8 comments

[ 3.0 ms ] story [ 32.4 ms ] thread
Found this to be a great introduction to an alternative way to model behaviors than a FSM. I used behavior trees for robotics and AI applications and found them generally simpler and that they produced much more modular and reusable code.
I would say that behavior trees are a higher-level abstraction on top of FSMs, such that behavior trees can be represented by equivalent FSMs (although with more states/transitions).
Interesting thanks. I wonder if anyone has combined BT with Evolutionary Algorithms, to automate the generation of them?
I had a similar idea when I was working with them but I never got around to trying it. I was primarily using them to compose other behaviors I had developed. It is not my specialization so I'm not up on the latest research around them but I would wager that there is some interesting research that could be done in that area.
You may be interested in chapter 8, Genetic Programming applied to BTs
So it does! I hadn't got that far when I skimmed the book.
Essential. BT's form the universal "language" with which to communicate with autonomous agents in an environment

Behavior Trees in Action: A Study of Robotics Applications

https://arxiv.org/abs/2010.06256