ABSTRACT
In this chapter, we describe how learning algorithms can be used to automatically create BTs, using ideas from [14,57]. First, in Section 8.2, we present a mixed learning strategy that combines a greedy element with genetic programming (GP) and show the result in a game and a robotic example. Then, in Section 8.3, we present a reinforcement learning (RL) algorithm applied to BTs and show the results in a game example. Finally in Section 8.5 we overview the main approaches used to learn BTs from demonstration.
