Techniques for solving inverse kinematic problems using trained machine learning models
In various embodiments, a computer animation application automatically solves inverse kinematic problems when generating object animations. The computer animation application determines a target vector based on a target value for a joint parameter associated with a joint chain and at least one of a target position or a target orientation for an end-effector associated with the joint chain. The computer animation application executes a trained machine learning model on the target vector to generate a predicted vector that includes data associated with multiple joint parameters associated with the joint chain.
1 . A computer-implemented method for automatically solving inverse kinematic problems when generating object animations, the method comprising:
determining a first target vector based on a first target value for a first joint parameter associated with a first joint included in a joint chain and at least one of a first target pose or a first target orientation for an end-effector associated with the joint chain, wherein the joint chain includes a plurality of joints associated with a skeleton of an object being animated and the first target value for the first joint parameter comprises at least a maximum angle or a minimum joint angle for constraining a rotation of the first joint; and
executing a trained machine learning model on the first target vector to generate a first predicted vector that includes data associated with a first plurality of joint parameters associated with the joint chain.
2 . The computer-implemented method of claim 1 , wherein the first joint parameter is an input parameter of the trained machine learning model and is not included in the first plurality of joint parameters.
3 . The computer-implemented method of claim 1 , wherein the data associated with the first plurality of joint parameters comprises a different predicted value for each joint parameter included in the first plurality of joint parameters.
4 . The computer-implemented method of claim 1 , wherein the data associated with the first plurality of joint parameters comprises at least one of a first predicted sine of a first predicted joint angle or a first predicted cosine of the first predicted joint angle.
5 . The computer-implemented method of claim 1 , further comprising determining the first target value for the first joint parameter based on input received via an interactive graphical user interface component that provides control over the first joint parameter for inverse kinematics and is displayed within a graphical user interface.
6 . The computer-implemented method of claim 1 , wherein determining the first target vector comprises performing one or more encoding operations on the first target value to generate at least one element included in the first target vector.
7 . The computer-implemented method of claim 1 , further comprising determining the at least one of the first target pose or the first target orientation for the end-effector based on at least one of a first pose or a first orientation for a start of the joint chain.
8 . The computer-implemented method of claim 1 , further comprising performing one or more decoding operations on at least a first element included in the first predicted vector to compute a predicted value for a second joint parameter included in the first plurality of joint parameters.
9 . The computer-implemented method of claim 1 , wherein the joint chain is included in the skeleton of the object that is being animated.
10 . One or more non-transitory computer readable media including instructions that, when executed by one or more processors, cause the one or more processors to automatically solve inverse kinematic problems when generating object animations by performing the steps of:
determining a first target vector based on a first target value for a first joint parameter associated with a first joint included in a joint chain and at least one of a first target pose or a first target orientation for an end-effector associated with the joint chain, wherein the joint chain includes a plurality of joints associated with a skeleton of an object being animated and the first target value for the first joint parameter comprises at least a maximum angle or a minimum joint angle for constraining a rotation of the first joint; and
executing a trained machine learning model on the first target vector to generate a first predicted vector that includes data associated with a first plurality of joint parameters associated with the joint chain.
11 . The one or more non-transitory computer readable media of claim 10 , wherein the first target vector comprises first data associated with the first target value, a second target value for a second joint parameter associated with the joint chain, and the at least one of the first target pose or the first target orientation for the end-effector.
12 . The one or more non-transitory computer readable media of claim 10 , wherein the data associated with the first plurality of joint parameters comprises a different predicted value for each joint parameter included in the first plurality of joint parameters.
13 . The one or more non-transitory computer readable media of claim 10 , wherein the data associated with the first plurality of joint parameters comprises at least one of a first predicted sine of a first predicted joint angle or a first predicted cosine of the first predicted joint angle.
14 . The one or more non-transitory computer readable media of claim 10 , further comprising determining the first target value for the first joint parameter based on input received via an interactive graphical user interface component that provides control over the first joint parameter for inverse kinematics and is displayed within a graphical user interface.
15 . The one or more non-transitory computer readable media of claim 10 , wherein determining the first target vector comprises computing at least one of a sine of the first target value or a cosine of the first target value to generate at least one element included in the first target vector.
16 . The one or more non-transitory computer readable media of claim 10 , further comprising determining the at least one of the first target pose or the first target orientation for the end-effector based on at least one of a first pose or a first orientation for a start of the joint chain.
17 . The one or more non-transitory computer readable media of claim 10 , further comprising performing one or more decoding operations on at least a first element included in the first predicted vector to compute a predicted value for a second joint parameter included in the first plurality of joint parameters.
18 . The one or more non-transitory computer readable media of claim 10 , wherein a second joint parameter included in the first plurality of joint parameters comprises a joint rotation parameter or a joint translation parameter.
19 . A system comprising:
one or more memories storing instructions; and
one or more processors coupled to the one or more memories that, when executing the instructions, perform the steps of:
determining a first target vector based on a first target value for a first joint parameter associated with a first joint included in a joint chain and at least one of a first target pose or a first target orientation for an end-effector associated with the joint chain, wherein the joint chain includes a plurality of joints associated with a skeleton of an object being animated and the first target value for the first joint parameter comprises at least a maximum angle or a minimum joint angle for constraining a rotation of the first joint; and
executing a trained machine learning model on the first target vector to generate a first predicted vector that includes data associated with a first plurality of joint parameters associated with the joint chain.