The Effect of Latency on the Performance of Remote Driving with Video Streaming

Remote driving by humans offers more flexibility than self-driving vehicles in terms of road hazard perception. However, the latency between a vehicle and a driver depends on the transport protocol of data and images. This study aims to investigate the impact of video streaming latency on driving performance during constant-radius cornering maneuvers. For safety reasons, the experiments were conducted on a one-tenth-scale RC car installed by the Cloud RC's AdvenX3 transmitter and receiver to enable participants to drive the car remotely from the cockpit through the steering wheel and pedals. The low-latency video streaming was achieved through a cable from the on-board web camera while the specified time delays ranged from 0 to 300 milliseconds were added. The drivers needed to maintain the car speed and track the center of the road while cornering through the quarter-circular path of radius two meters. The camera was installed to capture the top-view footage of the maneuvers above the grid of three-by-three reference markers on the smooth floor. For each trial, the coordinates of the car varying against time are plotted to display the cornering paths for analyzing the driving performance across different delay settings. The higher latencies led to increased deviations from the intended trajectory. The effect of latency on the driving performance depends on the individual drivers. Our findings highlight the need for optimizing latency thresholds to guarantee safety for remote driving.