Capturing complex behaviour for predicting distant future trajectories

• DOI: 10.1145/3004725.3004730
• Publication Date: 2017-02-10T13:39:43Z
• Authors: Bertil Chapuis; Arielle Moro; Vaibhav Kulkarni; Benoît Garbinato
• Keywords: 11. Sustainability; 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

We put forth a system, to predict distant-future positions of multiple moving entities and index the forecasted trajectories, in order answer predictive queries involving long time horizons. Today, proliferation mobile devices with GPS functionality internet connectivity has led rapid development location-based services, accounting for user mobility prediction as key paradigm. Mobility is already playing major role traffic management, urban planning advertising, which demand accurate horizon forecasting movements. Existing methodologies either use motion patterns or techniques based on frequently visited places predicting next move. However, when it comes distant-future, human too complex be represented by such statistical functions. Therefore, existing are not well suited satisfactory level accuracy. To tackle this problem, we introduce novel spatial object, 'Representative Trajectory', embodies movements users amongst their zones interest. propose means empirically evaluate quality object dynamically adapt its extraction method behaviour. rely an inverted store predicted trajectories that scales number entities. Our evaluation results show technique achieves more than 70% predictions best technique. This shows longer query horizons do necessarily indexing schemes, have rebalanced they grow constantly experienced problem while answering queries.