A reliability-based particle filter for humanoid robot self-localization in RoboCup Standard Platform League.

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Humanoid robot localization is crucial for autonomous navigation in dynamic environments like the RoboCup Standard Platform League.
• Accurate position estimation is essential for effective decision-making and task execution in robot competitions.
• Resource limitations on robot platforms necessitate efficient algorithms for real-time localization.

Purpose of the Study:

• To develop and validate a novel self-localization method for humanoid robots in the RoboCup Standard Platform League.
• To improve the accuracy and computational efficiency of robot localization algorithms.
• To address the challenges of position estimation degradation and loss in realistic game scenarios.

Main Methods:

• Implementation of a vision system on the Nao robot for detecting field markers.
• Development of extrinsic and intrinsic camera calibration procedures to minimize measurement errors.
• Application of a particle filter algorithm with a new particle selection strategy for position estimation.

Main Results:

• The proposed method demonstrated effective localization even in challenging situations such as falls or penalization.
• The new particle selection strategy significantly reduced CPU computing time per iteration.
• Experimental results confirmed the algorithm's quality in terms of both localization accuracy and computational efficiency.

Conclusions:

• The developed localization method offers a robust and efficient solution for humanoid robots in the RoboCup environment.
• The novel particle selection strategy effectively overcomes resource limitations common in robot platforms.
• This work contributes to advancing autonomous navigation capabilities in humanoid robotics.