Fuzzy integral-based gaze control architecture incorporated with modified-univector field-based navigation for...

Area of Science:

• Robotics
• Artificial Intelligence
• Computer Vision

Background:

• Humanoid robot navigation in dynamic environments is challenged by limited sensor data.
• Effective dynamic obstacle avoidance requires frequent local map updates via gaze control.
• Conventional methods struggle with real-time adaptation to environmental changes.

Purpose of the Study:

• To propose a fuzzy integral-based gaze control architecture for humanoid robots.
• To enhance dynamic obstacle avoidance capabilities in complex environments.
• To improve robot navigation performance by integrating multiple environmental cues.

Main Methods:

• A fuzzy integral-based gaze control architecture was developed.
• Four criteria (map confidence, waypoint, self-localization, obstacles) were defined with partial evaluation functions.
• Fuzzy integral was used for global evaluation of candidate gaze directions.
• Modified-univector field navigation incorporated self-localization error and obstacle data for virtual dynamic obstacles.

Main Results:

• The proposed architecture demonstrated improved performance compared to the weighted sum-based approach.
• Simulations using the HanSaRam-IX (HSR-IX) robot validated the system's effectiveness.
• The fuzzy integral approach provided a more robust global evaluation for gaze control.

Conclusions:

• The fuzzy integral-based gaze control architecture significantly enhances humanoid robot dynamic obstacle avoidance.
• Integrating multiple criteria through fuzzy integral leads to more competent navigation.
• The proposed method offers a promising solution for real-world humanoid robot applications.