We're developing an infrastructure to support multiplayer games for our game engine.

In simple terms, each client (player) engine sends some pieces of data regarding the relevant game objects at a given time interval.

On the receiving end, we step the incoming data to current time (to compensate for latency), followed by a smoothing step (which is the subject of this question).

I was wondering how smoothing should be performed ?

Currently the algorithm is similar to this:

• Receive incoming state for an object (position, velocity, acceleration, rotation, custom data like visual properties, etc).

• Calculate a diff between local object position and the position we have after previous prediction steps.

• If diff doesn't exceed some threshold value, start a smoothing step:

  1. Mark the object's CURRENT POSITION and the TARGET POSITION.
  2. Linear interpolate between these values for 0.3 seconds.

I wonder if this scheme is any good, or if there is any other common implementation or algorithm that should be used?

(For example - should i only smooth out the position? or other values, such as speed, etc)

any help will be appreciated.