2014 and pets are still mentally challenged
examples:
1. when the target player suddenly changes direction in movement, pet just blanks out for a few seconds (then struggles to catch up)
2. whenever you jump down (not even big cliffs but even small stairs) the pet takes the long way around
3. all pet spells that are ‘cast in a cone’ or spells like lightning bolt from drake: please make the pet turn to follow the target
The problem is that all AI agents in this game are handled at a very low tickrate for performance reasons. Thats why pet AI which is handled by the same systems is so bad. They are just as bad as AI mobs and can’t do anything AI mobs can’t do. Improved AI behaviour like finding alternative routes that involve jumping down cliffs can be a very demanding task for a computer, especially if you have to do it for hunders of pets at the same time. An elegant solution for this would be to push pet AI to the client and handle it in a seperate thread (since the games doesn’t seem to take advantage of multiple cores anyway there is alot of spare processing power to be harvested). Bandwidth from the server to the client wouldn’t really need to increase since pets are always close to the player anyway, so no addional data would be needed. But this probably involves a completely new system for client side AI so something like this will probably never see the light of day, since anets way of fixing hard problems is unsatisfying bandaid fixes.
We were told the problem is too deeply rooted into the games meachnics as Pfanne has explained and therefore they aren’t going to bother. I believe they said they’d look to change “low hanging fruit” first.