← CHI 2026 map

CHI '26 · Honorable mention · full-paper review · confidence medium-high

WELDAR: Augmenting Live Hands-On Training with In-Situ Guidance for Novice Learners

Chuhan(Franklin) Xu , Lia Sparingga Purnamasari , Zhenfang Chen , Daragh Byrne , Dina EL-Zanfaly

DOI PDF Program page Page JSON

WELDAR is a credible CHI-style systems-and-study paper: the contribution is a live, helmet-integrated AR training system plus a controlled novice study showing measurable gains over video instruction. Its main value is shifting XR training from simulation toward in-situ embodied guidance, though the claims remain bounded to one task, one learner group, and a constrained welding setup.

Axes Lens

Rare contribution shape, typical evidence profile. The point here is not a score. It is to show what kind of claim the paper makes, and whether the evidence pattern is unusual or baseline in this 268 -review set.

Review Summary

WELDAR’s strongest contribution is conceptual as much as technical: it takes a domain where training has traditionally relied on simulation, demonstration, or after-the-fact critique and shows that real-time in-situ guidance can be made workable during the actual physical task. That is a meaningful departure from common practice in XR training, because welding is not a forgiving environment; glare, helmet occlusion, noise, and motor precision all make live guidance difficult. The paper’s novelty is therefore not merely “AR for welding,” but the specific system architecture that integrates a headset into a welding helmet and couples stepwise instruction with feedback on multiple performance measures during live MIG welding. The validation is also solid for CHI standards: a within-subjects study with 24 novices, comparing AR guidance to video instruction across assisted practice and unassisted tests, gives the paper a clear empirical footing. The abstract’s reported improvements in composite performance, especially travel speed and work angle, support the claim that the system helps novices carry embodied knowledge into independent performance. At the same time, the evidence is appropriately bounded. The study is narrow in participant population and task context, and the paper does not support broad claims about long-term skill acquisition, expert-level performance, or general deployment across trades. The discussion is unusually candid about limitations: controller drift, audio-trigger latency, reduced visual resolution through the auto-darkening shield, and the lack of a VR benchmark all constrain how far the results can be generalized. That makes the paper read as a strong, well-scoped systems contribution with credible experimental support, rather than a universal solution for physical skill training. Its significance lies in demonstrating that in-situ AR guidance can augment live hands-on training in a safety-critical craft, while its limitations remind readers that robustness, ergonomics, and transfer remain open questions.

What Changed

Canon before

Prior CHI/XR training work has often centered on simulation, video, or post-hoc feedback rather than live in-situ guidance during the real task. In that common framing, the training system is expected to prepare learners before they enter the physical task, or to evaluate them after the fact, rather than intervene while the novice is actually welding. WELDAR challenges that default by treating the live booth itself as the training site and by making the guidance available at the moment of action, when the learner is still coordinating torch motion, distance, angle, and speed under real sensory constraints.

Departure from common sense

The paper’s core move is to challenge the default assumption that welding instruction should happen before or after the task, or in simulation. Instead, it argues that novices can still benefit from real-time AR guidance while performing live welding, even under harsh conditions such as glare, noise, and helmet constraints. That is a non-obvious claim because the live setting is exactly where many systems would retreat to simpler video or simulator-based instruction, yet the paper shows that in-situ feedback can still be usable and educationally meaningful.

Actual novelty

WeldAR is a live, in-situ AR training system for MIG welding that provides stepwise instruction and real-time feedback on multiple welding parameters during the actual task. The novelty is not just AR visualization, but the coupling of helmet-integrated guidance with embodied motor control support in a real welding booth. The paper also contributes a carefully instrumented workflow for sensing, calibration, and feedback delivery that is tailored to the constraints of welding practice, making the system contribution more than a demo artifact and closer to a reusable training architecture for a demanding physical skill domain.

Evidence

The paper presents a live welding AR system and evaluates it in a within-subjects study with 24 novices. The abstract states that XR training has emphasized simulation, while WeldAR overlays real-time guidance during live welding. It also reports improved performance in assisted practice and unassisted tests, with gains driven by travel speed and work angle. The evidence supports both the system contribution and a bounded empirical claim about novice welding training, while the discussion and appendix add concrete limitations and validation details that keep the claims appropriately scoped.

Limits

Method limits

The study is a within-subjects evaluation with 24 novices, so the evidence is strongest for this participant group and task setting. The paper’s own discussion also notes that the absence of a VR benchmark limits interpretation of whether the embodied live-welding context is essential versus the visual overlay alone. In addition, the reported analyses are tied to a specific curriculum, a specific welding process, and a specific set of performance metrics, so the method supports a focused claim rather than a broad general theory of all physical-skill training.

Deployment limits

The system is tied to live welding hardware and helmet/torch integration, so deployment depends on compatible equipment, safe booth conditions, and reliable sensing/feedback infrastructure. The paper also reports controller drift, audio trigger latency, reduced visual resolution, and torch-attachment constraints that would affect robustness outside the study setup. These constraints mean the system is promising for controlled vocational training environments, but not yet a drop-in solution for every workshop or every welding configuration.

Boundary conditions

Findings are bounded to novice learners, live MIG welding, and the specific AR guidance design tested against video instruction. Performance gains are reported for assisted practice and unassisted tests, but the paper does not establish generality across other trades, other welding processes, or unsupervised real-world deployment. The strongest interpretation is therefore task-specific transfer within the studied curriculum, with the embodied benefits most visible when learners are still acquiring foundational torch control.

Manual check: No manual check required from the provided evidence packet, but the full paper should be consulted if a stricter audit of the reported statistics, effect sizes, implementation details, or the exact scope of the VR-comparison limitation is needed.

Position in field

This sits at the intersection of XR training, embodied skill acquisition, and vocational instruction. It is positioned as a move from simulation-centric training toward in-situ guidance for physically demanding tasks, with welding as a concrete and safety-critical exemplar. In field terms, the paper is a strong CHI systems-and-study contribution that helps shift the conversation from simulated skill rehearsal to live, situated augmentation of practice.

Abstract

Extended Reality (XR) systems for physical skill training have largely emphasized simulation rather than real-time in-situ instruction. We present WeldAR, an Augmented Reality (AR) system with five learning modules that overlays real-time guidance during live welding using a headset integrated into a welding helmet and a torch attachment. We conducted an in-situ within-subjects study with 24 novices, comparing AR guidance to video instruction for live welding across practice and unassisted tests. AR improved performance in both assisted practice and unassisted tests, primarily driven by gains in travel speed and work angle. By offering real-time feedback on four performance measures, AR supported novices in carrying embodied knowledge into independent tasks. Our contributions include: (1) WeldAR for in-situ physical skill training; (2) empirical evidence that AR enhances composite welding performance and key physical skills; and (3) implications for the development of AR systems that support in-situ, embodied skill training in welding and related trades.