pith. sign in

arxiv: 2605.15952 · v1 · pith:2Y7CLL46new · submitted 2026-05-15 · 💻 cs.HC · cs.RO

Driving Through the Network: Performance and Workload Under Latency and Video Impairment

Ines Trautmannsheimer , Ahmed Azab , Frank Diermeyer This is my paper

Pith reviewed 2026-05-20 16:31 UTC · model grok-4.3

classification 💻 cs.HC cs.RO
keywords teleoperationnetwork latencyvideo bitrateworkloaddriving performanceautomated vehiclessimulator study
0
0 comments X

The pith

Latency of 300 ms with 2000 kbit/s video keeps remote driving velocity equivalent to best-case conditions.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper investigates the effects of network latency and video bitrate on human operators teleoperating automated vehicles using a fixed-base driving simulator. It finds that both increased latency and reduced bitrate raise operator workload and modestly reduce performance measures such as speed and increase steering reversals. Equivalence tests indicate that a combination of 300 ms latency and 2000 kbit/s bitrate produces speeds statistically equivalent to the no-impairment baseline within a 2 km/h smallest effect size of interest, whereas 500 kbit/s does not. The authors conclude that latency and video quality function as largely independent design factors and recommend physiology-aware system adaptations to prevent overload.

Core claim

Through a 2x2 study of added latency and bitrate in a simulator, the authors show that latency and bitrate each increase operator load with modest performance effects. Equivalence testing establishes that 300 ms latency at 2000 kbit/s maintains velocity equivalence to the best-case condition, supporting the view that these impairments can be managed independently in teleoperation design.

What carries the argument

The 2x2 experimental manipulation of added latency (100 or 300 ms) and bitrate (500 or 2000 kbit/s) against a best-case baseline, combined with equivalence testing on velocity measures.

If this is right

  • Operators can sustain near-optimal driving speeds under moderate latency if video bitrate remains high.
  • System designers can adjust latency and video quality somewhat independently without compounding performance losses.
  • Physiological signals like heart rate may allow early detection of increasing workload.
  • Teleoperation systems could adapt network parameters dynamically to maintain performance.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Real-world teleoperation might benefit from bitrate prioritization over latency minimization in bandwidth-constrained networks.
  • Extending the study to include actual vehicle dynamics could reveal additional effects not captured in the fixed-base simulator.
  • These findings could guide the development of quality-of-service standards for remote vehicle control over 5G networks.

Load-bearing premise

The fixed-base driving simulator accurately captures the real-world challenges of teleoperating automated vehicles under network impairments.

What would settle it

Observing in a real teleoperated vehicle trial that average speeds under 300 ms latency and 2000 kbit/s differ by more than 2 km/h from best-case conditions.

Figures

Figures reproduced from arXiv: 2605.15952 by Ahmed Azab, Frank Diermeyer, Ines Trautmannsheimer.

Figure 1
Figure 1. Figure 1: Overview of representative teleoperation scenar [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Pictures of the nine scenarios, as perceived by the [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the route geometry and shows, using [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The experimental setup incorporated a worksta [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Paired comparison for Velocity between Best [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Difference vs. Best-Case for Velocity (Cell [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
read the original abstract

Teleoperation promises to extend the operational envelope of automated vehicles, yet it critically depends on network latency and video quality. We report a fixed-base driving-simulator study (N=25) with a 2x2 manipulation of added latency (100/300 ms) and bitrate (500/2000 kbit/s), plus a best-case baseline (0 ms added, 9000 kbit/s). We measured effective glass-to-glass (G2G) latency per condition (baseline approx. 413 ms; effective totals approx. 500-700 ms) and verified stable framerate and encoder settings. Multimodal measures covered performance (speed, steering reversals, crashes), oculomotor behavior (blink rate, fixation duration), physiology (RR interval, heart rate, skin conductance), and subjective workload. Latency and bitrate each increased operator load and modestly affected performance. Physiological measures (heart rate, RR interval) exhibited sub-additive interactions, whereas performance and oculomotor interactions were small or non-significant. Equivalence tests showed that 300 ms with 2000 kbit/s was velocity-equivalent to best-case (SESOI +/- 2 km/h), while 300 ms with 500 kbit/s was not. We argue that latency and video quality should be treated as largely independent design levers, and that physiology-aware adaptation can anticipate overload before safety is compromised.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript reports findings from a fixed-base driving simulator study involving 25 participants. It employs a 2x2 design manipulating added latency (100 ms vs. 300 ms) and video bitrate (500 kbit/s vs. 2000 kbit/s), alongside a best-case baseline. Dependent variables encompass performance metrics (speed, steering reversals, crashes), oculomotor measures (blink rate, fixation duration), physiological signals (RR interval, heart rate, skin conductance), and subjective workload. The authors conclude that latency and bitrate act as largely independent levers increasing operator load, with physiological interactions being sub-additive, and that the 300 ms latency combined with 2000 kbit/s condition demonstrates velocity equivalence to baseline within ±2 km/h.

