The Development Of Open-Source Humanoid Robot Platform And Stability Control System

This thesis work focused on developing an open source and 3D-printable multifunctional platform that researchers can use to create and verify their machine learning algorithms to a humanoid robot.In our case,it can help to solve one of the balancing problems-a humanoid balancing on a real time moving planar surface.The presented mechanical platform is durable,easv to fix.upgrade and fast to assemble.In addition,using created platform,we present an approach that solves a humanoid balancing problem,which uses only fully connected neural network method as a basic idea for real time balancing.We can divide our balancing algorithm into 3 parts:1)sensors detect the current position of the body and generate the input data for the NN,2)according to that data,NN produce the output commands,3)using servomotors,NN produces movements that will keep a balanced state of the entire robot.During testing,we proved that the robot could balance on the moving planar surface that randomly tilts up to 10 degrees in any direction.We demonstrate that using our created platform researchers can easily verify their algorithms in real physical conditions and compare the performance of various neural networks.All these can be useful to researchers for analysis in the field of robotics and machine learning.Finally,the thesis work solves the actual problems,such as:·Developing an open source platform that can be easily recreated by researchers.·The new balancing method that replaces PID-systems with NN-systems.