This week we hosted a seminar on AI Task Planning and Execution in ROS, given by Dr. Michael Cashmore (University of Strathclyde) and Dr. Gerard Canal (King’s College London). Drs Cashmore and Canal are the lead developers of the framework for embedding AI Planning tools in ROS.
The seminar covered four main topics:
– What is task planning for robotics? Including its key problem characteristics, such as non-determinism, uncertain duration, temporal constraints, and real-time environments.
– Techniques for modelling temporal task planning problems for robot scenarios, including tools for generating, optimising, and parsing task plans that describe behaviours for autonomous robots.
– Formalisms for flexible task plan execution, such as Simple Temporal Networks (STN) STNs with Uncertainty (STNU), and probabilistic STNs (PSTN).
– Common libraries and tools used for task planning in autonomous robotics.