Page 1 of 1

Physics simulation over the network

Posted: Wed May 24, 2017 12:31 pm
by B1TZ3R0
Hi, I'm curious about how RocketLeague has implemented its perfect replication of physics movement on the network.

I readed about this physics engine that is deterministic, but I think that only this can't be sufficent to allow a smooth movement over the network without errors and scats. In fact send only the acceleration, and position (at fixed time rate) not allow the server to be always 100% in sync with client.

Usually to replicate a normal character movements are sent a lot of infos like position, acceleration, delta time and etc..., almost each tick to allow the server to replicate perfectly each movement, and the server is always little behind the client movement.

This approach (with best conditions, of course) allow the server to execute always the movements like the client does before.

So what I thought is that each client execute the simulation only on its controlled actor and when send these data to the server, the server can simulate the movement only for this actor with custom DeltaTime.

This approach is way less performant than the first, especially for the server (that for each client update must execute a simulation), and I don't know if this approach is used by RocketLeague.

I don't know much about this engine and if it has a function that can be called each time to simulate a particular actor. However what do you think about how RocketLeague simulate the movement of a car so nice over the network?