1. Home
  2. RIP
  3. View Record
https://www.nationalacademies.org/trb/events

Leveraging AI to Improve the Safety of Connected and Automated Vehicles in Winter Weather

Connected and automated vehicles (CAVs) are socially important because they have the potential to reduce accidents, expand access to transportation, and improve efficiency. However, broader adoption of CAVs in the Minnesota region is hampered by many challenges. First, winter weather snowfall corrupts sensor data as LiDAR returns become noisier, camera imagery loses contrast, and both modalities suffer from occlusions and missing data. These corruptions lower the reliability of perception modules and introduce uncertainty into all subsequent decision-making stages. Second, cold weather affects battery chemistry, reducing range and increasing charging time, which increases uncertainty for drivers and automated vehicles alike. For instance, during a cold snap in January 2024 in Chicago, charging stations saw long lines of electric CAVs [15], many with dead batteries, exacerbating the range anxiety hampering adoption in cold-weather states. This project investigates the potential of recent artificial intelligence (AI) breakthroughs, such as diffusion models and generative AI, to enhance winter safety and increase the adoption rate of CAVs. To improve safety, the project evaluates the ability of diffusion models, a recent AI advance, to reduce noise in winter-degraded LiDAR, camera, and other sensor data during snowfall. To spur adoption and support the sustainability of CAVs, the project examines the value of AI in improving models for estimating CAV energy use and emissions, extending the range of CAVs, and facilitating eco-routing to help choose routes that reduce energy use and emissions.