Real-time Intent Estimation using IOC for Motion Prediction of Pedestrians on the Urban Roads

This thesis presents a new method to help self-driving cars predict the movements of vulnerable road users in real time. Using a mathematical approach called inverse optimal control, the system estimates a person’s intent based on their observed actions, even when only part of their path is visible. Unlike data-heavy AI models, this method is transparent, accurate, and works with limited data. A real-time motion prediction module was developed and tested on both synthetic and real-world pedestrian data, achieving high accuracy. The results show strong performance across different scenarios, improving safety and decision-making for autonomous vehicles in complex environments.