Abstract : Real-time rigid-body simulation commonly employs simplified aerodynamic models, where drag forces are applied only to translational motion. Although such models efficiently reduce velocity, they fail to generate stable and physically plausible rotational behavior, because asymmetric aerodynamic effects are not explicitly modeled. As a result, rotational motion often appears transient or inconsistent, especially in interactive real-time environments. This paper proposes an air-mesh-based aerodynamic torque model that approximates rotational effects using local geometric variation and relative velocity, without explicitly solving fluid dynamics. The proposed method constructs a lightweight proxy representation around a moving rigid body and estimates resistance from area variation signals. These resistance vectors are accumulated as torque, allowing rotational motion to emerge naturally from asymmetric aerodynamic effects. All computations are designed to be element-wise parallel, making the method suitable for GPU-based real-time simulation. Experimental results demonstrate that, compared with linear drag-only models, the proposed approach produces sustained and directionally consistent rotation while maintaining stable real-time performance. Trajectory and pose comparisons in both 2D and 3D scenarios show stronger coupling between translational and rotational motion, resulting in more physically plausible behavior without the computational cost of CFD-based methods. Overall, the proposed model provides a computationally efficient yet physically plausible approximation of aerodynamic torque, making it practical for real-time physics simulation, interactive graphics, and virtual environments.

[paper]