Michigan researchers' STEMbot climbs plant stems for early pest detection

A team led by Zachary Charlick from the University of Michigan (likely) introduced STEMbot, a miniature climbing robot designed to autonomously navigate under plant canopies for early pest detection. Unlike existing platforms that lack onboard perception or are limited to unbranched trunks, STEMbot integrates a fully geometric PIN-SLAM pipeline with a semantic OcTree for robust localization and mapping while climbing. It uses a manifold-constrained A* planner with ray-tracing goal specification, enabling branch-aware traversal and inspection of occluded targets. In hardware experiments, STEMbot reliably climbed stems ranging from 7–33mm and navigated across four distinct plant specimens. Quantitative evaluations show high-fidelity geometric reconstructions with an average Chamfer distance under 1cm compared to offline photogrammetry, confirming globally consistent odometry critical for autonomous navigation.

The paper, accepted to IROS 2026, addresses a key bottleneck in organic agriculture scalability: labor costs for pest monitoring. Many pests live on leaf undersides or stems, detectable only after significant damage. STEMbot’s ability to inspect these hidden areas could enable early intervention, reducing crop loss and pesticide use. Its compliant design allows safe interaction with delicate plant structures, while the SLAM-based autonomy eliminates the need for GPS under dense foliage. The work opens pathways for swarms of such robots to monitor large fields, potentially transforming precision agriculture.