Significance. The study offers valuable insights into the human factors of vehicle teleoperation under network impairments. By demonstrating largely independent effects of latency and bitrate, it supports modular design approaches for remote driving systems. The use of equivalence testing provides practical guidance on tolerable impairment levels. Multimodal data collection, including physiology, highlights potential for early detection of overload. These contributions are relevant to the development of safe and efficient teleoperation interfaces for automated vehicles, provided the simulator-based findings translate to real-world conditions.

major comments (2)
  1. [Methods] The fixed-base simulator lacks vehicle motion cues, haptic feedback, and variable environmental factors inherent to real teleoperation. This is a load-bearing assumption for extrapolating the independence of latency and bitrate effects and the equivalence results to practical teleoperation design, yet the manuscript provides no cross-validation with motion-platform simulators or on-road data.
  2. [Results] In the equivalence tests, the SESOI of ±2 km/h for velocity is used to conclude equivalence for the 300 ms + 2000 kbit/s condition. However, the justification for this specific threshold and any sensitivity analysis to alternative SESOI values are not reported, which could affect the robustness of the claim that this condition is acceptable while the lower bitrate is not.
minor comments (2)
  1. [Abstract] The approximate effective G2G latencies (500-700 ms) and verification of stable framerate should be detailed with precise measurements and statistical tests in the main body to enhance reproducibility.
  2. [Discussion] The suggestion for physiology-aware adaptation could reference existing literature on real-time workload monitoring in HCI to strengthen the forward-looking implications.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their constructive review and valuable feedback on our manuscript. We have addressed each major comment below, making revisions to the manuscript where feasible to strengthen the presentation of our findings.

read point-by-point responses

  1. Referee: [Methods] The fixed-base simulator lacks vehicle motion cues, haptic feedback, and variable environmental factors inherent to real teleoperation. This is a load-bearing assumption for extrapolating the independence of latency and bitrate effects and the equivalence results to practical teleoperation design, yet the manuscript provides no cross-validation with motion-platform simulators or on-road data.

    Authors: We agree that the fixed-base simulator design limits ecological validity due to the absence of motion cues, haptic feedback, and dynamic environmental factors. In the revised manuscript, we have substantially expanded the Limitations subsection within the Discussion to explicitly address how these factors could influence the observed independence of latency and bitrate effects as well as the equivalence results. We emphasize that the controlled laboratory setting was intentionally chosen to isolate network impairment variables, enabling clear causal attribution that would be difficult in less controlled real-world conditions. While we cannot provide cross-validation data in this revision, we have highlighted the need for future studies using motion platforms or on-road teleoperation as a key direction for extending these findings. revision: partial

  2. Referee: [Results] In the equivalence tests, the SESOI of ±2 km/h for velocity is used to conclude equivalence for the 300 ms + 2000 kbit/s condition. However, the justification for this specific threshold and any sensitivity analysis to alternative SESOI values are not reported, which could affect the robustness of the claim that this condition is acceptable while the lower bitrate is not.

    Authors: We thank the referee for identifying this gap. In the revised Results section, we now provide a clear justification for the SESOI of ±2 km/h, based on the standard deviation of velocity in the baseline condition and practical considerations for maintaining safe and consistent teleoperated driving speeds. We have also added a sensitivity analysis testing equivalence at alternative SESOI values of ±1 km/h and ±3 km/h. The equivalence conclusion for the 300 ms latency with 2000 kbit/s condition remains robust across these thresholds, while the 500 kbit/s condition does not meet equivalence criteria. These updates are reported with the corresponding statistical details. revision: yes

standing simulated objections not resolved
  • Cross-validation of results with motion-platform simulators or on-road teleoperation data, which would require new empirical studies outside the scope of the current work.

