Automated planning in hybrid domains for in-space robot assembly tasks

Mariana Dias Cunha

Key information

Authors:

Mariana Dias Cunha (Mariana Dias Cunha)

Supervisors:

Rodrigo Martins de Matos Ventura (Rodrigo Martins de Matos Ventura)

Published in

December 21, 2021

Abstract

Recent developments in the area of automation of in-space additive manufacture and assembly of structures has resulted in the development of in-space mobile robots that perform assembly and logistics operations inside the International Space Station (ISS). Consequently, studies and improvements in the area of robot Task-Motion Planning (TMP) need to be made in order to improve robot navigation. Inspired by this, we have created a hybrid domain using PDDL that describes mobile robots navigating through a 2D approximation sampled roadmap that are responsible for moving, loading and assembling modules. We have chosen a state of the art TMP approach called MPTP. However, in this approach, the motion planner is not aware of the world's configuration by the time it is invoked by the task planner, for we have contributed with an added logic to the approach that allows the motion planner to be informed of the map availability according to the actions expanded by the task planner. We designed a scenario as a proof-of-concept to test and compare the use of task and motion planners separately versus a mixed TMP approach. Our results illustrate that using a mixed approach in domains such as the space assembly domain presents advantages when compared to a separate approach since it is aware of physical motion constraints while task planning.