Multi-objective optimization technique for trajectory planning of multi-humanoid robots in cluttered terrain.

Area of Science:

• Robotics
• Artificial Intelligence
• Control Systems

Background:

• Humanoid robots offer advantages over wheeled robots due to their human-like mobility.
• Navigating cluttered environments presents a significant challenge for autonomous robots.
• Existing trajectory planning methods may lack robustness or efficiency in complex terrains.

Purpose of the Study:

• To propose a novel hybridized controller for trajectory planning in cluttered environments for humanoid robots.
• To enhance the navigation capabilities and robustness of humanoid robots.
• To address deadlock conditions in multi-humanoid systems.

Main Methods:

• A two-step control mechanism combining modified Multiple Adaptive Neuro-Fuzzy Inference System (MANFIS) and Multi-Objective Sunflower Optimization (MOSFO) techniques.
• Utilizing obstacle distances and target direction as inputs for the MANFIS controller to generate initial steering angles.
• Employing the MOSFO technique for final steering angle determination and implementing a dining philosopher controller for deadlock resolution.

Main Results:

• Simulations in the WEBOT simulator demonstrated a deviation under 5% when validated with real-time experiments, indicating robustness.
• The proposed technique showed an average improvement of 6.12% (ankle), 7.05% (knee), and 15.04% (hip) in joint torques compared to the NAO default controller.
• The strategy also presented improvements in travel length when compared against existing navigation strategies in multi-robot systems.

Conclusions:

• The hybridized MANFIS-MOSFO controller offers a superior and robust solution for humanoid robot trajectory planning in cluttered terrains.
• The implementation effectively resolves deadlock issues in multi-humanoid systems.
• The proposed method significantly enhances navigational efficiency and joint torque performance, outperforming existing approaches.