synchronized multi-character motion editing

Synchronized Multi-character Motion Editing

Manmyung Kim, Kyunglyul Hyun, Jongmin Kim, Jehee Lee

Seoul National University

Multi-character Interaction: synchronization in space and time

Cumbersome to Maintain Synchronization

Edit while Maintaining Multiple Character Interaction

Related Work

A hierarchical approach to interactive motion editing for human-like figures. LEE, SIGGRAPH 99.

• Continuous motion editing

– make a smooth change to the motion to satisfy user-specified constraints

Related Work

A hierarchical approach to interactive motion editing for human-like figures. LEE, SIGGRAPH 99.

Motion path editing. GLEICHER, I3D 2001.

• Continuous motion editing

– make a smooth change to the motion to satisfy user-specified constraints

Related Work

Interactive control of avatars animatedwith human motion data. LEE, SIGGRAPH 2002.

Motion GraphsKOVAR, SIGGRAPH 2002.

• Structural motion synthesis

– splice motion segments to synthesize a novel motion sequence

Related Work

Group Motion Editing.Kwon, SIGGRAPH 2008.

• Group Motion Editing

– the locomotion of pedestrians

Overview

Multiple character interaction

Interactive motion path manipulation

Handling large deformation

Overview

Multiple character interaction

Interactive motion path manipulation

Handling large deformation

Overview

Multiple character interaction

Interactive motion path manipulation

Handling large deformation

Multiple Character Interaction

Multiple Character Interaction

Pinningposition

Multiple Character Interaction

Pinningposition

Pinning direction

Multiple Character Interaction

Pinningposition

Pinning direction

Relativepostion & direction

Multiple Character Interaction

Pinningposition

Pinning direction

Relativepostion & direction

Variational relative

Multiple Character Interaction

Pinningposition

Pinning direction

Relativepostion & direction

Variational relative

End-effector

Multiple Character Interaction

Pinningposition

Pinning direction

Relativepostion & direction

Variational relative

End-effector

Multiple Character Interaction

Pinningposition

Pinning direction

Relativepostion & direction

Variational relative

End-effector

Absolute time

Multiple Character Interaction

Pinningposition

Pinning direction

Relativepostion & direction

Variational relative

End-effector

Absolute time

Synchronization

Multiple Character Interaction

Variational relativePinningposition

Pinning direction

Absolute time

End-effector

Synchronization

Relativepostion & direction

Formulated as linear equations

Absolute Position, Direction, and Timing

Absolute Position, Direction, and Timing

Relative Position, Direction, and Timing

Relative Position, Direction, and Timing

End-effector Constraints

Motion Path Editing

Based on Laplacian mesh editing [Igarash 2005; Sorkine 2004]

– deform curve in as-rigid-as possible manner

Linear least squares problems : efficient

Applying Laplacian formulation to Motion Path

Project root trajectory onto the ground

Applying Laplacian formulation to Motion Path

Project root trajectory onto the ground

Define the direction by tangent and normal vectors

Tangent Vector

Normal Vector

Handling Degenerate Cases : Stationary path

Stationary motion tends to stretch unrealistically

Treat stationary portion as rigid segment using hard con-straints

Treat as rigid segmentStretch unrealistically

Handling Degenerate Cases : Stationary path

Stationary motion tends to stretch unrealistically

Treat stationary portion as rigid segment using hard con-straints

Treat as rigid segmentStretch unrealistically

Handling Degenerate Cases : Tangent Flipping

Small deformation could flip tangent directions

Handling Degenerate Cases : Tangent Flipping

Small deformation could flip tangent directions

Handling Degenerate Cases : Tangent Flipping

Small deformation cause a tangent direction to flip

Determine new tangent vector by linear interpolation

Tangent interpolationTangent flipping

Post-processing touch-up

End-effector constraints involve non-linear equations

: iterative inverse kinematics solver

Pragmatic solution : Motion path editing IK-based refinement

Full-body Refinement

Time Warping

Smooth time-warp to meet timing constraintsAbsolute time Synchronization

Time Warping

Smooth time-warp to meet timing constraintsAbsolute time Synchronization

User Manipulation

Time Warping

Smooth time-warp to meet timing constraints

Timeline and spatial path are motion curves

− the same Laplacian curve editing method

Handling Large Deformation

Only Laplacian path editing

Laplacian path editing Discrete motion editing

Handling Large Deformation

Only Laplacian path editing

Laplacian path editing Discrete motion editing

Motion graph

− identify similar frames and create transitions

Discrete Transformations

Motion graph

− identify similar frames and create transitions

There are exponentially many sequences of dis-crete transformations

− structurally-varied motion path

Discrete Transformations

Interactive editing is inherently incremental

− motion path change gradually

Incremental Change

Interactive editing is inherently incremental

− motion path change gradually

Three local transformations : delete, insert, replace

− interactive performance & predictable control

Incremental Change

Deletion

Types of Discrete Transformation

Deletion

Types of Discrete Transformation

Insertion

Types of Discrete Transformation

Replacement

Types of Discrete Transformation

Evaluation of Discrete Transformation

E = Espatial Etemporal Epenalty

Espatial : spatial deformation energy

Etemporal : temporal deformation energy

Epenalty : penalize lengthening and shortening of motion path

Evaluate deformation energy of Laplacian path editing to meet user constraints

Evaluation and Selection

Evaluation and Selection

Our Algorithm

Update user constraints

Update user constraints

Enumerate all possible transformations

Our Algorithm

Update user constraints

Enumerate all possible transformations

Evaluate each transformation to select the best

Our Algorithm

Update user constraints

Enumerate all possible transformations

Evaluate each transformation to select the best

Laplacian path editing

Our Algorithm

Update user constraints

Enumerate all possible transformations

Evaluate each transformation to select the best

Laplacian path editing

Full-body refinement

Deformed motions

Our Algorithm

Update user constraints

Deformed motions

Enumerate all possible transformations

Evaluate each transformation to select the best

Laplacian path editing

Full-body refinement

Our Algorithm

Update user constraints

Deformed motions

Enumerate all possible transformations

Evaluate each transformation to select the best

Laplacian path editing

Full-body refinement

Performacebottleneck

Our Algorithm

Pruning Discrete Transformations

Prune transformations for interactive performance

– Duration

Pruning Discrete Transformations

Prune transformations for interactive performance

– Duration

– Enclosing

Pruning Discrete Transformations

Prune transformations for interactive performance

– Duration

– Enclosing

– Constraints

Deletion

Subsampling Acceleration Technique

For each discrete transformation, we evaluate its energy by solving Laplacian equations

Subsample motion paths in evaluating its deformation energy

Subsampling ratio is sparse such as 125

Discussion

Contribution

– a unified formulation of space, time, interaction

– combining continuous and discrete motion editing

– intuitive interface

Discussion

Contribution

– a unified formulation of space, time, interaction

– combining continuous and discrete motion editing

– intuitive interface

Future works

– handling 3D motion path

– non-linear constraints

Synchronized Multi-character Motion EditingManmyung Kim, Kyunglyul Hyun, Jongmin Kim, Jehee Lee