What is better Rigidbody View or Transform view (fastest)

Options
I have cars in my game that I want other users to see. I have regular cars that I will just use transform view and I have the players cars which are more complex.

The player car has 15 transform (wheels, a arms... etc) and the main car is a rigid body. Would it be faster to send all 15 transforms as Transform view or should I send the main body as a rigidbody view and the rest as transform views. I don't care if the players collide, I don't plan on having any physical behaviour when they collide, other than stopping the cars.

Comments

  • KingDruid
    KingDruid
    edited December 2015
    Options
    Any suggestions? I'm trying to make a car game. What I just tried right now is getting the worker demo and instead of spawning the worker, I spawned a car prefab (with wheel colliders) which I added a Photon Rigidbody View script to which automatically added the Photon View script. I can see both cars and have also figured out how to control them on each side, but what I noticed is that the rigid body has a small shake up and down to them on the other players screen. I understand that I have to animate the wheels and everything on the other car, but how do I fix the shaking on the models on the remote side so it looks smooth.

    Photon RIgidbody has:
    Synchronize Velocity: Checked
    Synchronize Angular Velocity: Checked

    Photon view has:
    Unreliable on Change for observe option
    Transform serialization is set to Position and Rotation.

    The shaking is a small shake usually more shaking when it comes to a stop, but its noticeable enough that I think people would ask why does the car always shake as it comes to a stop. Maybe its lag, but I'm running this on PC in 2 separate clients. I also tried on mobile and had the same results.
  • KingDruid
    KingDruid
    edited December 2015
    Options
    I just tried using the cube lerp script on the car model and it was a bit smoother, but buggy because at times the car would move sporadically, but other times it worked pretty good.

    The downside to this method is that I use the cars velocity, and suspension position to determine where to put the wheels and how to rotate them. Using this method would mean that I cant use that because im just sending the transform and none of the actual car data will be in every players scene, so I will have to send all the wheels and suspension pieces like I did the car using the cube lerp script. I counted the number of moving objects under my car model and its roughly 15 to 30 objects that move around so each player would be sending transforms location, rotation and like 4 of them also scale.

    ---I read that there is a bolt tutorial somewhere where they send the car input from each player then that is just translated in each users game by sending the input to each car in the scene. I haven't tested this, but I'm wondering how well that works because doesn't the physics behave differently on different devices??? I'm not sure what method is the best to use for unity3d cars with wheel colliders.
    ---(THIS METHOD is a NOGO because the physics are dynamic and unreliable when I tried it)
  • KingDruid
    Options
    This forum seems dead. If none of the developers want to help out, do any users have experience with both these methods, what is better to use for a unity3d car with wheel colliders.
  • crodriguez08
    edited January 2016
    Options
    I'm no expert at this but why would you add Transform Views to each part of the car if they're already children to the main body?