Self Localization Based On Neighborhood Probability Mapping for Humanoid Robot

Abstract

The humanoid robot competition is an autonomous robot with a human-like body platform with a single camera as a vision sensor and balancing sensor to support them to play soccer in the specific field. The technical challenges in this competition such following the ball, running during search the ball, dynamic walking, kicking while maintaining the balance body condition, decision making with other robot, localization and mapping as research issues investigated in the Humanoid competition. Localization and mapping still big challenges in humanoid competition, it was only single camera is used in competition rule and no others sensor to support the position and orientation during playing the game. The proposed system was developed is neighborhood probability mapping. The long-term goal of this research is to realize an ideal system to accelerate the redesign field condition and implementation process in a humanoid robot that can be monitored in real-time. The aim of this research is to take the opportunities: (a) increasing the robot's performance of vision and intelligence on the humanoid robot; (b) with this SLAM method the robot can distinguish between the balls that are in the field and outside the field; (c) able to distinguish the enemy goal from the goal itself based on goal detection and line detection; (d) the goal keeper robot capable of acting as an attacker and scanning the kick towards the enemy goal. The testing condition was implemented between simulation testing and real testing in same times. Based on the data experimental result, the robot can estimate their position and orientation during searching the ball position, goal position and obstacle coordinate with high real time accuracy. The result shows that the proposed system can be applied to the humanoid soccer robot in the real time directly and it worked with less error.