Deep Learning-Based Pedestrians Controlling System on Zebra Crossing

The 'Vision zero' goal represents the apex of the European Commission's road safety plan, andenhancing road safety is seen as a primary objective by governments and transport policymakers across theglobe. Due to the greater death rate associated with crashes involving pedestrians, this vulnerable usergroup has been the focus of increased study. Consequently, there is a growing need for studies thatexamine pedestrian behavior and the implementation of Intelligent Transport Systems designed to aidpedestrians. Using a Countdown Signal Timer (CST) and some machine learning techniques, this researchaims to forecast how pedestrians would behave at crossings. Presenting this case study is an analysis of asimulated environment model intersection with pedestrian traffic lights equipped with countdown signaltimings. An X-Convolutional Neural Network (X-CNN) and a decision-making model were bothemployed to meet the requirements of the analysis. Both models demonstrated satisfactory performance,according to the results. To be more specific, the X-CNN model used a Mean Squared Error value topredict the pedestrians' crossing speed. By accurately predicting how pedestrians cross the street, we maylearn more about the impact of countdown signal timers and how to improve infrastructure safety for thisdemographic so that they receive timely and relevant information through the speaker.