Event
Open-Source Imitation Learning Pipeline for Strawberry Peduncle Grasping
1 reports · 1 sources
Nitin Rai et al. proposed and open-sourced a visuomotor imitation learning pipeline for strawberry peduncle grasping, comprising human teleoperation demonstration collection, ACT policy training, and autonomous deployment on an edge platform, with the dataset, model weights, and code made publicly available. The policy uses ResNet-50 as its visual backbone, was trained on 130 demonstrations over 100K gradient steps, with loss converging from 2.84 to 0.091, and was autonomously deployed at 15 Hz on a Jetson AGX Orin edge platform. The paper has been accepted to the IROS 2026 Agricultural Robotics Workshop.
Report timeline
- Research on Pedicel-Targeted Strawberry Harvesting via Imitation Learning on a Low-Cost Teleoperated Robotic Arm
This work proposes a visuomotor imitation learning pipeline for pedicel-targeted strawberry harvesting that unifies perception, motion planning, and contact manipulation: data is collected from human teleoperation demonstrations, an Action Chunking with Transformers (ACT) policy with a ResNet-50 visual backbone is trained, and the policy is deployed autonomously on a Jetson AGX Orin edge platform at 15 Hz.
arXiv cs.RO ·