IA-LSTM: Interaction-Aware LSTM for Pedestrian Trajectory Prediction.

Area of Science:

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

Background:

• Pedestrian trajectory prediction is crucial for autonomous systems like self-driving cars and mobile robots to prevent collisions.
• Modeling complex human-human interactions and individual movement patterns in crowded scenes remains a significant challenge.

Purpose of the Study:

• To develop a novel mechanism for measuring the relative importance of human-human interactions in crowd scenarios.
• To enhance pedestrian trajectory prediction by incorporating an interaction-aware architecture.

Main Methods:

• Introduction of a correntropy-based mechanism to quantify interaction importance and define personal space.
• Development of a data-driven interaction module to extract dynamic interaction features and calculate interaction weights.
• Design of an interaction-aware architecture using a long short-term memory network for trajectory prediction.

Main Results:

• The proposed mechanism effectively measures the relative importance of human-human interactions.
• The interaction-aware architecture successfully extracts and utilizes social messages among pedestrians.
• Experimental results on public datasets show superior performance compared to state-of-the-art methods.

Conclusions:

• The novel correntropy-based approach significantly improves pedestrian trajectory prediction accuracy in complex crowd scenarios.
• The developed interaction module and architecture offer a robust solution for understanding and predicting pedestrian behavior.
• This research contributes to safer and more efficient autonomous navigation in human-populated environments.