ABSTRACT
In this book, we have tried to present a broad, unified picture of BTs. We have covered the classical formulation of BTs, highlighting the implementation details for both computer game AI and robotics applications. We have shown that BTs generalize many successful control architectures, such as sequential behavior compositions, teleo-reactive programs, subsumption architectures, and decision trees. We have described a number of practical design principles that are useful in both computer game AI and robotics to fully exploit the potential of BTs. We have presented a new statespace formalism that allows us to analyze properties of BTs in terms of efficiency, safety, and robustness as well as estimates on execution times and success probabilities using a stochastic framework. We also showed the connections between BTs and the important areas of planning and learning.
