A camera-based vehicular visible light communication system using modulated taillights in public road scenarios

This paper proposes a camera-based vehicular visible light communication system using modulated LED taillights for vehicle-to-vehicle communication. The taillights are modulated using on-off keying with a carrier frequency above the critical flicker frequency threshold to appear steady to the human eye and only a camera with a short exposure time can perceive the signal. For receiving the signal, a common camera with a recording frame rate of 30 frames per second was used for the experiments. This limits the throughput of the communication system as only one bit per frame and per individually modulated light source can be transmitted, thus the system is not suitable for transmitting time-critical data. However, the main advantage is that the transmitted data as well as the transmitting vehicle are visible in the camera image, hence we can associate the message and its origin. For the communication between two cooperative autonomous vehicles, e.g., members of a platoon, a secure wireless channel with low latency is needed. To establish an encrypted connection in this main channel our system can be used as an out-of-band channel to verify the identity of the vehicle in front by transmitting a verification code. In this paper we evaluate how the communication quality in terms of bit error rate of the raw data transmission without any error correction methods is influenced by various environmental factors, e.g., communication distance, traveling speed, weather, brightness, etc. This is done by testing our system in controlled scenarios as well as on public roads. The results show that our system is capable of transmitting data with a raw bit error rate of 1.59% on average in platooning relevant situations on the highway in dry conditions. Rain affects the system due to raindrops on the windshield, spray thrown up by vehicles in front, and the windshield wipers occluding the modulated taillights of the transmitting car. In light to medium rain conditions the mean bit error rate is between 2.7% and 6.2%, depending on the intensity.