Register now After registration you will be able to apply for this opportunity online.
Learning Rigid Objects Dynamics with Graph Networks
Researchers have started to explore data-driven physics simulations, particularly with Graph Neural Networks for rigid objects collisions. However, simulating rigid collisions among arbitrary shapes is notoriously difficult due to complex geometry and the strong non-linearity of the interactions. In this project, you will focus on the task of learning/simulating rigid objects dynamics with Graph
Neural Networks (GNNs), with the end-goal of predicting future or alternative trajectories for physical rigid objects in a scene.
In this project, you will focus on the task of learning/simulating rigid objects dynamics with Graph
Neural Networks (GNNs), with the end-goal of predicting future or alternative trajectories for the
objects in the scene.
This task includes the modeling of 3D objects with graphs (particles, meshes, ...), the design
of models suited for physics simulations (accounting for physical and mechanical laws) and the
prediction of future (or alternative) trajectories (also called rollout).
In this project, you will focus on the task of learning/simulating rigid objects dynamics with Graph Neural Networks (GNNs), with the end-goal of predicting future or alternative trajectories for the objects in the scene.
This task includes the modeling of 3D objects with graphs (particles, meshes, ...), the design of models suited for physics simulations (accounting for physical and mechanical laws) and the prediction of future (or alternative) trajectories (also called rollout).
The objectives of the project are:
- Understanding the core concepts of GNNs and their application to modeling and simulation
- Identify the current state-of-the-art GNNs methods for simulating rigid objects dynamics
- Reproducing a subset of the selected baseline methods on existing datasets
- If possible, extend the existing approaches by a method of your choice (e.g. embedding physical inductive biases in the architecture, improving the explainability, reducing the data or computational requirements, ...)
The objectives of the project are:
- Understanding the core concepts of GNNs and their application to modeling and simulation
- Identify the current state-of-the-art GNNs methods for simulating rigid objects dynamics
- Reproducing a subset of the selected baseline methods on existing datasets
- If possible, extend the existing approaches by a method of your choice (e.g. embedding physical inductive biases in the architecture, improving the explainability, reducing the data or computational requirements, ...)