Joint Motion Control Inputs for Smooth Physics-Based Animation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current motion control systems in physics-based animation rely heavily on human animators, which is time-consuming and impractical for scenarios requiring multiple characters with diverse movement characteristics, and machine learning-based neural networks often produce unnatural, jittery movements.
Innovation Solution
Training neural networks to accept sensor readings as inputs and output joint actuator commands, specifically using first or second derivatives of servo control commands, and incorporating reinforcement learning to optimize movement generation, ensuring smooth and life-like character motions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If hand animation is used to create life-like movements, then movement quality and realism are improved, but production time and cost increase significantly
Solution Approach 1:
The patent uses motion capture technology to record real human movements and creates virtual motion libraries that can be copied and applied to multiple characters. This allows realistic movements to be captured once and reused extensively, dramatically reducing production time while maintaining movement quality.
Solution Approach 2:
The system enables characters to generate their own movements through physics-based animation and procedural generation techniques. Characters can autonomously navigate, react to environment, and perform actions without requiring frame-by-frame hand animation, significantly reducing animator workload while maintaining natural movement characteristics.
2Reliability
If motion capture is used to record human movements, then movement realism is improved, but cost and time increase for large numbers of characters
Solution Approach 1:
The patent creates universal motion libraries from motion capture data that can be applied across multiple character types and scenarios. A single motion capture session can generate reusable animation assets that serve multiple purposes and character archetypes, eliminating the need to capture motions separately for each character.
Solution Approach 2:
The system uses procedural animation and parameter-driven motion synthesis to adapt captured movements to different characters by modifying parameters such as scale, speed, and joint constraints. This allows a single motion capture dataset to generate varied realistic movements for characters with different anatomies and movement styles.
3Productivity
If neural networks are used to generate character movements, then production time is reduced, but movement quality becomes unnatural and jittery
Solution Approach 1:
The patent implements feedback mechanisms where neural network-generated movements are evaluated against physics constraints and motion quality metrics. The system uses reinforcement learning and iterative optimization to refine generated movements, providing feedback loops that correct unnatural or jittery motions while maintaining generation speed.
Solution Approach 2:
The system performs preliminary training of neural networks using high-quality motion capture data and physics-based animations during an offline phase. This pre-training establishes realistic movement patterns and constraints that guide subsequent real-time generation, ensuring natural movements are produced quickly without requiring extensive runtime computation.
4Productivity
If random movement models are applied to multiple characters, then animator workload is reduced, but character individuality and scene realism decrease
Solution Approach 1:
The patent applies different motion styles, animation parameters, and behavioral characteristics to individual characters based on their specific attributes and roles. Each character can have customized motion profiles, preferred movement patterns, and unique animation parameters that preserve individuality while maintaining overall scene coherence and realism.
Data Source
AI summary
A method, system and non-transitory instructions for control input, comprising, taking an integral of an output value from a Motion Decision Neural Network for a movable joint to generate an integrated output value. Generating a subsequent output value using a machine learning algorithm that includes a sensor value and the integrated output value as inputs to the Motion Decision Neural Network and imparting movement with the moveable joint according to an integral of the subsequent output value.