Geoelements is an interdisciplinary research group at UT Austin focused on advancing AI and numerical methods for robotics and control. Our research combines cutting-edge techniques including Gaussian Splatting, Graph Neural Networks (GNNs), Material Point Method (MPM), and Reinforcement Learning (RL) to enable robots and autonomous systems to navigate and interact with complex, unstructured environments.
Gaussian Splatting: 3D scene reconstruction and rendering for robot perception.
Graph Neural Networks: Physics-informed learning for dynamics prediction and control.
Material Point Method: High-fidelity simulation of robot-environment interactions.
Reinforcement Learning: Adaptive control strategies for complex robotic tasks.
Differentiable Simulation: End-to-end learning through physics simulators.
XR Visualization: Immersive interfaces for robot teleoperation and training.