Circularity Check

0 steps flagged

Empirical measurements from simulator study contain no circular derivations

full rationale

This paper reports an empirical fixed-base simulator experiment (N=25) with direct 2x2 manipulations of added latency and bitrate plus baseline, followed by multimodal measurements of speed, steering, crashes, blink rate, fixation, RR interval, heart rate, skin conductance, and workload. All central claims (latency and bitrate effects on load, sub-additive physiological interactions, velocity equivalence via equivalence tests with SESOI of ±2 km/h) rest on observed data and standard statistical tests rather than any equations, fitted parameters, self-citations, or ansatzes that reduce outputs to inputs by construction. No mathematical derivation chain exists to inspect for circularity.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Empirical study with few free parameters beyond the equivalence test threshold; relies on domain assumptions about simulator fidelity.

free parameters (1)
  • SESOI for velocity equivalence = +/- 2 km/h
    Smallest effect size of interest used in equivalence testing for velocity.
axioms (1)
  • domain assumption Simulator-based teleoperation performance generalizes to real-world conditions.
    The study relies on a fixed-base simulator to draw conclusions about teleoperation.

pith-pipeline@v0.9.0 · 5785 in / 1266 out tokens · 86992 ms · 2026-05-20T16:31:58.929260+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

27 extracted references · 27 canonical work pages

  1. [1]

    Brecht, D., Gehrke, N., Kerbl, T., Krauss, N., Majstorovi \'c , D., Pfab, F., Wolf, M.-M., and Diermeyer, F. (2024). Evaluation of teleoperation concepts to solve automated vehicle disengagements. IEEE Open Journal of Intelligent Transportation Systems

    work page 2024

  2. [2]

    CARLA Simulator: Python API Reference

    CARLA Simulator Team (2025). CARLA Simulator: Python API Reference . CARLA. Accessed: 14 July 2025

    work page 2025

  3. [3]

    Committee, O.-R. A. D. O. (2021). Taxonomy and definitions for terms related to driving automation systems for on-road motor vehicles . SAE international

    work page 2021

  4. [4]

    Dosovitskiy, A., Ros, G., Codevilla, F., Lopez, A., and Koltun, V. (2017). CARLA : An open urban driving simulator. In Proceedings of the 1st Annual Conference on Robot Learning , pages 1--16

    work page 2017

  5. [5]

    Unreal engine

    Epic Games . Unreal engine

    work page

  6. [6]

    Gstreamer

    GStreamer Project . Gstreamer. Accessed September 2025

    work page 2025

  7. [7]

    Hart, S. G. and Staveland, L. E. (1988). Development of NASA-TLX ( Task Load Index ): Results of Empirical and Theoretical Research . In Hancock, P. A. and Meshkati, N., editors, Advances in Psychology , volume 52 of Human Mental Workload , pages 139--183. North-Holland

    work page 1988

  8. [8]

    Hoffmann, S., Willert, F., Hofbauer, M., Schimpe, A., and Diermeyer, F. (2022). Quantifying the influence of image quality on operator reaction times for teleoperated road vehicles. In 13th International Conference on Applied Human Factors and Ergonomics (AHFE 2022)

    work page 2022

  9. [9]

    Advanced video coding for generic audiovisual services (h.264)

    ITU-T (2019). Advanced video coding for generic audiovisual services (h.264). Technical Report Recommendation H.264, International Telecommunication Union. Version approved February 2019

    work page 2019

  10. [10]

    B., Lu, Q., Bouali, F., Haas, O

    Kamtam, S. B., Lu, Q., Bouali, F., Haas, O. C., and Birrell, S. (2024). Network latency in teleoperation of connected and autonomous vehicles: A review of trends, challenges, and mitigation strategies. Sensors (Basel, Switzerland) , 24(12):3957

    work page 2024

  11. [11]

    Kerbl, T., Brecht, D., Gehrke, N., Karunainayagam, N., Krauss, N., Pfab, F., Taupitz, R., Trautmannsheimer, I., Su, X., Wolf, M.-M., et al. (2025). Tum teleoperation: Open source software for remote driving and assistance of automated vehicles. arXiv preprint arXiv:2506.13933

    work page arXiv 2025

  12. [12]

    Lichiardopol, S. (2007). A survey on teleoperation

    work page 2007

  13. [13]

