How would you reduce lag in RTS game?

In my RTS game prototype, I could not avoid extreme lag when there's only 20 units in a small map.



As you can see, 1 unit has 2 triggers and 1 collider, and they are, respectively: Line of Sight, Attack Range, and RigidBody.

The game in the editor is running 2 FPS on a gaming laptop. And it's really pushing my laptop fans.

I would like to know what is the best way to reduce lag. Does that mean I have to remove my triggers?

 

Check your profiler. What you described should do way better than 2fps. There is probably something else causing the problem.

 

Ok, I found out I'm calling Physics.OnTriggerStay() a lot. 100x a lot. It's bringing down the CPU, utilizing up to 98% of it.

The most weirdest thing is that without OnTriggerStay(), the units themselves couldn't identify if the unit in the triggers is an enemy or a friend.

At least I found my culprits. Thanks.

 

Yup. Don't use the built in physics for an RTS. And don't use OnXXXStay for anything.

 

How would I create an alternate?

 

Usually you'd use ontriggerenter and exit... if you need to do something as they are in the trigger use enter/exit to set a bool so they can determine if they need to do something special. But if it is for something such as enemies in range... on enter add them to a list and use the list to find closest or whatever.

 

How many units do you want? Agent based simulations in unity tend to break down somewhere between 200-1000 agents, depending on the device and agent complexity.

For small level simulations simply using OnTriggerEnter and OnTriggerExit will work. I would also suggest making the rigidbodys kinematic and using navmesh agents..

For higher numbers of units you want to start using your own solution for storing their location. Something like a quad tree. Start treating Unity as a rendering engine only.

For über high numbers you'll want to leave behind the GameObject altogether. You'll need to start using imposters. And your simulation will run entirely outside of Unity. There are techniques specifically designed for high numbers of units.