All-Weather Autonomous Vehicle: Performance Analysis of an Automated Heavy Quadricycle in Non-snow and Snowstorm Conditions Using Single Map

Abstract

Bad weather (e.g. snowstorm, tropical rain) still brings issues to the automated vehicle (AV) performance due to several factors, such as limited visibility of the environment and bad positioning. As such, most of the testing and study of AV has been done in optimal condition weather. However, for a fully automated vehicle experience, a reliable AV is required to be able to maneuver in all type of weather and environment condition. In this work, an experimental series has been held in the Arctic Circle to evaluate the positioning performance of a driverless vehicle system in the rural area and harsh weather condition. The experiment’s aim is to enable an all-weather AV performance using a single map which can endure harsh northern environment weather. This work briefly reports the experiment finding as well as concisely analyzing the effects of the positioning strategy in the varied weather conditions towards the automated vehicle dynamics and controller performance. A sudden snowstorm during the test sessions allows the system reliability to be validated in both non-snow and snowstorm conditions. The computed average lateral positioning error value of the automated driving is 18.7 cm in non-snow weather, and 22.0 cm for the snow condition. The results show that the inclusion of our solution into the AV system helps to provide reliable localization in all-weather condition, thus providing dependable tracking performance by the motion control. This work is important as it is the first public test of the automated road vehicle in the Arctic Circle ever done .