Virtual Field Trip Sync: Metaverse Classroom Latency

Classroom Network

Avatars 16

Latency 120 ms

Packet loss 8%

Update rate 10 Hz

Sync Readouts

Avg drift

0.00 m

Sync accuracy

100%

Applied updates/s (all avatars)

0.0

Legend

Rendered avatar (dead-reckoned on client) True server position (ground truth)

How it works

In a shared VR/metaverse classroom the server only broadcasts each avatar's exact position a few times per second — sending every rendered frame would flood the network. Between updates, every other client must guess where the avatar is right now. The standard technique is dead reckoning (a.k.a. client-side prediction):

p_render(t) = p_last + v_last · (t − t_last)

where p_last/v_last are the position and velocity from the most recently applied network update, and t_last is when it was applied. This simulation runs each avatar's true, continuous motion on an imaginary "server," then simulates the network path to every "client":

Update rate — how often the server samples and sends a position+velocity packet per avatar.
Latency — one-way delay before a sent packet is applied on the client.
Packet loss — probability a given packet never arrives; the client keeps extrapolating from the older packet, so drift grows until the next one lands.

The cyan avatar is what a real classmate would actually see rendered; the faint red dot is the true position at that instant. Avg drift is the mean 3D distance between them across all avatars, and sync accuracy is the share of avatars currently within 0.3 m of ground truth — the threshold below which humans stop noticing desync in a shared VR space.

Field trip: Solar System Hall

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.