Rig My Ride: Automatic Rigging of Physics-based Vehicles for Games

1 McGill University
2 École de Technologie Supérieure
3 Roblox

Presented at the Symposium on Computer Animation (SCA) 2025
★ Received the best poster award ★

Our pipeline automatically produces game-ready vehicles.

Abstract

We extend the concept of traditional rigging, which links polygonal meshes to an underlying skeleton for 3D characters, to the creation of physics-based wheeled vehicle models directly from surface geometry. Unlike character rigging, physics-based rigging involves assigning joints and collision proxies to animate the surface geometry. We present an automated pipeline that transforms a polygon soup into a physics-based, multi-wheeled vehicle model. The pipeline begins by using text-driven 2D image segmentation to identify vehicle components, which are then mapped onto the 3D mesh. A rough estimate of collision geometries and joint parameters is then used to initialize a rigid body simulation of the vehicle. Then, a numerical optimization refines these parameters in order to produce more realistic vehicle behaviour. The final result is a functioning physics-based vehicle for real-time simulations, which is demonstrated across a variety of vehicles, including cars, tricycles, lunar rovers, and even a semi-truck with 10 wheels.

Pipeline of our method

Given an input mesh of a vehicle (left), our method automatically produces a rigged version of the vehicle mesh (right) that we can drive in an interactive physics-based simulation. The pipeline of our approach first uses multi-view rendering with a prompt based segmentation to produce 2D wheel masks via GLIP and SAM. Masks are projected onto the model and clustered to identify sets of 3D points that make up the wheels. Cylinders are fit to the point sets, and a gradient free minimization uses physics simulation to optimize the wheel geometry and joint constraints so as to produce a functional rig for interactive simulation.

Supplemental Video

BibTeX


@article{Katz2025RMR,
  author = {Katz, Melissa and Kry, Paul G. and Andrews, Sheldon},
  title = {Rig My Ride: Automatic Rigging of Physics-based Vehicles for Games},
  year = {2025},
  journal = {Proceedings of the ACM on Computer Graphics and Interactive Techniques},
  volume = {8},
  number = {4},
  doi = {10.1145/3747861}
}