Click, Don’t Steer: A Quantitative Comparison of Tele-Driving and Tele-Assistance User Interfaces for Remote Operation of Autonomous Vehicles

Paper Title

Click, Don’t Steer: A Quantitative Comparison of Tele-Driving and Tele-Assistance User Interfaces for Remote Operation of Autonomous Vehicles

Publication Info

• Topic area: Comparative evaluation of teleoperation paradigms for autonomous vehicle remote control.
• Keywords: Teleoperation, tele-driving, tele-assistance, autonomous vehicles, mental workload, situation awareness, user experience, remote operation, human-machine interaction, cognitive offloading.

Background and Problem

• Problem / challenge: Autonomous vehicles (AVs) struggle with edge-case scenarios like heavy traffic, poor visibility, and construction zones, necessitating human intervention via teleoperation. Existing paradigms, tele-driving and tele-assistance, lack systematic comparison to determine their relative merits.
• Significance: Understanding the trade-offs between tele-driving and tele-assistance is critical for designing scalable, efficient, and safe teleoperation systems, which are essential for the deployment of AVs in real-world environments.
• Motivation and related work: Prior research has focused on teleoperation interfaces, latency challenges, and interaction techniques, but no comprehensive comparison of tele-driving and tele-assistance has been conducted. This study addresses this gap by empirically evaluating both paradigms.

Solution

• Proposed approach: A controlled laboratory experiment comparing tele-driving (continuous control via steering wheel and pedals) and tele-assistance (high-level commands via a graphical interface) across four edge-case scenarios.
• Novelty:
  1. First systematic, quantitative comparison of tele-driving and tele-assistance paradigms.
  2. Empirical evidence demonstrating cognitive offloading benefits of tele-assistance.
  3. Analysis of design trade-offs and implications for future teleoperation systems.
• Procedure and key techniques:
  • Participants (N=40) were randomly assigned to tele-driving or tele-assistance conditions.
  • Four scenarios (e.g., busy junction, road construction) were simulated using the CARLA simulator.
  • Metrics included mental workload (NASA-TLX, pupil diameter), situation awareness (SAGAT), task completion time, and user experience (UEQ).
  • A Wizard-of-Oz (WoZ) approach was used to simulate the AV’s execution of commands in tele-assistance.

Results

• Concrete findings:
  • Tele-assistance significantly reduced mental workload (NASA-TLX: U = 5.09, p = 0.024; pupil diameter: trend toward lower values).
  • Situation awareness was higher in tele-assistance (F(1, 38) = 5.48, p = 0.025).
  • Tele-driving enabled shorter task completion times (F(1, 38) = 88.60, p < 0.001).
  • User experience ratings were broadly positive for both paradigms, with tele-assistance rated higher for perspicuity (p = 0.0138).
• Advantage over baselines:
  • Tele-assistance reduced cognitive load and improved situational awareness compared to tele-driving.
  • Tele-driving achieved faster task completion but at the cost of higher workload and lower situational awareness.
• Experiments / evaluation:
  • Four scenarios were tested twice per participant, with randomized order to minimize learning effects.
  • Dependent variables included physiological (pupil diameter), subjective (NASA-TLX, UEQ), and performance-based (completion time, SAGAT) measures.
• Limitations and future work:
  • Participants were not professional teleoperators, and the study was conducted in a controlled simulator environment.
  • The WoZ approach may not fully reflect real-world AV execution, limiting generalizability of completion time results.
  • Future research should explore real-world systems, multi-vehicle supervision, and hybrid teleoperation paradigms.

Summary

This study provides the first systematic comparison of tele-driving and tele-assistance for remote operation of AVs. Tele-assistance reduced mental workload and improved situational awareness, making it suitable for sustained teleoperation tasks, while tele-driving enabled faster task completion, favoring dynamic scenarios requiring direct control. These findings highlight the complementary strengths of both paradigms and suggest potential for hybrid systems. Future work should address scalability, real-world deployment, and operator training to optimize teleoperation systems for AV fleets.