Autonomous Winter Road Maintenance Decision Making Enabled by Boosting Existing Transportation Data Infrastructure with Deep and Reinforcement Learning
Project Information
Project ID:
FHWA-PROJ-20-0062
Project Abstract:
This research aims to boost the current model-driven winter maintenance with an artificial-intelligence (AI)-enhanced framework in which data can be analyzed in real-time for autonomous decision-making, and such decision-making ability can be improved continuously as more data is obtained. For this purpose, recurrent neural networks will be used to obtain a data-driven environment (i.e., road conditions) prediction ability; deep reinforced learning will be employed to reach an autonomous decision-making ability; convolutional neural networks (CNN) will be investigated for accurate and real-time road condition sensing.
The proposed AI work will form a closed-loop consisting of data acquisition, condition predictions, decision-making, validation, and human intervention. If successful, the project deliverables will significantly improve winter road maintenance, directly benefit safety, operations, mobility and workforce training, and indirectly contribute to pavement design and management.
Project Status:
Active
Project Funding Amount (Contract Award Amount):
$689,239.00
Start Date:
End Date:
Public Access Plan:
Yes
FHWA Program Area:
- Exploratory Advanced Research
Fiscal Year / AMRP Program / AMRP Activity:
- FY 2002-2022 / Exploratory Advanced Research
- FY 2002-2022 / Operations / Automation and Connectivity
- FY 2002-2022 / Operations / Transportation Systems Management and Operations
- FY 2002-2022 / Safety / Safety Design and Operations
AMRP = Annual Modal Research Plan
Contact Information
Team:
Infrastructure Analysis and Construction Team
Office:
Office of Infrastructure Research and Development
Office Code:
HRDI-20