Sensor-Guided Same-Layer Repair of Material-Loss Gaps

From measured gaps to localized repair

Material interruptions during robotic construction 3D printing can leave gaps in an already deposited layer. This research localizes those material-loss regions from measured geometry, returns the robot to the affected locations, and deposits material within the same layer. Post-repair laser measurements verify the geometric response.

Sensing and repair planning

Laser profiles describe deposited geometry and distinguish genuine material-loss regions from invalid scanner observations. The measured gaps define where localized repair is needed. The RGB camera documents the experiments; geometric feedback comes from the laser scanner.

Grey-Box Repair uses a semi-empirical deposition model and constrained inverse command selection to relate material delivery, robot travel, and material age to the requested deposition.

Grey-Box + Residual-Informed Repair adds an offline-trained spatial ridge-residual correction to guide the deposition request. Its coefficients remain frozen during execution. This strategy was implemented and evaluated in physical single-gap and double-gap repairs; it does not use reinforcement learning.

The physical evaluation includes single- and double-gap repair. Repair does not imply complete gap filling or elimination of overfill. Experimental conditions differ, so observed improvements cannot be attributed solely to residual learning.

Research workflow

Robotic material deposition — Deposit the mortar layer.
Laser-based geometric measurement — Measure the deposited geometry.
Material-loss gap localization — Identify the regions requiring repair.
Grey-box and residual-informed repair planning — Select feasible, spatially targeted repair commands.
Same-layer repair execution and geometric verification — Repair the measured gaps and re-scan the layer.