Technical Name 5G C-V2X Enabled Intelligent Traffic Prediction and Warning System
Project Operator National Chiao Tung University
Project Host 李大嵩
We propose a 5G C-V2I enabled intelligent traffic prediction and warning system in a framework including roadside units (RSUs) with radars and edge computing servers (ECSs). The proposed system includes two key techniques:
The “low-complexity extended traffic prediction” realizes prediction of complex dynamics with ultra-low computation time. A deep-learning neural network integrates predictive traffic information (PTIs) into extended predictive traffic information (EPTI). Danger zone detection is conducted based on EPTI, which yields the enhanced EPTI (EEPTI).
The “low-latency augmented-awareness navigation” combines EEPTI and existing navigation information into one with augmented situational awareness. The system sends the EEPTI to 5G application servers to broadcast it to road users, who can use their APPs to decode the information on in-vehicle tablets or smart phones to acquire trajectories of ego-vehicle and surrounding vehicles, the intersection dynamics and warning message.
Scientific Breakthrough
We propose a trajectory prediction system that adopts mmWave radars with more dependable detection in harsh environments than GPS and image-based systems. The proposed system integrates PTIs generated by RSUs at multiple intersections to offer a more comprehensive field-of-view of the road environment. We reduce the latency by adopting a low-complexity light-weight prediction model. Through 5G C-V2I, EEPTI with warning message is transmitted to road users via URLLC network protocols at a target latency of 50 ms, which meets the requirement of traffic warning services.
The proposed system is the first attempt of an intelligent extended traffic prediction and warning system based on 5G C-V2I architecture and mmWave radars, and can assure prediction reliability in harsh environments.
Industrial Applicability
The proposed intelligent extended traffic prediction and warning system can provide more time margin for road users to conduct essential decision making and judgements. Road users can have access to this safety service through an easy-to-use APP operated on in-vehicle tablets or smart phones. It has high potential to become a business case in 5G intelligent transportation vertical applications. The proposed system meets the specification of 5G C-V2X, and can be integrated with existing traffic monitoring platforms to enter the intelligent transportation market. Road users at large can benefit from the system with augmented-awareness navigation and enjoy elevated level of active safety. The proposed solution can hopefully open up a new 5G service market with a high promising prospect.
Keyword millimeter wave radar trajectory prediction deep-learning ultra-reliable and low-latency communications edge computing smart sensing recurrent convolutional neural net roadside unit self-driving vehicle intelligent transportation system
  • Contact
other people also saw