Why doesn't AI work with WheeledVehicle


TL;DR: how to make BTTask with call AI Move To work with WheeledVehicle?

I am trying to implement simple car AI with only one behavior: move to a waypoint, reach it, then move to the next, and so on. I have implemented a custom pathfinding in c++ to reach the goal changing inputs of the controlled car, but this does not satisfy my goals due to important thing: I need to add some avoidance fot cars. I’ve tried RVOAvoidance, btw, but it seems not able to work both with custom moves (via SetThrottleInput and SetSteeringInput). Unfortunatelly, RVOAvoidance is not an option, becaise it doesn’t care about navmesh and world meshes, and the cars are, basically, won’t move.

I’ve heard about [DetourCrowds][2], so, to make it work I desired to remove the old controller, and make a new one, implementing the behavior in blueprints (inherited from a C++ class with UCrowdFollowingComponent), using BehaviorTrees, hoping it will work with ‘classic’ scheme.

Now, I am stuck in the stage where the car need to be moved.
When car was controlled by controller calling SetThrottleInput the control was fine


But my car won’t move using BTTask, calling AI Move To:

Inbefore: callings are correct, the values are correct and behavior tree is correct.

Also, if I am doing wrong, is there an AI Move To analog for cars?

The question is, how to make it work correctly?


How to make my old controller work with DetourCrowds using SetThrottleInput and SetSteeringInput?

Thanks in advance.

You cannot use navmesh and the current AI functions with vehicles right now. They work with humanoid characters, so you’ll have to make your own vehicle AI with custom made tasks, distance, speed, obstruction detections, etc.

Here you can find a simple car AI project provided by Peter and improve upon that.

Thanks. So, there is no other way except implement custom AI pathfinding manually?

Nope, not that i know of(I have been trying to make my own vehicle AI for a while now as well) It will be added eventually i hope, but not in the near future.