Data Rate Reduction for Video Streams in Teleoperated Driving

With the pioneering introduction of autonomous vehicles, system failures while driving from A to B are more likely to occur. In such scenarios one option is to hand back the control to the human driver, if someone suitable is inside the vehicle. Teleoperated Driving, the remote control of vehicles by human operators, can be a solution to scenarios without suitable drivers inside. A video stream is used to provide operators with an overview of the vehicle's environment and support for a safe remote control. By utilizing cellular networks as wireless communication medium for Teleoperated Driving, the available bandwidth is a limiting factor. This paper introduces a multi-step approach to lower the bandwidth requirements, which is achieved by initially splitting the single video stream into two parts: One part conveying the original video information restricted to important objects and the remainder, to which various filters are applied. Results show that this approach can lead to a decreased bandwidth consumption. These results are validated with a user study, where participants had to rate the perceived video quality and the driveability for the different combinations. This user study shows that, for every investigated scenario, at least one combination of parameters (applied filters) was rated driveable. Finally, the results are used to sketch a system that infers specific combinations of parameters based on the environmental conditions and the available bitrate.