ORLA*: Mobile Manipulator-Based Object Rearrangement with Lazy A Star

TL;DR

This work proposes ORLA*, which leverages delayed/lazy evaluation in searching for a high-quality object pick-n-place sequence that considers both end-effector and mobile robot base travel and can achieve global optimality.

Abstract

Effectively performing object rearrangement is an essential skill for mobile manipulators, e.g., setting up a dinner table or organizing a desk. A key challenge in such problems is deciding an appropriate manipulation order for objects to effectively untangle dependencies between objects while considering the necessary motions for realizing the manipulations (e.g., pick and place). To our knowledge, computing time-optimal multi-object rearrangement solutions for mobile manipulators remains a largely untapped research direction. In this research, we propose ORLA*, which leverages delayed (lazy) evaluation in searching for a high-quality object pick and place sequence that considers both end-effector and mobile robot base travel. ORLA* also supports multi-layered rearrangement tasks considering pile stability using machine learning. Employing an optimal solver for finding temporary locations for displacing objects, ORLA* can achieve global optimality. Through extensive simulation and ablation study, we confirm the effectiveness of ORLA* delivering quality solutions for challenging rearrangement instances. Supplementary materials are available at: https://gaokai15.github.io/ORLA-Star/

Figures (11)

• Figure 1: An example Mobile Robot Tabletop Rearrangement (MoTaR) setup.
• Figure 2: [Left] An example of the EE scenario, where the table is small and the robot can reach all poses from a fixed position. We count the traveling cost of the end-effector (EE) in the cost function. [Right] An example of the MB scenario, where the table is large and the robot can only reach a portion of tabletop poses from a fixed position. We count the traveling cost of the mobile base (MB) in the cost function.
• Figure 3: [Left] An example of MB scenario, where the robot (gray disc) travels along the green table following the black track along the table boundaries. To pick/place an object on the table, the robot moves to the nearest position before the manipulation. [Right] A working example where $o_1$ and $o_2$ block each other's goal pose. One must move to a buffer pose to finish the rearrangement.
• Figure 4: An example input of StabilNet when attempting to place a cup right on top of an apple in the environment. The ground truth label given by the simulation is a failure.
• Figure 5: Examples of disc instances. [Left] EE scenario with $\rho=0.2$; [Middle] MB scenario with $\rho=0.2$; [Right] EE scenario with $\rho=0.5$. Colored and transparent discs represent the initial and goal arrangements respectively.