Hi all, i have done an extensive amount of searching the past few weeks and haven’t seemed to find anything like this. I have raw simulation data that defines a timestamp, object info, and coordinates in a .csv file like the below.
I am able to get the data parsed, loaded into unreal editor, and Actor BPs based on the descriptors. Then by using the “move component to” function, along with a variety of structures, maps, and arrays, i can get this series of static meshes to move from coordinate to coordinate at the correct time intervals from this data.
The issue arises here; i would like to place characters, in this case vehicles, along these paths and have them move/animate/drive along these paths at the appropriate times. The simulation includes curves, stops, etc.
Is there a way to generate splines from this csv data and assign time values along them so that they reach each point at the correct timestamp?
If you assume that bicubic will work, then what you have is kind of enough.
Simply figure out what curve to apply between the locations you have.
In practice, its going to be eons better if you output a tranform for each spline point instead of just the LOC.
In that way you “assume” nothing, and the data you feed in is really in charge of what ends up happening.
Which would be critical if your Z weren’t all set to 0.
In all occurences, you should probably revise this so that you have separate stuff.
From one sheet you generate the splines.
From another, you set the variables that atart the cars and their speeds.
If you are then supposed to simulate, you can’t “force” movement along the spline, you should let it be automated based on whatever the vehicle speed is - because there’s no way you can absolutely guarantee the accuracy of loc over time id the vehicle is simulating.
For stuff like that, its usually best to drive via AI controller - “follow the closest spline as best you can” is what you get out of it.
And save any bugs, it allows the same road to have individual lanes that the AI can swap to and from.
Again, getting to where the object manages to be at the right LOC at the right time while simulating is near impossible.
So if you really need that, Don’t bother using characters. Just move meshes along…
What additional parameters would be beneficial? This is just a stripped down subset of the data, there is a ton more than can be pulled, I just pulled what I thought would be relevant.
The mesh solution is what I was afraid of. The coordinates come from a traffic microsimulator and models all facets of the traffic movements. This includes lane switches, intersections, vehicle interactions, and pedestrian movements. So technically the “simulation” is already done and represented by the data. We are now looking to bring it into unreal in order to create realistic and interactive visualizations.
Creating a transform is simple enough and I can just add that to my parsing script.
The transform - if you have the original data - will better define the locations for the spline you create.
If the data is real, I’m afraid you cannot just assume a bicubic tranform from point to point.
I mean, you can, but if that is not how the real thing goes then your visualization will have a margin or error that could be beyond what you would expect.
Also depends a lot on data location i suppose?
Continental US where streets follow the roman grid pattern should be simple enough, and bicubic will almost always work to define turns.
European like circular infrastructure with roundabouts and generally speaking - a mess that doesnt make sense even those who live the area?
Yea, thats going to be super tough to match when you assume bicubic between transforms…
Anyway,
If your data is real, maybe pulling splines of the road off OpenStreetMaps is a better idea?
I’d probably try that first instead of assuming bicubic.
If you have it, for additional info I would pull estimated speed (even if you can calculate/extrapolate between times).
In case you do simulate again you can set the object to that speed periodically, and it should end up getting less error prone.
Either way, if you assume a margin of error of about a second your simulation will end up being OK.
If you expect anything better you’d likely always be dissapointed unless you just place meshes at specific locations based on the time…
