Learning to Turn: Diffusion Imitation for Robust Row Turning in Under-Canopy Robots

TL;DR

This paper introduces a diffusion imitation learning approach enabling under-canopy agricultural robots to perform robust row turning using only visual and velocity data, addressing navigation challenges in complex environments.

Contribution

It presents a novel diffusion policy-based imitation learning method for robust row turning in under-canopy robots, trained from demonstrations.

Findings

Simulation results show successful learning of row turning behaviors.

The approach works with visual observations and velocity states.

Challenges remain in control within rows and varied initial conditions.

Abstract

Under-canopy agricultural robots require robust navigation capabilities to enable full autonomy but struggle with tight row turning between crop rows due to degraded GPS reception, visual aliasing, occlusion, and complex vehicle dynamics. We propose an imitation learning approach using diffusion policies to learn row turning behaviors from demonstrations provided by human operators or privileged controllers. Simulation experiments in a corn field environment show potential in learning this task with only visual observations and velocity states. However, challenges remain in maintaining control within rows and handling varied initial conditions, highlighting areas for future improvement.