Computer vision based detector integrated with V2X framework for vulnerable road user safety

The Federal Highway Administration (FHWA) Intelligent Transportation Systems Joint Program Office (ITSJPO) initiative connected and automated vehicle education (CAVe-In-A-Box) supports the use of cellular-vehicle-to-everything (C-V2X) to leverage the advanced 5G wireless network connectivity and establish robust communication between vehicles and the infrastructure. Although in early stage of implementation, the C-V2X protocol has been gaining popularity over dedicated short-range communication (DSRC) protocol due to the different channelization schemes and timing requirements and a wireless architecture like the cell phones, and, contrary to DSRC, uses multiple carriers in blocks of smaller subframes. The C-V2X framework used with the V2X-Hub and a C-V2X road-side unit (RSU) in conjunction with the CAVe-In-A-Box supports seamlessly adapting existing computer vision and machine learning models used in transportation data analytics. In this research, CAVe-In-A-Box infrastructure and mobile kits are assembled and integrated with Bosch Autodome Inteox 7000i camera that offers intelligent video analytics (IVA) to detect, localize, and classify vulnerable road users (VRUs) such as persons and bike-pedestrians in addition to vehicles. Furthermore, to send pedestrian detection calls to signal controller via synchronous data link control (SDLC) and transmit V2X personal safety messages (PSM), MH Corbin Connect:ITS is utilized.

Additionally, Cohda Wireless MK6c C-V2X RSU is integrated with the CAVe-In-A-Box infrastructure kit and the Bosch Autodome camera to broadcast SAE J3224 sensor data sharing message (SDSM) to nearby vehicle on board units (OBUs), when VRU detection occurred. SAE J3224 standard (published in August 2022) for V2X sensor-sharing for cooperative and automated driving supports innovative use cases such as detection of an unequipped VRU by a RSU or host vehicle. In this paper, V2X messages with SAE J3224 (SDSM) and SAE J2735 (PSM) standards have been tested and verified with lab tests. In addition, the CAVe-In-A-Box integrated with computer vision VRU detector final assembly is tested and verified to detect VRUs (pedestrian and bike-pedestrians) and send a bus interface unit (BIU) call to the signal controller which is programmed with a leading pedestrian interval (LPI).

First, the PSM and SDSM V2X messages provide real-time safety benefits particularly in vehicle-VRU conflicts at intersections. Second, the LPI has proven to significantly improve VRU safety. In addition, future mobility in terms of connected vehicles will naturally benefit from this research project outcomes. Finally, combining the V2X messaging, and infrastructure level signal control timing changes will advance VRU safety solutions.

Note: These preliminary results are shared for publication during an active research project sponsored by ODOT Office of Research.