Real-Time People Re-Identification and Tracking for Autonomous Platforms Using a Trajectory Prediction-Based Approach

Area of Science:

• Robotics and Artificial Intelligence
• Computer Vision
• Human-Robot Interaction

Background:

• Autonomous systems like social robots and self-driving cars require robust human re-identification for safe and effective operation.
• Real-time tracking and trajectory prediction are critical for applications demanding quick reactions, such as autonomous driving.

Purpose of the Study:

• To develop and evaluate a novel real-time people re-identification system.
• To enhance human-robot interaction by accurately tracking individuals in dynamic environments.
• To improve safety in autonomous driving through precise people trajectory prediction.

Main Methods:

• A real-time people re-identification system integrating semantic environmental information and social influence for trajectory prediction.
• Evaluation within social robotics (using the AMIRO framework) and autonomous driving scenarios.
• Quantitative and qualitative analysis using existing datasets and real-time acquired data.

Main Results:

• The proposed system demonstrates improved performance in people re-identification and trajectory prediction.
• An enhancement of over 5% in the Multiple Object Tracking Accuracy (MOTA) metric was achieved compared to existing modules.
• The system proved effective in both social robotics and autonomous driving case studies.

Conclusions:

• The combined approach of semantic and social information significantly improves people re-identification and trajectory prediction.
• The developed system offers a viable solution for enhancing safety and interaction in autonomous systems.
• The method shows promise for real-world applications in social robotics and autonomous vehicles.