SOUS VIDE: Cooking Visual Drone Navigation Policies in a Gaussian Splatting Vacuum

TL;DR

This paper introduces SOUS VIDE, a comprehensive system combining a novel simulator, training approach, and policy architecture for visual drone navigation that achieves robust zero-shot transfer from simulation to real-world environments using only onboard perception.

Contribution

The paper presents FiGS, a high-fidelity visual simulator, and SV-Net, an end-to-end visuomotor policy architecture, enabling robust drone navigation with minimal real-world training.

Findings

Policies are robust to 30% mass variations.

Effective under 40 m/s wind gusts.

Maintain performance despite scene changes and moving objects.

Abstract

We propose a new simulator, training approach, and policy architecture, collectively called SOUS VIDE, for end-to-end visual drone navigation. Our trained policies exhibit zero-shot sim-to-real transfer with robust real-world performance using only onboard perception and computation. Our simulator, called FiGS, couples a computationally simple drone dynamics model with a high visual fidelity Gaussian Splatting scene reconstruction. FiGS can quickly simulate drone flights producing photorealistic images at up to 130 fps. We use FiGS to collect 100k-300k image/state-action pairs from an expert MPC with privileged state and dynamics information, randomized over dynamics parameters and spatial disturbances. We then distill this expert MPC into an end-to-end visuomotor policy with a lightweight neural architecture, called SV-Net. SV-Net processes color image, optical flow and IMU data streams into low-level thrust and body rate commands at 20 Hz onboard a drone. Crucially, SV-Net includes a learned module for low-level control that adapts at runtime to variations in drone dynamics. In a campaign of 105 hardware experiments, we show SOUS VIDE policies to be robust to 30% mass variations, 40 m/s wind gusts, 60% changes in ambient brightness, shifting or removing objects from the scene, and people moving aggressively through the drone's visual field. Code, data, and experiment videos can be found on our project page: https://stanfordmsl.github.io/SousVide/.