GUIDE-CoT introduces a novel approach that enables large language models to predict pedestrian movements with higher accuracy and contextual awareness.

• Combines goal-driven reasoning with real-time visual information processing
• Implements a dynamic estimation module that continuously refines predictions
• Achieves superior trajectory prediction compared to traditional computer vision approaches
• Demonstrates effective integration of language models with spatial reasoning tasks

This research significantly advances security applications by improving systems that monitor crowd movement, detect anomalous behaviors, and enhance public safety in surveillance environments.

GUIDE-CoT: Goal-driven and User-Informed Dynamic Estimation for Pedestrian Trajectory using Chain-of-Thought