Finland.
The following tests are foreseen:
- Create a sensor framework at the test site in the North where the winter-time is relatively long starting in November and ending in March. The aim is to use a similar setup in the car to have various measures in urban areas, highways, etc. when drifting snow makes the sensors blind.
- Carry out a series of test drives in Northern Europe. The durability in artic conditions is measured with installing static measurement equipment in the North, and it will be remotely operated.