Incorporate Sound Localization Through Audio Detection in Fully Automated Vehicles to Enhance Detection Accuracy and Improve Ground Maintenance Vehicle Safety on Airfields.
Experimental Study Evaluated Fully Automated Vehicles, Including Tow Tractor in Airfield, Shuttle in Crowded Indoor and Last-Mile Delivery Vehicle, On-Demand Taxi, Health Service Vehicle, Infrastructure Monitoring Vehicle and Shuttle in Urban Areas.
Cincinnati, Ohio, United States
Kentucky, United States
Nationwide, South Korea
Results and Lessons Learned from Autonomous Driving Transportation Services in Airfield, Crowded Indoor, and Urban Environments
Summary Information
Research over recent decades has demonstrated the potential of automated vehicles (AVs) in urban settings, though they sometimes underperform compared to expert drivers in complex scenarios. This study presented the research results and lessons learned from fully automated driving transportation services in airfield, crowded indoor, and urban environments . Specifically, the study ran a fully automated tow tractor service at the Cincinnati/Northern Kentucky International Airport (CVG) airfield for about 107 hours, as well as operating indoor fully automated shuttle services for about 20 hours at Incheon International Airport arrival and departure halls in South Korea. This study also incorporated the findings from operating last-mile delivery and shuttle services in urban environments in Seoul, Sejong, Gwangju, Hwaseong, and Jinhae in South Korea.
- Incorporate sound localization through audio detection in fully automated vehicles to enhance detection accuracy. This is especially critical for tow-tractors operating in airfields, where there is air blasting near airplane engines (jetblast), to properly detect jetblast and avoid getting any significant damage.
- Develop a robust and adaptive control algorithm for variable tow-tractor dynamics. This is important since tow-tractor payloads operating in airfields can change excessively and this could significantly affect the efficiency and safety of fully automated tow-tractors.
- Provide human intervention for training fully automated vehicles, supported with reinforcement algorithms. To truly train fully automated vehicles in the real-world, substantial human intervention and supervision is required to intervene before dangerous situations ensue, to reset and let the vehicle learn through autonomous reinforcement learning algorithms, which could eventually minimize human intervention in various tasks.