    P., McDermott, P

    Luck, J. P., McDermott, P. L., Allender, L., and Russell, D. C. (2006). An investigation of real world control of robotic assets under communication latency. In Proceedings of the 1st ACM SIGCHI/SIGART Conference on Human-Robot Interaction , HRI '06, page 202–209, New York, NY, USA. Association for Computing Machinery

    work page 2006

  14. [14]

    Majstorovi \'c , D., Hoffmann, S., Pfab, F., Schimpe, A., Wolf, M.-M., and Diermeyer, F. (2022). Survey on teleoperation concepts for automated vehicles. In 2022 IEEE international conference on systems, man, and cybernetics (SMC) , pages 1290--1296. IEEE

    work page 2022

  15. [15]

    A., and El-Sayed, H

    Malik, S., Khan, M. A., and El-Sayed, H. (2022). Carla: Car learning to act—an inside out. Procedia Computer Science , 198:742--749

    work page 2022

  16. [16]

    Roadrunner – 3d-editor zum erstellen von szenen für das simulations- und testen von automatisierten fahrsystemen

    MathWorks (2025). Roadrunner – 3d-editor zum erstellen von szenen für das simulations- und testen von automatisierten fahrsystemen. Accessed on 11 September 2025

    work page 2025

  17. [17]

    Mindfield Biosystems Ltd. (2025). Skin response [apparatus and software]. Accessed: 21 August 2025

    work page 2025

  18. [18]

    Neumeier, S., Wintersberger, P., Frison, A.-K., Becher, A., Facchi, C., and Riener, A. (2019). Teleoperation: The holy grail to solve problems of automated driving? sure, but latency matters. In Proceedings of the 11th International Conference on Automotive User Interfaces and Interactive Vehicular Applications , AutomotiveUI '19, page 186–197, New York, ...

    work page 2019

  19. [19]

    State of connectivity between and within countries: Closing broadband connectivity divides for all

    OECD (2025). State of connectivity between and within countries: Closing broadband connectivity divides for all. Technical report, Organisation for Economic Co-operation and Development (OECD). Accessed on 11 Sep 2025

    work page 2025

  20. [20]

    Polar h10 heart rate sensor

    Polar Electro Oy (2025). Polar h10 heart rate sensor. Accessed: 21 August 2025

    work page 2025

  21. [21]

    Richardson, I. E. (2004). H. 264 and MPEG-4 video compression: video coding for next-generation multimedia . John Wiley & Sons

    work page 2004

  22. [22]

    a ring, I., Vogelbacher, G., H \

    Richter, A., Walz, T. P., Dhanani, M., H \"a ring, I., Vogelbacher, G., H \"o flinger, F., Finger, J., and Stolz, A. (2023). Components and their failure rates in autonomous driving. In Proceeding of the 33rd European Safety and Reliability Conference , pages 233--240

    work page 2023

  23. [23]

    Saparia, S., Schimpe, A., and Ferranti, L. (2021). Active safety system for semi-autonomous teleoperated vehicles. In 2021 IEEE Intelligent Vehicles Symposium Workshops (IV Workshops) , pages 141--147. IEEE

    work page 2021

  24. [24]

    Next stop for waymo one: Washington, d.c

    Waymo (2025). Next stop for waymo one: Washington, d.c. Accessed: 2025-09-12

    work page 2025

  25. [25]

    Xiao, Y., Krunz, M., Volos, H., and Bando, T. (2019). Driving in the Fog : Latency Measurement , Modeling , and Optimization of LTE -based Fog Computing for Smart Vehicles . In 2019 16th Annual IEEE International Conference on Sensing , Communication , and Networking ( SECON ) , pages 1--9, Boston, MA, USA. IEEE

    work page 2019

  26. [26]

    Zhao, L., Nybacka, M., Aramrattana, M., Rothhämel, M., Habibovic, A., Drugge, L., and Jiang, F. (2024). Remote Driving of Road Vehicles : A Survey of Driving Feedback , Latency , Support Control , and Real Applications . IEEE Transactions on Intelligent Vehicles , 9(10):6086--6107

    work page 2024

  27. [27]

    Zoox opens its las vegas robotaxi service to the public

    Zoox (2025). Zoox opens its las vegas robotaxi service to the public. Accessed: 2025-09-12

    work page 2025