Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Human-AI Interaction in Industrial Robotics: Design and Empirical Evaluation of a User Interface for Explainable AI-Based Robot Program Optimization

Benjamin Alt, Johannes Zahn, Claudius Kienle, Julia Dvorak, Marvin May, Darko Katic, Rainer Jäkel, Tobias Kopp, Michael Beetz, Gisela Lanza

TL;DR

The paper tackles the adoption barrier of deep-learning–driven robot program optimization in manufacturing by introducing an Explanation User Interface (XUI) that embeds Explainable AI (XAI) features. The core method, spi, trains a differentiable shadow model $\hat{P}$ of the parameterized robot program $P$ from data $(x,\Theta)$ and performs model-based iterative optimization over the input parameters $x$ to maximize a task objective, guided by a three-step workflow (dataset definition, model training, program optimization). It contributes a design framework centered on user adaptability (Guided vs Expert modes) and explainability across the workflow, including LRP-based explanations, plus a concrete UI implementation within ArtiMinds lar and a preliminary user study demonstrating feasibility for both AI novices and experts, with a plan for a larger, controlled study. The preliminary findings indicate the XUI supports practical robot-program optimization with acceptable cognitive load and perceived usefulness, though novices benefit from more guidance and deeper explainability remains a challenge; the authors propose a rigorous large-scale study to derive generalizable guidelines for trustworthy human-AI interaction in industrial robotics. Overall, the work presents a practical pathway for deploying explainable, user-centered AI tooling to reduce the skill gap and enable safer, more transparent robot programming in manufacturing environments.

Abstract

While recent advances in deep learning have demonstrated its transformative potential, its adoption for real-world manufacturing applications remains limited. We present an Explanation User Interface (XUI) for a state-of-the-art deep learning-based robot program optimizer which provides both naive and expert users with different user experiences depending on their skill level, as well as Explainable AI (XAI) features to facilitate the application of deep learning methods in real-world applications. To evaluate the impact of the XUI on task performance, user satisfaction and cognitive load, we present the results of a preliminary user survey and propose a study design for a large-scale follow-up study.

Human-AI Interaction in Industrial Robotics: Design and Empirical Evaluation of a User Interface for Explainable AI-Based Robot Program Optimization

TL;DR

The paper tackles the adoption barrier of deep-learning–driven robot program optimization in manufacturing by introducing an Explanation User Interface (XUI) that embeds Explainable AI (XAI) features. The core method, spi, trains a differentiable shadow model of the parameterized robot program from data and performs model-based iterative optimization over the input parameters to maximize a task objective, guided by a three-step workflow (dataset definition, model training, program optimization). It contributes a design framework centered on user adaptability (Guided vs Expert modes) and explainability across the workflow, including LRP-based explanations, plus a concrete UI implementation within ArtiMinds lar and a preliminary user study demonstrating feasibility for both AI novices and experts, with a plan for a larger, controlled study. The preliminary findings indicate the XUI supports practical robot-program optimization with acceptable cognitive load and perceived usefulness, though novices benefit from more guidance and deeper explainability remains a challenge; the authors propose a rigorous large-scale study to derive generalizable guidelines for trustworthy human-AI interaction in industrial robotics. Overall, the work presents a practical pathway for deploying explainable, user-centered AI tooling to reduce the skill gap and enable safer, more transparent robot programming in manufacturing environments.

Abstract

While recent advances in deep learning have demonstrated its transformative potential, its adoption for real-world manufacturing applications remains limited. We present an Explanation User Interface (XUI) for a state-of-the-art deep learning-based robot program optimizer which provides both naive and expert users with different user experiences depending on their skill level, as well as Explainable AI (XAI) features to facilitate the application of deep learning methods in real-world applications. To evaluate the impact of the XUI on task performance, user satisfaction and cognitive load, we present the results of a preliminary user survey and propose a study design for a large-scale follow-up study.
Paper Structure (17 sections, 5 figures)

This paper contains 17 sections, 5 figures.

Table of Contents

  1. Introduction
  2. An Explanation User Interface for AI-based Robot Program Optimization
  3. AI-based Robot Program Optimization
  4. Guiding Principles for UI Design
  5. User Adaptability
  6. Explainability
  7. XUI Workflow
  8. Dataset definition
  9. Model training
  10. Program optimization
  11. Preliminary User Study
  12. Study Design
  13. Results
  14. Discussion
  15. Large-Scale User Study
  16. ...and 2 more sections

Figures (5)

  • Figure 1: Overview of the proposed system: A user interface enables intuitive interaction of a human user with an ai system for robot program optimization.
  • Figure 2: Workflow and corresponding ui elements, with variation points for user adaptability () and explainability features (). Screenshots for data visualization (left), lrp (center left, bars illustrate the relative impact of individual program parameters on model output), hyperparameter specification (center right), specification of the optimization objective (2nd from right) and visualization of optimization results (far right) are shown.
  • Figure 3: ui for the optimization step, Guided (left) and Expert modes (right).
  • Figure 4: Experiment set-up for robotic gearbox assembly.
  • Figure 5: Survey results of 12 participants, 8 of which are classified as ai novices and 4 as ai experts. $\blacktriangle$ indicates the median response.