Latency Effects on Multi-Dimensional QoE in Networked VR Whiteboards

This study systematically investigates how network latency degrades the pragmatic and hedonic dimensions of Quality of Experience in networked VR whiteboards by analyzing their varying impacts across sequential and free collaboration modes, as well as between avatar-enhanced and traditional VR or PC-based systems.

The Gist

Imagine you are trying to draw a picture with a friend who lives in another country. You are both standing in front of a giant, magical whiteboard that exists in a virtual world. You can see your friend's "digital ghost" (an avatar) standing next to you, and you can both draw on the board at the same time.

This is what Networked Virtual Reality (NVR) Whiteboards are. They are amazing tools for teamwork, but they have a sneaky enemy: Latency.

Think of Latency as a "laggy telephone line." Even if you speak, your friend hears you a split second later. In a virtual whiteboard, if you draw a line, your friend sees it a moment later. If that moment is too long, the whole experience falls apart.

This paper is like a detective story where researchers tried to figure out exactly how much lag is too much and what kind of lag hurts the most.

Here is the breakdown of their investigation, explained simply:

1. The Two Ways People Work Together

The researchers realized that people don't just "draw" in the same way. They identified two main styles of teamwork:

• The Relay Race (Sequential Collaboration): Imagine you and your friend are drawing a picture, but you have to take turns. You draw a line, then stop and wait for your friend to draw the next part. If the "laggy telephone" is slow, you stand there awkwardly waiting, and the rhythm breaks. This is highly sensitive to lag.
• The Jam Session (Free Collaboration): Imagine you and your friend are both drawing at the same time, adding your own ideas wherever you want. If your friend's line appears a second late, you don't care as much because you are busy drawing your own part. This style is more forgiving of lag.

2. The Three "Pillars" of a Good Experience

The researchers asked: "What actually makes the experience feel bad when there is lag?" They found that while people care about having fun (hedonic), they care much more about three practical things (pragmatic):

• Interactivity (The "Flow"): Does it feel like I'm talking to a real person, or am I talking to a robot with a broken microphone?
• Efficiency (The "Speed"): Can I actually get my work done, or am I wasting time waiting for things to catch up?
• Believability (The "Trust"): Do I trust that my partner is actually there and paying attention, or does it feel like a glitchy simulation?

3. The Experiment: Testing the Limits

The team built a custom virtual whiteboard system where they could artificially add lag, like turning a dial from "Instant" to "Glacial." They tested three setups:

1. VR+ (With Avatars): You see your friend's digital body.
2. VR (No Avatars): You only see the drawing lines, no bodies.
3. PC (Computer): You are drawing on a regular 2D screen with a mouse.

They tested these setups with both the "Relay Race" and the "Jam Session" at different lag levels (from 100ms to 2500ms).

4. The Big Discoveries

The "600ms" Magic Number

They found a critical threshold.

• Below 600ms: Everything feels great. You can do a "Relay Race" perfectly fine.
• Above 600ms: The "Relay Race" (Sequential) starts to fall apart. People get frustrated because they can't wait for each other.
• The "Jam Session" (Free): Even with high lag (up to 2000ms), people could still get work done because they weren't waiting for each other.

The Avatar Paradox (The "Ghost" Effect)

This was the most surprising part.

• When lag is low: Seeing your friend's avatar (VR+) is amazing. It makes you feel connected, like you are in the same room. It boosts trust and fun.
• When lag is high: Seeing your friend's avatar becomes painful. If your friend's hand moves but the line they draw appears 2 seconds later, it looks like a glitch. It breaks the "magic" and makes you feel like the system is broken.
• The Lesson: If the internet is slow, it might actually be better to hide the avatars and just show the drawings. It's like turning off the video on a Zoom call when the connection is bad so you can still hear the audio clearly.

The "Trust" Factor

In the "Jam Session," even with lag, people still trusted their partner if they could see the avatar. The avatar acted as a "social anchor," proving the other person was still there, even if the drawing was delayed.

5. The Takeaway for the Real World

If you are building a virtual whiteboard app for teams, here is the cheat sheet:

1. Keep it under 600ms: If you want people to take turns drawing (like in a meeting), the internet speed must be very fast.
2. Switch modes for bad internet: If the connection is slow, switch the app to "Free Mode" (let everyone draw at once) rather than "Turn-taking Mode."
3. Be smart with Avatars: Show the avatars when the connection is good to build trust and fun. But if the connection gets bad, consider hiding the avatars to stop the "glitchy" feeling and focus on the work.

In a nutshell: Latency is the enemy of teamwork. But if you understand how people work together (taking turns vs. drawing freely) and what they see (avatars vs. just lines), you can design systems that keep the team connected even when the internet isn't perfect.

Technical Summary

Here is a detailed technical summary of the paper "Latency Effects on Multi-Dimensional QoE in Networked VR".

1. Problem Statement

Networked Virtual Reality (NVR) whiteboards are emerging as critical tools for remote, immersive collaboration. However, their effectiveness is heavily compromised by network latency (caused by transmission delays, rendering, or synchronization). While prior research has compared VR whiteboards to traditional ones in local settings, there is a lack of systematic understanding regarding:

• How latency differentially impacts specific Quality of Experience (QoE) dimensions (Pragmatic vs. Hedonic).
• How these impacts vary across different collaboration modes (Sequential vs. Free).
• How platform configurations (specifically the presence of embodied avatars) modulate latency sensitivity.
• The specific mechanisms through which latency degrades the collaborative experience.

