Optimization-Based Path Planning for Industrial Manipulators: Experimental Validation in Additive Manufacturing

TL;DR An offline planner optimizes a UR10e's whole joint path to meet process tolerances in carbon-fiber winding and 3D printing, and an iterative scheme sets its cost weights from KPI gaps. The planned winding path's worst fiber-alignment error is 1.90°, under a 2° limit, and the 479,523-pose printing path stays within 0.066 mm of the reference.
What they do
In robotic additive manufacturing the process dictates how the tool or workpiece moves. The robot still needs a joint path that respects joint limits, self-clearance and smoothness. The authors build one offline planner for two processes. In carbon-fiber winding, a UR10e carries a grooved core past a fixed fiber feeder. In 3D printing, it carries an extrusion head along prescribed nozzle poses. Each process supplies only its reference, residuals and tolerances, and Iterative Weight Tuning (IWT) picks the cost weights.
How it works
- Joint-path optimization. All configurations along the reference are optimized together. Costs penalize first- to third-order joint differences and self-clearance violations, joint bounds are hard constraints, and the process cost is evaluated through forward kinematics. The solver is JAXIPM on an H100 GPU.
- Winding. A first stage orders the core’s vertices with Hierholzer’s algorithm and optimizes fiber directions for each groove transition. The robot stage then aligns the free fiber with that reference, keeps it clear of the links and bounds the groove-end error to 2°.
- Printing. The 479,523-pose reference is optimized in chunks of at most 90,000. Boundary penalties couple each chunk to its fixed neighbors, and forward and reverse sweeps alternate.
- IWT. Weights are updated in log space from empirical sensitivities of the KPI gaps, without differentiating through the optimizer.
Results
- Winding (2,401 samples over 20 edges): fiber-direction error at groove ends is 1.236° RMS and 1.900° max, under the 2° tolerance. The largest joint step between samples is 2.28°.
- Winding IWT meets all four KPI targets (three clearances and fiber error) at repetition 5. Repetition 1 met two of them.
- Printing: IWT meets its three KPI targets after 3 repetitions. Nominal TCP error is 4.38×10⁻⁴ mm RMS and 0.066 mm max; orientation error peaks at 1.32×10⁻³°.
- Both paths were executed on a physical UR10e with joint tracking logged.
Limitations
Accuracy is reported for the planned path (nominal, from forward kinematics), not measured on placed fiber or printed parts; the authors name that as the next step. Printing ran without extrusion. The planner assumes a fixed feeder position, base registration and environment geometry, so any change means solving again. In the winding IWT study, four weights were tied together and two held fixed.
Figures




Abstract
Robotic additive manufacturing requires joint paths that reconcile robot feasibility with process-specific geometry. We develop an offline optimization-based path planner that combines joint-path regularity and self-collision penalties with process costs and constraints, so that a new process enters through its own reference, residuals, and tolerances. Full-path and boundary-coupled chunked optimization handle references of different lengths. Iterative Weight Tuning (IWT) determines the cost weights from the gaps between Key Performance Indicators (KPIs) and their targets. In carbon-fiber winding and robotic 3D printing, the planned paths meet the geometric tolerances and all KPI targets, and both paths are executed on a UR10e, without extrusion for printing, with joint-level tracking recorded during execution.
Summaries are written automatically from each paper's text. Figures and videos come from the paper and its project page; all credit goes to the authors.
