Human motion deformation for obstacle avoidance by optimizing joint positions

In recent years, human motion deformation techniques have been widely utilized in animation editing and motion synthesis to adapt existing motion data to various environmental conditions. One key application is obstacle avoidance, where motion must be deformed to avoid collisions while maintaining natural motion. Traditional methods, such as lattice-guided deformation, have shown effectiveness but often result in unnecessary avoidance or unnatural postures. This study proposes a novel motion deformation method that optimizes joint positions based on a weighted evaluation function. By computing joint positions through forward kinematics, optimizing them to satisfy constraints, and converting them back into joint angles using inverse kinematics, the method enables real-time, frame-by-frame deformation. The evaluation function balances obstacle avoidance, motion smoothness, and joint structure preservation, and gradient descent is used for optimization. Experimental results have confirmed that this method is capable of real-time processing in small-scale obstacle avoidance and minimizes unnatural postures. On the other hand, issues remain for large-scale deformation and improvement of computational efficiency. Future research will aim to apply this method to various motion data and obstacles of other shapes.