2. Methodology

The study employed a two-stage experimental approach using a custom-built, latency-controllable NVR whiteboard system.

A. Experimental Platform

• System: Built on Unity (Meta Quest 2, Horizon OS) using WebRTC for peer-to-peer (P2P) communication.
• Latency Control: Implemented a precise End-to-End (E2E) latency simulation mechanism using a time-based buffer queue. The system could artificially introduce latency ranging from 100ms to 2500ms while maintaining a baseline of ~80ms.
• Comparative Platforms:
  1. VR+: VR with visible user avatars (embodied actions).
  2. VR: VR without avatars (only drawing strokes visible).
  3. PC: Traditional mouse-based electronic whiteboard (baseline).
• Collaboration Modes:
  • Sequential Collaboration (SC): Structured, turn-taking workflow (tightly coupled).
  • Free Collaboration (FC): Open, simultaneous drawing (loosely coupled).

B. Participants & Procedure

• Participants: 36 university students (24 in the formal study, 18 in a pilot study).
• Design: A within-subjects design testing 7 latency levels (100ms to 2500ms) across 3 platforms and 2 modes.
• Metrics:
  • Pilot Study: Evaluated 8 QoE sub-dimensions (3 Pragmatic: Interactivity, Efficiency, Believability; 5 Hedonic: Enjoyment, Immersivity, etc.) to identify sensitivity.
  • Formal Study: Focused on the three identified pragmatic dimensions plus Overall QoE, rated using the 5-point Absolute Category Rating (ACR) scale.

3. Key Contributions

1. Custom Research Tool: Developed and open-sourced a flexible NVR whiteboard system capable of precise latency manipulation and data collection, overcoming limitations of commercial platforms.
2. QoE Dimension Identification: Identified that Pragmatic dimensions (Interactivity, Efficiency, Believability) are significantly more sensitive to latency and more critical to users than Hedonic dimensions.
3. Mechanism Analysis: Uncovered how latency shapes QoE differently based on collaboration mode and platform, revealing that 300ms is a critical threshold for interactivity divergence and 600ms for efficiency collapse in sequential tasks.
4. Avatar Impact: Demonstrated that while avatars (VR+) generally enhance QoE by providing social cues, they can amplify the perception of inefficiency under high-latency conditions in tightly coupled tasks.

4. Key Results

A. Pilot Study Findings

• Pragmatic vs. Hedonic: Users prioritize Efficiency (100% affected, 94% cared), Believability (89% affected, 91% cared), and Interactivity (86% affected, 81% cared).
• Hedonic Dimensions: While "Enjoyment" was perceived as affected, it was rarely considered critical. Dimensions like "Immersivity" and "Novelty" showed low sensitivity to latency.

B. Impact on Specific Dimensions

• Interactivity:
  • VR+ consistently outperformed VR and PC due to embodied cues.
  • Threshold: Above 300ms, Free Collaboration (FC) maintained higher interactivity than Sequential Collaboration (SC) because SC relies on tight turn-taking, which is disrupted by delay.
• Efficiency:
  • SC is highly fragile; efficiency drops sharply after 600ms.
  • VR+ suffered the steepest decline in SC at high latencies (>1000ms) because the delay in avatar synchronization made the inefficiency more salient.
  • FC remained resilient across platforms even at higher latencies due to its parallel nature.
• Believability (Trust):
  • VR+ maintained the highest believability scores even at high latencies. Avatars provided social cues (presence) that sustained trust in the partner's engagement, even when actions were delayed.
  • SC was more sensitive to latency than FC regarding trust.

C. Overall QoE & Dominant Factors

• Low Latency (≤300ms): All platforms performed well; VR+ was superior.
• Moderate Latency (600–2000ms):
  • SC: QoE is dominated by Interactivity. Users are sensitive to flow disruption.
  • FC: QoE is dominated by Believability. Users care more about partner presence than immediate responsiveness.
• High Latency (>2000ms):
  • SC: QoE collapses; Efficiency becomes the dominant factor (can the task be done?).
  • VR+ vs. Others: At extreme delays, the benefit of avatars diminishes in SC as the delay in avatar movement exacerbates the feeling of inefficiency.

5. Significance and Implications

• System Optimization: Designers of NVR whiteboards should prioritize keeping E2E latency below 600ms for sequential workflows to prevent QoE collapse.
• Avatar Strategy:
  • Beneficial: Avatars are crucial for maintaining trust (Believability) in free-form discussions and low-latency sequential tasks.
  • Detrimental: In high-latency sequential scenarios, avatars may amplify negative perceptions of efficiency. Systems might consider dynamically reducing avatar fidelity or hiding them when latency exceeds critical thresholds to preserve efficiency.
• Mode Adaptation: Systems should adapt interaction logic based on network conditions. For example, shifting from strict turn-taking (SC) to parallel input (FC) when latency rises can preserve user experience.
• Evaluation Framework: The study provides a validated framework for evaluating NVR collaboration, emphasizing pragmatic dimensions over hedonic ones for latency-sensitive applications.

Conclusion

The paper establishes that latency does not degrade QoE uniformly; rather, it selectively erodes specific pragmatic dimensions depending on the collaboration mode. While avatars generally enhance the experience by fostering trust, their utility is conditional on network stability. These findings offer actionable guidelines for optimizing NVR whiteboard systems to ensure robust performance under real-world network constraints.