They are just going to be "smaller" - robots are amazingly efficient at short range operations, but at scale trains >> car hacks or mods >> belts >> robots. You *can* invest enough resources in robots to do anything you can do with trains or belts, but it is resource inefficient and has extremely complicated bottlenecks. Worse, the bottlenecks are often very hard to see, compared to belts or trains, and harder still to both fix, and measure if your fix worked.
On the other hand, in most later games, resources are essentially free, so my hot take is that a "fast enough" super-huge logistic robot network is fine. Sure, you need tens of thousands of robots, plus supporting roboports, plus the power to drive them, but none of that costs anything but a little time scaling the automated construction of the ingredients (ei: building robots and roboports), and a tiny amount of blueprint time to stamp down "moar roboports!" where you need them.
Anyway, an easy way to make good logistics network through bad rules: you can have one network with <= 50 roboports, and all others are limited to <= 5 roboports. By "roboports" I mean "logistic network area", so if you use modded ones with, say, a 200x200 logistic area you get one and only one of those.
I'd spend the big network on a convenient mall: all the stuff I don't need high production volume, or where I want to add "one tiny assembler" to build, eg, ammunition for just me, while the bulk work was done elsewhere.
PS: by "car mods" I mean AAI Programmable Vehicles, basically.
