Machine-Vision-Assisted Automated Closed-Loop Manipulation of Open-Surface Droplets

Automated manipulation of open-surface droplets is important for microscale sample handling, reaction assembly, and biochemical analysis, yet existing systems often lack obstacle-aware planning, control-state monitoring, and disturbance recovery. Here, we present a machine-vision-assisted closed-loop system for programmable, noncontact droplet manipulation on a photogated corona microfluidic (PGCM) platform. The system integrates real-time localization, rapidly exploring random tree planning, robotic laser actuation, and hierarchical feedback control. An inner loop guides the laser spot along the planned path, while an outer loop monitors the droplet–laser binding state. Obstacle changes trigger online replanning, whereas loss of binding triggers laser return, droplet recapture, and task recovery. The system transported 1–50 μL droplets, performed sequential coalescence, static and dynamic obstacle avoidance, complex-path manipulation, multistep reactions, and droplet-mediated transport of a flexible porous solid target. Following a manually imposed voltage interruption, it automatically detected loss of binding and resumed the task after actuation was restored. This work provides a programmable and disturbance-tolerant visual closed-loop framework for automated open-surface droplet manipulation.