MSc.Thesis Defense:Buğra Saraç

HUMANOID ROBOT LEG AND ARM COORDINATION FOR FORCE CONTROL APPLICATIONS

Buğra Saraç

Mechatronics Engineering MSc. Thesis 2024

Thesis Jury

Assoc. Prof. Kemalettin Erbatur (Thesis Advisor),

                                                                                                   Prof. Mustafa Ünel,

                                                                                             Assoc. Prof. Dr. Özkan Bebek

Date & Time: 22nd of July 2024 – 3:30 PM

Place: FENS L056

Keywords: Humanoid, Biped, Pushing Control, Full-Body Motion Control, Ground Interaction Forces, Balance Control, Zero Moment Point

Abstract

Humanoid robots are designed for a future world where they act in a multiplicity of roles in the human environment. Imitation of human kinematic arrangement and size lends the robot ability of reaching, manipulating and operating user interfaces and tools designed for humans. The field is rich and vast covering subjects of soft robotics and safety for robot-human coexistence and cognitive research for conditional awareness and planning. This thesis deals with the force interaction of the robot through its hand end effectors. The emphasis is on applying force to an immovable object.

Pushing a tool or an object in the robot hand against a fixed surface has many application places in the industrial environment. Exertion of highest force achievable with the available mechanical system and actuators is desired. The bipedal structure, however, poses balance restrictions on the humanoid robot. When the hands push the reactive forces can cause the robot loose balance and fall.

The advantages of the bipedal humanoid robot structure in the human environment are accompanied by serious control challenges due to complex dynamics and balancing requirements.

The thesis presents a full-body reference generation and control technique in which the robot maintains balance while pushing the wall in front of it. A criterion based on the Zero Moment Point concept is employed to compute leg joint references. Added to leg joint control torques are feedforward torque components obtained from the end contact force references. These additional components are computed using optimized foot-ground interaction forces. Robot arms are controlled over hybrid position-force control. Various reference synthesis and control modes are coordinated for leaning on the wall and enlarging the support polygon on the ground.

A full-dynamics three-dimensional simulation and animation environment is employed for the development and tests of the proposed technique