Developing an End-to-End Vision-to-Control Pipeline
Developing an End-to-End Vision-to-Control Pipeline
Created using ChatSlide
This presentation delves into a sophisticated robotics project leveraging the SmolVLA in a LIBERO/MuJoCo simulation to achieve task-oriented robotic motions, characterized by non-random actions. The project predominantly focuses on selecting a fine-tuned model compatible with the Franka Panda robot arm, and utilizing advanced simulation environments like PyTorch and LeRobot within MuJoCo. Key infrastructure choices include using WSL2 Ubuntu and Conda with Python 3.12 to optimize LIBERO's...