Abstract:

The main objective of this thesis is to design a footstep-planning algorithm for humanoid robots. Unlike wheeled robots, humanoid robots can negotiate obstacles and climb stairs, and is therefore more suitable for navigation in more complex environments. As the algorithm takes height and horizontal coordinates into account, the humanoid robot can plan its steps in three dimensions, allowing it to avoid collisions and remain stable throughout.

Both computational error and impact during the landing of a foot might give rise to a difference between the command trajectory and the simulation result. The robot can remain stable, but there is a possibility of collision in a complex environment. To reduce this, we use feedback information from each foot to modify the trajectory of the next step.

The momentum generated by a moving foot might cause an unexpected rotational error, and we compensate for it by swinging the robot body and arms, while restricting the available range of swing to avoid exaggerated swinging motions.

The simulation physical environment was constructed using ADAMS, and all of the control code was built in MATLAB. These two environments are connected by Simulink in MATLAB.