Event
Sim-to-real framework proposed for agricultural grasping VLA
1 reports · 1 sources
The research team proposed a sim-to-real robot learning framework for agricultural reaching and grasping: an agricultural environment containing multiple tomato plant structures was built in MuJoCo, expert demonstrations were collected via joystick teleoperation, and two pretrained generalist policies, Octo and π0.5, were first adapted using simulated demonstrations and then fine-tuned with real demonstrations on a Franka Research 3 robotic arm. Results show that π0.5 achieved a higher task performance score than Octo and degraded less under distribution shifts such as viewpoint and spatial offsets; the study also examined the impact of proprioceptive information on policy performance. Evaluation covered in-distribution, viewpoint/spatial-shift, and out-of-distribution conditions in simulation, and real-robot experiments were also conducted.
Report timeline
- Vision-Language-Action Model Robot Learning Framework for Agricultural Manipulation
A research team proposed a simulation-to-real robot learning framework for agricultural reaching and grasping, building a MuJoCo-based agricultural environment with diverse tomato plant structures and collecting expert demonstrations via joystick teleoperation. Two pretrained generalist policies, Octo and π0.5, were first adapted with simulated demonstrations, then fine-tuned with real demonstrations on a Franka Research 3 robotic arm. Results showed that π0.5 scored higher than Octo on task performance and degraded less under distribution shifts such as viewpoint and spatial offsets; the study also examined the effect of proprioceptive information on policy performance.
arXiv cs.RO ·