Simulation Results on Selector Adaptation in Behavior Trees

Behavior trees (BTs) emerged from video game development as a graphical language for modeling intelligent agent behavior \cite{halo,lim2010evolving}. BTs have advantages of modularity and scalability with respect to finite state machines. When implementing intelligent behavior with BTs, the designer of a robotic control system breaks the task down into modules (BT leaves) which return either "success" or "failure" when called by parent nodes. All higher level nodes define composition rules to combine the leaves including: Sequence, Selector, and Parallel node types. A Sequence node defines the order of execution of leaves and returns success if all leaves succeed in order. A Selector node (also called "Priority" node by some authors) tries leaf behaviors in a fixed order, returns success when a node succeeds, and returns failure if all leaves fail. Decorator nodes have a single child and can modify behavior of their children with rules such as "repeat until $X>0$". BTs have been explored in the context of humanoid robot control \cite{marzinotto2014towards,colledanchise2014performance,bagnell2012integrated} and as a modeling language for intelligent robotic surgical procedures \cite{hu2015semi}.