Modular Multi-Level Replanning TAMP Framework for Dynamic Environment

Abstract

Task and Motion Planning (TAMP) algorithms can generate plans that combine logic and motion aspects for robots. However, these plans are sensitive to interference and control errors. To make TAMP algorithms more applicable and robust in the real world, we propose the <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">m</u> odular <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">m</u> ulti-level <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">r</u> eplanning TAMP <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">f</u> ramework(MMRF), expanded existing TAMP algorithms by combining real-time replanning components. MMRF generates an nominal plan from the initial state and then reconstructs this nominal plan in real-time to reorder manipulations. Following the logic-level adjustment, MMRF attempts to replan a new motion path, ensuring that the updated plan is feasible at the motion level. Finally, we conducted several real-world experiments. The result demonstrated MMRF swiftly completing tasks in scenarios with moveing obstacles and varying degrees of interference.