Multiplayer Theory

The problem with multiplayer would most likely come from the huge amount of node information that everybody needs to send and receive from the server, but couldn't it be optimized?


lets say the only simulated vehicle is on your local machine, and this vehicle had a matching "hitbox" that lined up with the appropriate coordinates, and where ever you went, your hotbox would follow and sent to the server.

When you see another player, you'd really be seeing is their hitbox and not the fully simulated vehicle.
In theory, you would be able to crash into other players, exchanging hitbox information, impacting your local vehicle simulation. Hitbox wheels would have to be "animated" to sell the effect.

No doubt this would be a lot of work and time from the Devs, but I'd love to imagine this could be possible.

 

Interesting

 

It doesnt sound like a bad idea but honestly i think that even that would be too much, but honestly i dont know.
Maybe if we took the refnodes location/movement and etc and only send the information of those nodes wouldnt be such a bad idea, leave to rest to be calculated on the other users computer, it would be less data to send then the entire jbeam, its around 4-6 nodes most of the times.

 

I still think the model would need to be heavily simplified. maybe like blacked out windows and no interior, less polys, simplified suspension, ect

 

Graphically speaking, there are two parts to a simulated vehicle from what I can tell. The simulated beam body (nodes and beams), and the 3D mesh that's imposed on that. The 3D mesh could be imposed onto the hitboxes, with their existing LODs (level of detail) providing graphical optimization. You would need to "fake" "animate" the wheels and such to create the sense that the hitbox server vehicles behave naturally.

 

no matter how much you "simplify" it, you still can't send information fast enough to get truly real-time performance because of lag, not the amount of data

 

exactly my thoughts
--- Post updated ---

Considering other games have similar methods I don't think it's impossible. All network systems fail under heavy lag, that's a given. With the ideal connection, I'm optimistic in the concept. We're talking about 50-100 coordinates of information, carrying mass and velocity, vs the thousands it takes to simulate it locally.

 

other games don't rely on a 2000fps physics model

 

If I wasn't clear before, there is no need to send that "2000fps" physics model

Your PC will simulate a single beam physics car, same old, nothing new there.
We attach a hitbox to that beam car at key reference points, this hitbox will compose of 50-100 nodes/beams/polys how ever you want to call them. Their only function is to follow the simulated car, record the coordinates, record the weight, and record the velocity, to provide collision information. This information will be relayed to the server, your simulated car will not.
When your crashing into a players car, your local simulated car is reacting to the hitbox of the other players car only, and vice versa on the other players end.

 

i.e. not real time, not accurate, results will not be identical on both ends

i.e. garbage

 

Since the early days of multiplayer it was and still is all interpolation, identical is not needed, just close enough to have some fun.

 

interpolation has always ruined fun, and will definitely make the fun non-existent here

 

Play a game of counter strike without interpolation and get ready to pull your hair out

 

I prefer tf2

 

The idea isn't bad.. I thought of smth similar where the vehicle has a low poly model and a simplyfied jbeam or no jbeam on distance, as soon as you are in range of like 30 meters or so, the model would load in, at 2 meters the jbeam would load in

The model and jbeam could be saved in the servers cache or ram to decrease loading times

 

Y'all are forgetting something. The devs said the replay system laid the groundwork for multiplayer, and that looks to be very true.
The replay app doesn't hold all the information, it interpolates stuff remember? It is possible to live transmit the replay file, and have the other PC(s) live play the file(s) with a similar app, that uses the same interpolation. Collision is something I haven't figured out yet, but this lays a solid ground for multiplayer using an existing feature.

 

Ah yes, I forgot about the replay system, Interesting hearing how you explain it. Through my experience in playing with it I'm pretty sure it simulates the jbeam body, and when playing the replays, crashes would be slightly different each time, but the cars would always follow the recorded path.

Ideally the hitbox system would go much more further in compressing that information.
Example being the players car you see in front of you from the server would only consist of that hitbox, with their full 3D model laid over top, the hitbox would show a basic form of deformation that happened on their machine, no parts of the car could fall off, not like a fully detailed simulation.

 

I am not being mean... I am just too lazy to type this again...

This second one I was a little aggravated at how they were treating the OP so its worded a little aggressive, but it still makes a point.

 

Heck, if the Devs finished a stable multiplayer version of the replay system, I'd have a blast just being able to street race/rock climb/drift with some buddies online as ghost cars and no collisions, but collisions aren't impossible.

Does everybody remember when we all thought real time soft body physics couldn't be done in games, yet Beamng told us to hold their beer? (for those who didn't know about RoR)