IT-Av Automotive experiment example: video-based assisted driving

A quick summary of IT-Av Automotive experiment example: video-based assisted driving


The objective of the assisted overtaking is to gather all needed information in real-time for a driver to make a decision of overtaking or not the front vehicle, while the visibility of the road from the vehicle is not sufficient to take a decision. Important information can be used to take the decision, such as: video images from the road ahead.

System architecture

The concrete scenario for video-camera-based assisted driving is presented in Figure 1. Each vehicle contains an On Board Unit (OBU) that provides the communication between the vehicle and the infrastructure and a RaspberryPi Cam recording the front vehicle. The OBU also connects to an Android device, smartphone, through WiFi, providing visual information for the driver.

Figure 1 - IT-Av video based assisted driving: system architecture.

Both rear and front vehicles may belong to different brands and have no agreement between them in terms of joint services. This means that the transmission of the video streaming from the camera in the front vehicle to the visual screen in the rear vehicle may need a 3rd party agreement to agree on the service to be established between the vehicles. Moreover, the video quality to be chosen in the camera will depend on the capabilities of the video screen. Therefore, a video transcoding system may be needed to adapt the video capabilities between both vehicles.

Recreating the experiment

Using the 5GinFIRE deploying system, the user can create or instantiate a VxF and submit its experiment using a Network Service Descriptor (NSD) that describes the network service to be created. In this case, the experimenter should request the deployment of opencv_transcoder_ns from 5GinFIRE portal. After successful deployment run [VxF IP] in the front vehicle and [VxF IP] in the rear vehicle.