Specifying Failure and Progress Conditions in a Behavior-Based Robot Programming System

Behavior-based robot programs are designed by splitting the overall robot decision making process among concurrent processes, each of them being in charge of a well defined simple goal-oriented behavior. However, the combination of simple behaviors into more complex ones often incur global interactions that cannot be debugged by just considering individual behaviors. The normal way of dealing with global interactions that may cause failure in the robot execution is to program additional heuristic processes that monitor behaviors to check that they progress normally towards their intended goals. In this paper, we explain how Linear Temporal Logic (LTL) can be used as a declarative language for specifying an interesting class of such monitoring processes. This simplifies the task of writing robot control programs, increases the design modularity by clearly separating the control components from the monitoring ones, and augments the reliability of control programs thanks to a precise and clear LTL semantics.