Temporal Behavior Trees — Segmentation

Sebastian Schirmer, Jasdeep Singh, Emily Jensen, Johann Dauer, Bernd Finkbeiner, Sriram Sankaranarayanan

We present our tool for the segmentation of temporal behavior trees (TBT), a novel formalism for monitoring specifications. TBTs can be easily retrofitted to behavior trees, commonly used to program robotic applications. Our tool supports the robustness semantics of TBT and generates trace segmentations. In other words, given a TBT specification and a trace, it determines the optimal assignment of TBT nodes to sub-traces. To illustrate its application, we use the example of an autonomous ship deck landing. We showcase the user inputs required and demonstrate how the outputs can be interpreted to identify challenging task aspects, contributing to a comprehensive system analysis.

27th ACM International Conference on Hybrid Systems: Computation and Control (HSCC 2024).