Generating Humanoid Robot Motions based on a Procedural Animation IK Rig Method

Abstract

Many researchers have explored using humanoid robots to perform various tasks within living environments. Although well-designed motions are essential for providing friendly interactions with humanoid robots, previous works have emphasized task efficiency over user-friendliness in generated motions. To bridge this gap, we propose a motion generation method for humanoid robots which is based on procedural animation with inverse-kinematics (IK) rig methods, commonly used in video games and computer graphics (CG). First, we prepared the robot’s 3D model by rigging it with two different rig structures, humanoid armature and robot armature. Then we create the IK rig from the humanoid armature, and set the robot-armature’s motion objectives based on the humanoid-armature transform information. Second, to make the motion adjustable, we defined a pole-vector method on the IK rig to define the middle joint’s position dynamically. Therefore, using our approach, we are able to use commonly available CG character animations on humanoid robots, and adjust motions to match various contextual or task requirements. We evaluated our approach’s ability to generate humanoid robots motions that mimic humanoid character’s animations, as well as the adjustability of generated motions to different contextual requirements. We used three 3D CG models of humanoid robots with different body configurations. The results show that our approach is successful in generating appropriate motions on the humanoid robot rigs based on five animations. The results also show the potential of our approach to adjust motion to correspond to factors like non-planar terrain or task-specific requirements. In light of our results, we discuss the advantages of our approach and potential applications to generate interactive motions or for task requirements.