Enhancing Pedestrian Safety in Autonomous Vehicles Using Neural Networks for Informal Communication

Abstract: Pedestrian safety remains a global concern, particularly in South Africa. Despite various initiatives, the slow decline in traffic fatalities contradicts the goals outlined in the Sustainable Development Goals. A key gap in the literature is the limited attention to human factors, especially the dynamics of formal and informal communication among road users. This oversight suggests that pedestrians are often not prioritized in road safety discussions and initiatives. However, with the rise of autonomous vehicles, understanding their interaction with pedestrians and human drivers is essential for enhancing road safety. As the interaction between autonomous vehicles and other road users is crucial, as it can significantly influence pedestrian safety outcomes. This paper explores how autonomous vehicles can employ informal communication, such as car lights patterns and movements, to communicate effectively with pedestrians and drivers. By examining how autonomous vehicles can utilize non-verbal cues and informal signals to interact with pedestrians and drivers. This study utilizes Convolutional Neural Networks (CNNs) to extract spatial features and Recurrent Neural Networks (RNNs) to model temporal sequences, identifying patterns in autonomous vehicle communication. By detecting recurring informal signals, this approach improves vehicle-to-vehicle (V2V) communication, bridging the gap between technology and human factors. It ultimately aims to enhance road safety and traffic flow in autonomous systems.