Social Robot Navigation Tasks: Combining Machine Learning Techniques and Social Force Model

Area of Science:

• Robotics
• Artificial Intelligence
• Human-Robot Interaction

Background:

• Robot navigation in dynamic public and private spaces is challenging due to environmental constraints and unpredictable pedestrian movement.
• Ensuring human comfort and safety during robot navigation requires a "human-aware" approach.

Purpose of the Study:

• To develop and validate novel navigation strategies for social robots that prioritize human comfort and safety.
• To integrate machine learning techniques with the Social Force Model (SFM) for enhanced social navigation capabilities.

Main Methods:

• Two navigation tasks were developed: social robot navigation and robot accompaniment.
• Both tasks utilized sensor data for environmental awareness and pedestrian motion analysis.
• The Social Force Model (SFM) was employed to model pedestrian behaviors, combined with supervised deep learning and Reinforcement Learning (RL) for robot motion control.

Main Results:

• The integrated SFM and machine learning models enabled robots to navigate in a socially aware manner.
• Simulations and real-world experiments with humanoid (IVO) and aerial robots demonstrated the system's effectiveness.
• The approach successfully addressed human-aware navigation challenges in complex, dynamic environments.

Conclusions:

• The combination of the Social Force Model (SFM) and machine learning provides a robust solution for human-aware social robot navigation.
• This methodology facilitates safe and comfortable robot movement alongside pedestrians and during accompaniment tasks.
• The developed systems show significant promise for real-world deployment of social robots in diverse environments.