Dataset: Beyond User-Reported Experience — Convergent Evidence for Visual Interface Quality Assessment in Teleoperation

# Dataset: Beyond User-Reported Experience — Convergent Evidence for Visual Interface Quality Assessment in Teleoperation **Reference:** Rafiei, S., Brunnström, K., Schenkman, B., Pifferi, G., Djupsjöbacka, A., & Andrén, B. (2026). Beyond User-Reported Experience: Convergent Evidence for Visual Interface Quality Assessment in Teleoperation. *AutomotiveUI Adjunct '26*. https://doi.org/10.1145/3828158.3838255 **License:** CC BY 4.0 — free to share and adapt with attribution. --- ## Study overview A within-subjects laboratory study (n = 24 main-study participants, M1–M24; 2 pilot participants, P1–P2) comparing five visual interface configurations for a teleoperated mobile platform: | Condition | FPV feed | Depth-cue overlay | TPV feed ||-----------|----------|------------------|----------|| FPV | ✓ | | || AFPV | ✓ | ✓ | || FTPV | ✓ | | ✓ || AFTPV | ✓ | ✓ | ✓ || TPV | | | ✓ | Conditions counterbalanced with a Latin-square scheme (24 of 30 orderings). Eye tracking: Tobii Pro Fusion, 120 Hz, calibrated to FPV monitor only. All analyses in the paper use M1–M24; P1–P2 are included here for completeness. --- ## Files All files are at the top level. Set your working directory to this folder before running any script. ### Data files | File | Description ||------|-------------|| `background_questionnaire.xlsx` | Participant demographics (26 rows: P1–P2, M1–M24). Columns: participant ID, age, gender, profession (generalised category), vision, glasses, RC experience, truck/machinery experience, gaming frequency. Timestamps, health flag, and free-text responses removed for privacy. || `recurring_questionnaire.xlsx` | Per-trial questionnaire responses. Row 1: short column names. Row 2: full question text. Rows 3+: data (one row per trial). Columns: participant ID, condition, letters found, distance estimate (cm), depth confidence, presence, controllability, task confidence, overall experience, presence (remote), helpfulness (1–5), depth perception (1–5). Timestamps and free-text comments removed for privacy. || `eye_tracking_raw.zip` | Raw Tobii Pro Fusion gaze data. Structure inside zip: `<PID>/<CONDITION>.csv`. Each CSV: line 1 = recording start time, line 2 = blank, lines 3+ = columns `system_time_stamp` (µs, monotonic), `left_x`, `left_y`, `right_x`, `right_y` (normalised to FPV display [0,1]², y=0 top). Last two lines = recording end time and total duration. Values outside [0,1]² indicate off-screen gaze; `nan` = missing (blink/loss). || `aoi_results.csv` | Derived: per-trial AOI dwell (%), TTFF (s), and sample counts. Generated by `01_process_eyetracking.R`. || `fpv_dwell.csv` | Derived: per-trial on-FPV gaze dwell (%) and trial duration (s). Generated by `01_process_eyetracking.R`. || `helpfulness.csv` | Derived: per-trial helpfulness and depth-perception ratings. Generated by `02_prepare_questionnaire_data.R`. || `task_performance.csv` | Derived: per-trial task performance (proportion of 9 target letters correctly identified). Generated by `02_prepare_questionnaire_data.R`. | ### Figure files | File | Description ||------|-------------|| `gaze_heatmap.pdf` | **Figure 3 in paper.** Pooled on-FPV gaze density for FPV-based conditions (fire palette, black background). Yellow dashed box = AOI. n values = valid gaze samples per condition. || `fig_aoi_dwell.pdf` | **Figure 4 in paper.** AOI dwell (%) per condition with SE bars and individual data points. Significance brackets for planned augmentation contrasts. || `fig_convergence.png` | **Figure 5 in paper.** Per-condition means (±SE) for helpfulness, on-FPV gaze dwell, duration, and task performance. | ### R analysis scripts Run scripts in order (01 → 02 → 03/04/05). Scripts 03–05 can be re-run independently once 01 and 02 have been run once. Set your working directory to this folder before running. | File | Description ||------|-------------|| `01_process_eyetracking.R` | Reads `eye_tracking_raw.zip` and extracts per-trial gaze metrics. Output: `aoi_results.csv`, `fpv_dwell.csv`. Runtime: ~5 minutes. || `02_prepare_questionnaire_data.R` | Reads `recurring_questionnaire.xlsx` and extracts helpfulness and task performance. Output: `helpfulness.csv`, `task_performance.csv`. || `03_statistical_analysis.R` | Reproduces all statistical results in the paper (Friedman tests, Wilcoxon contrasts, correlations). Prints to console. || `04_generate_figures.R` | Reproduces `fig_aoi_dwell.pdf` and `fig_convergence.png`. || `05_generate_heatmap.R` | Reproduces `gaze_heatmap.pdf` (fire-palette gaze density heatmap). Saves to a path set by `OUT_DIR` in the script — change this to a folder without special characters if your working directory path contains non-ASCII characters (e.g. Swedish å, ö). | **Requirements:** R ≥ 4.3. Packages: `dplyr`, `readxl`, `MASS`. Install with:```rinstall.packages(c("dplyr", "readxl", "MASS"))``` --- ## AOI definition The area of interest (AOI) marks the region of the FPV display containing the depth-cue overlay (green rectangle + red trapezoid in the lower portion of the image). In normalised FPV display coordinates (y = 0 at top): - x ∈ [0.20, 0.80]- y ∈ [0.45, 0.80] --- ## Ethics and privacy All participants provided written informed consent. Data have been anonymised: no names or contact details are included. Participant IDs are arbitrary codes (M1–M24, P1–P2). Timestamps, free-text responses, and one self-reported health flag have been removed. Professions have been generalised to broad categories. --- ## Contact Corresponding author: Kjell Brunnström ([e-mail]), RISE Research Institutes of Sweden.