Pedestrian Traffic Flow Prediction based on ANN Model and OSM Data
One of the main challenges that blind pedestrians have to cope with in their daily life is orientation and navigation while walking the urban space. In our previous research, a computerized network route calculation algorithm for blind pedestrians was developed, which relies on OpenStreetMap (OSM) mapping data, aimed at calculating optimal routes in terms of accessibility and safety. Despite the potential and practicability of our solution, critical mapping data is still missing in OSM to allow a comprehensive and scalable solution. One data type is related to pedestrian traffic flow that has been found to influence the path blind pedestrians will choose to walk. Artificial Neural Networks (ANN) model, which allows learning and predicting different phenomena from training samples by investigating the correlation and effects among various environmental features, is developed and used to model and predict pedestrian traffic flow, while relying on the existing OSM data. To model the ANN, we have relied on parameters and factors related to the streets’ geometrical and topological configuration, as well as points of interest nearby, e.g., public transportation and shops, to name a few. The ANN model was trained using training samples of pedestrians traffic flow collected in the center of Tel-Aviv. Implementing the computed ANN model, nearly 90% of the testing data was successfully predicted. We believe that this ANN model can accurately generate new data, which together with the existing OSM data, can greatly contribute and augment the reliability of route calculation algorithms for blind pedestrians.