PhD Candidate: Wanxin Wang

Abstract

Inverse kinematics or IK is widely used for robotic control, animation generation in the film industry, and computer video games. IK usage for robotic control requires high fidelity -- the IK solution must be feasible so it could be applied directly to control the robotic articulated parts movement without violating the defined constraints, but there is no strict latency requirement. Animation generation generally does not require such high fidelity, the animators generate postures through manipulating IK joints (end effectors). Lower latency is required for animators’ working efficiency, it is not expected for animators to wait too long until the posture is generated. IK usage for computer video games requires very high interactive rate on a low DOF kinematic chain. For example, a hand pushing the door has a kinematic chain from its shoulder to hand that consists only two or three joints. The IK problems are usually predictable for game scenarios, a game scenario includes many characters equipped with IKs, thus an individual IK must work fast and require little computational resource.

In this dissertation, I developed a hybrid inverse-kinematics (IK) solution for real time usage, consisting of different numerical computation methods working on different kinematic chains, the results of which are integrated in a sequential IK framework, with help from recorded postures and transitions in forward-kinematics or FK configurations, will yield a satisfactory result in terms of both fidelity, computational speed, and robustness. 

Advisor: Joe Kearney

Please contact Wanxin Wang for further details, if you wish to join his exam.