LLM-Driven Embodied AI Task Generation via Genesis
This project investigates scalable task generation for Embodied AI by leveraging Genesis, a state-of-the-art, high-fidelity simulation platform for general-purpose robotics and physical interaction. A digital twin of the Kinova Gen3 robotic arm was developed in Genesis, with inverse kinematics and Python-based control enabling the arm to perform manipulation tasks such as picking and poking. To automate large-scale scenario creation, a large language model (LLM) was used to dynamically generate and modify control scripts and scene parameters, supporting the procedural synthesis of diverse, sensorimotor-grounded robotic tasks.