Improving the localization of humanoid soccer robots in specified fields: A neural network approach

This article presents a novel approach for minimizing the errors of localization to achieve more precise results on humanoid soccer robots. The proposed approach combines four major localization methods which are 3 flags, 2 flags and sqrt, 2 flag and sin/cos, and 1 flag and Gyro with taking advantage of neural networks. The approach enables a robot do localization with the least errors related to the actual position. We prepared some experiments to support our technique. Experimental results show the precise real position of robot calculated efficiently which make it enable to use the approach in real-time environments.