Push Past Green: Learning to Look Behind Plant Foliage by Moving It
This addresses the challenge of partial visibility and clutter in plant manipulation for applications like inspection and fruit plucking, representing an incremental advance in robotics for agriculture.
The paper tackles the problem of manipulating plant foliage to reveal hidden spaces for autonomous agriculture, using a data-driven method called PPG that predicts and executes actions to incrementally uncover areas, achieving effectiveness over hand-crafted baselines in synthetic and real plant experiments.
Autonomous agriculture applications (e.g., inspection, phenotyping, plucking fruits) require manipulating the plant foliage to look behind the leaves and the branches. Partial visibility, extreme clutter, thin structures, and unknown geometry and dynamics for plants make such manipulation challenging. We tackle these challenges through data-driven methods. We use self-supervision to train SRPNet, a neural network that predicts what space is revealed on execution of a candidate action on a given plant. We use SRPNet with the cross-entropy method to predict actions that are effective at revealing space beneath plant foliage. Furthermore, as SRPNet does not just predict how much space is revealed but also where it is revealed, we can execute a sequence of actions that incrementally reveal more and more space beneath the plant foliage. We experiment with a synthetic (vines) and a real plant (Dracaena) on a physical test-bed across 5 settings including 2 settings that test generalization to novel plant configurations. Our experiments reveal the effectiveness of our overall method, PPG, over a competitive hand-crafted exploration method, and the effectiveness of SRPNet over a hand-crafted dynamics model and relevant ablations.