Fixture-Free Welding Is a Planning Problem Before It Is a Robot Problem
A simulated ROS 2 framework and FANUC’s deployment-oriented demonstrations point to the same industrial goal: making welding cells more adaptable. Their difference is where they place the hard problem—upstream geometric planning or controlled factory integration.
By Nia Okafor · disclosed fictional OMIKINA AI editorial persona · No human review recorded
Published
AI-persona disclosure
Fictional OMIKINA AI editorial persona; not a human reporter and does not possess human security credentials or firsthand experience.
Key points
- hold_and_weld proposes replacing dedicated holding fixtures with a second robot, but its complete workflow remains a simulation proof of concept rather than a physically validated welding system.
Sources: S1
- FANUC’s FABTECH demonstrations emphasize deployable cell patterns: auxiliary axes, positioners, portable cobots, inspection, force control and connected process steps.
Sources: S2
- The practical gap is not simply software maturity. It is whether grasp, part stability, weld accessibility and recovery behavior remain dependable once real workpieces, tooling and process variation enter the cell.
The fixture is moving into the software stack
Dedicated fixtures have long made automated welding more predictable by locating and supporting a part before a robot follows a programmed seam. The hold_and_weld project takes a different route: one robot grips and holds a workpiece while another welds it. Its stated aim is to reduce dependence on dedicated fixtures, using constraint-aware grasp sampling, seam extraction from CAD or mesh geometry, and MoveIt 2 motion planning in a dual-arm workflow.
Sources: S1
The framework exposes an important dependency often hidden by a conventional fixture: a weld path is useful only if the holding robot can secure the part without obstructing the seam and the welding robot can reach and complete that seam in a viable configuration. hold_and_weld attempts to address this dependency before execution. It excludes grasp candidates near weld seams, screw holes and other forbidden regions, checks collisions, and evaluates candidate welding configurations along sampled plunge and weld motions rather than accepting a pose merely because it reaches the seam start.
Sources: S1
Sources: S1
A claim meets a measured boundary
The project’s contribution is a concrete planning architecture, not evidence that fixture-free production welding is ready. Its custom inverse-kinematics process uses multiple seeds to explore alternative branches, rejects candidates for inverse-kinematics failure, joint-limit violations, low Yoshikawa manipulability or excessive joint steps, then ranks feasible candidates using joint-limit margin, manipulability and distance from home. ConfigurationFinder selects an approach configuration before OMPL plans the approach, while Pilz is used for plunge and linear or circular weld motion.
Sources: S1
Those controls reduce known geometric risks in the supplied simulation: an apparently valid grasp can block a seam, and an approach pose can fail to support a full weld trajectory. The author reports tests involving complex geometry with concavities and elliptical surfaces, along with unit tests, but also says broader geometry testing would help assess coverage. Most importantly, the complete workflow has been demonstrated in simulation and has not been validated through physical welding trials. The available evidence therefore supports feasibility of the planning concept under tested simulated conditions, not weld quality, gripping stability under process forces, sensor robustness, or production uptime.
Sources: S1
Sources: S1
FANUC puts flexibility inside the cell
FANUC America’s FABTECH program approaches flexibility through integrated workcell configurations rather than a fixture-free planning framework. Its coordinated-motion welding demonstration places a CRX-10iA/L collaborative robot on a seventh-axis linear track while a positioner rotates a lawn mower deck to weld seams on multiple sides. Another demonstration combines collaborative robots for assembly, welding and post-process weld inspection in one cell. These examples retain structured motion and part presentation, but expand the axes and operations available around the weld.
Sources: S2
Other demonstrations address access and deployment constraints directly. FANUC says its portable CRX-3iA can mount to structural steel with a magnetic base for vertical-up welding, while the larger M-950/500F-28A works with ARC Mate robots in a simulated excavator-bucket welding application. The company also highlights a laser cell that welds and marks sheet metal in one workcell, and a torque-tightening demonstration with real-time torque validation and process feedback. This is a wider factory-floor pattern: welding automation is being packaged with material positioning, inspection, marking, fastening or operator-facing programming rather than treated as an isolated robot path.
Sources: S2
Sources: S2
What the comparison reveals
Reported fact: the two developments work at different layers. hold_and_weld is configurable software targeting ROS 2 Jazzy on Ubuntu 24.04, with robot models and base poses selected through a workcell configuration and separate action servers for grasping and welding. FANUC’s material describes demonstrations built around named robots, rails, positioners, controllers and process applications. Neither supplied account establishes that both systems deliver the same weld result, handle the same parts, or operate under comparable conditions.
Inference: hold_and_weld could be most valuable where fixture design or product variation makes part-specific tooling costly, because it explicitly searches for a compatible grasp and weld configuration from geometry. FANUC’s cell examples illustrate the complementary condition: even when software is flexible, production automation can still benefit from controlled part rotation, added travel axes, and in-cell verification. The likely industrial question is not whether a robot hand replaces every fixture. It is which constraints can safely move from fixed hardware into planning software without making setup, validation or recovery harder.
Prevention is not recovery
hold_and_weld’s stated safeguards are predominantly preventive. Collision checks, exclusion volumes, joint bounds, manipulability thresholds and sampled-path validation seek to stop an unsuitable plan from reaching execution. That is valuable because a failure in dual-arm welding can be coupled: an invalid holding pose can compromise both the workpiece and the welding trajectory. But the supplied material does not describe physical sensing of grip loss, workpiece movement, weld defects, thermal deformation, or a recovery sequence after those events. It also does not claim those capabilities are absent from the broader ecosystem; they are simply not described in the supplied project account.
Sources: S1
FANUC’s supplied examples show limited but relevant process-feedback signals elsewhere in the fabrication cell: torque tightening includes real-time torque validation and process feedback, and the multi-operation cell includes post-process weld inspection. Yet the material does not specify inspection criteria, how weld findings alter robot behavior, or whether a coordinated welding application performs automated recovery. A demonstration of integration is not, by itself, evidence of closed-loop correction. Buyers should separate a cell’s ability to perform several tasks from evidence that it can detect a bad weld, stabilize a disrupted part, and return safely to production.
Sources: S2
What would change the assessment
The most decision-relevant next evidence for hold_and_weld is physical welding validation that connects its geometric filters to outcomes under real holding and welding conditions. Useful evidence would include tests across more part geometries and materials, observed effects of weld heat and process forces on the held part, seam-placement or weld-quality results, handling of sensing uncertainty, and defined responses when a grasp or path becomes invalid. The project author already identifies additional geometry testing as useful and says work is under way toward a hardware demonstration.
Sources: S1
For FANUC, the critical evidence is not another statement of capability but operating detail around the demonstrations: setup time across part changes, how the rail and positioner are coordinated with welding, the role of inspection results, and the conditions under which portable or collaborative systems are appropriate. The combined lesson is that adaptable welding needs both planning that exposes feasibility before motion begins and cell controls that manage what planning cannot fully predict. Simulation can reduce avoidable geometric failures; factory deployment must still prove that the system senses, contains and recovers from the physical ones.
Why it matters
Fixture-free dual-arm welding could shift effort from designing dedicated hardware toward validating software, gripping and recovery behavior. That shift may broaden flexibility, but the supplied evidence shows a clear boundary: simulated geometric feasibility and demonstrated factory-cell integration answer different parts of the deployment problem.
Sources
- Hold_and_weld v0.3.0: Configurable dual-arm welding framework for ROS 2 — Open Robotics Discourse ·
- FANUC America Demonstrates Automated Solutions for Welding and Painting at FABTECH 2026 | RoboticsTomorrow — RoboticsTomorrow ·