Machine Learning Analysis of Autonomous Vehicle Sensors Under Extreme Conditions in Alaska

Abstract

Abstract Autonomous vehicles are part of an expanding industry that encompasses various interdisciplinary fields such as dynamic controls, thermal engineering, sensors, data processing, and artificial intelligence. Exposure to extreme environmental conditions, such as changes to temperature and humidity, affects sensor performance. To address potential safety concerns related to sensor perception used in autonomous vehicles in extremely cold real-world situations, specifically Alaska, examination of frosts and water droplets impact on vehicle optical sensors is conducted in both real-world and laboratory-controlled settings. Machine learning models are utilized to determine the vision impediment levels. Potential hardware and software tools are then introduced as solutions for the environmental impacts. Through this research, a better understanding of the potential caveats and algorithm solutions can be suggested to improve autonomous driving, even under challenging weather conditions.