Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Force-Driven Validation for Collaborative Robotics in Automated Avionics Testing

Pietro Dardano, Paolo Rocco, David Frisini

TL;DR

The paper tackles the challenge of safe, repeatable avionics testing by combining cobot automation with DL/XAI to validate cockpit interactions using wrench data. It proposes a preprocessing pipeline to filter disturbances, transforms force/torque signals into 2D scaleograms via Continuous Wavelet Transform, and trains four architectures (FF_ANN, 1D-CNN, 2D-CNN, Hybrid-CNN) to classify actions as Success or Fail and identify failure causes, with Grad-CAM providing interpretability. Results show 1D-CNNs achieve the best performance across most actions, with 97–99% F1-scores on several targets, while 2D and Hybrid models offer competitive performance at higher computational cost; knob interaction remains challenging due to data imbalance. The work demonstrates feasibility of a wrench-based validation framework integrated with ROS2, enabling real-time validation and improved reliability for automated avionics testing, and outlines directions for broader generalization and CV-based redundancy.

Abstract

ARTO is a project combining collaborative robots (cobots) and Artificial Intelligence (AI) to automate functional test procedures for civilian and military aircraft certification. This paper proposes a Deep Learning (DL) and eXplainable AI (XAI) approach, equipping ARTO with interaction analysis capabilities to verify and validate the operations on cockpit components. During these interactions, forces, torques, and end effector poses are recorded and preprocessed to filter disturbances caused by low performance force controllers and embedded Force Torque Sensors (FTS). Convolutional Neural Networks (CNNs) then classify the cobot actions as Success or Fail, while also identifying and reporting the causes of failure. To improve interpretability, Grad CAM, an XAI technique for visual explanations, is integrated to provide insights into the models decision making process. This approach enhances the reliability and trustworthiness of the automated testing system, facilitating the diagnosis and rectification of errors that may arise during testing.

Force-Driven Validation for Collaborative Robotics in Automated Avionics Testing

TL;DR

The paper tackles the challenge of safe, repeatable avionics testing by combining cobot automation with DL/XAI to validate cockpit interactions using wrench data. It proposes a preprocessing pipeline to filter disturbances, transforms force/torque signals into 2D scaleograms via Continuous Wavelet Transform, and trains four architectures (FF_ANN, 1D-CNN, 2D-CNN, Hybrid-CNN) to classify actions as Success or Fail and identify failure causes, with Grad-CAM providing interpretability. Results show 1D-CNNs achieve the best performance across most actions, with 97–99% F1-scores on several targets, while 2D and Hybrid models offer competitive performance at higher computational cost; knob interaction remains challenging due to data imbalance. The work demonstrates feasibility of a wrench-based validation framework integrated with ROS2, enabling real-time validation and improved reliability for automated avionics testing, and outlines directions for broader generalization and CV-based redundancy.

Abstract

ARTO is a project combining collaborative robots (cobots) and Artificial Intelligence (AI) to automate functional test procedures for civilian and military aircraft certification. This paper proposes a Deep Learning (DL) and eXplainable AI (XAI) approach, equipping ARTO with interaction analysis capabilities to verify and validate the operations on cockpit components. During these interactions, forces, torques, and end effector poses are recorded and preprocessed to filter disturbances caused by low performance force controllers and embedded Force Torque Sensors (FTS). Convolutional Neural Networks (CNNs) then classify the cobot actions as Success or Fail, while also identifying and reporting the causes of failure. To improve interpretability, Grad CAM, an XAI technique for visual explanations, is integrated to provide insights into the models decision making process. This approach enhances the reliability and trustworthiness of the automated testing system, facilitating the diagnosis and rectification of errors that may arise during testing.
Paper Structure (10 sections, 2 equations, 7 figures, 3 tables, 1 algorithm)

This paper contains 10 sections, 2 equations, 7 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Automated Testing in Avionics
  4. Wrench analysis and Success Detection
  5. Setup and Data Preprocessing
  6. Classification Task
  7. Experiments and Results
  8. 1D Convolutional Neural Networks
  9. 2D and Hybrid Convolutional Neural Networks
  10. Conclusion and Discussion

Figures (7)

  • Figure 1: ARTO setup
  • Figure 2: Button pressing
  • Figure 3: FLAP actuation
  • Figure 4: (a) Measured force feedback during multiple button pressing actions, highlighting oscillations and variability across trials. (b) Signal segmentation using the developed algorithm, which isolates the most informative transient phase by detecting key energy-based thresholds and applying a fixed-length window.
  • Figure 5: Comparative analysis of 2D and 1D Grad-CAM for feature attribution. (a) The 2D variant maps activation on the wavelet-transformed signals. (b) The 1D approach preserves temporal information, offering a clearer interpretation of model focus on raw force signals.
  • ...and 2 more figures