Development of a new motion algorithm for walking, stair climbing and obstacle avoidance in humanoid robots

Humanoid robots are robot platforms that most closely resemble human physiology in terms of their physical properties. Therefore, humanoid robots are generally used for human-robot interaction. However, controlling humanoid robots is very difficult due to their similarity to human physiology. To achieve ideal control of a humanoid robot, kinematic and dynamic analyses must be performed. The analyzed robot is then programmed for different tasks with control algorithms. In this thesis, kinematic calculations were performed using the geometric properties of humanoid robots to obtain the working space. Motion trajectories were generated using Polynomial and Bezier curve generation methods. During the created movements, the balance state of the robot was observed through dynamic analysis, and the ideal step distances for walking and climbing stairs were calculated by determining the Zero Moment Point (ZMP). Thanks to the kinematic and dynamic calculations, Model Predictive Control (MPC) was utilized for obstacle avoidance, enabling ideal trajectory planning. ROBOTIS OP3 humanoid robot was used for the experimental studies. The control algorithm was developed in the Robot Operating System (ROS) environment. Simulation studies were realized in the GazeboSim environment of ROS. The walking and control algorithms developed in the simulation environment were directly transferred to the real robot thanks to ROS. The joint angles of the robot were obtained and the performance of the optimal trajectory was calculated. Motion trajectories developed for walking and climbing stairs were applied both in the simulation environment and on the real OP3 robot. The motion performance of the robot without falling was analyzed using ZMP calculations. The results were presented in figures. The obtained results demonstrated that the developed trajectories and control algorithms work successfully on the real robot.