MoMa-Pos: An Efficient Object-Kinematic-Aware Base Placement Optimization Framework for Mobile Manipulation

TL;DR

MoMa-Pos is a framework that optimizes base placement for mobile manipulators, focusing on navigation-manipulation tasks in environments with both rigid and articulated objects, offering improved efficiency, precision, and adaptability across diverse settings and robot models.

Abstract

In this work, we present MoMa-Pos, a framework that optimizes base placement for mobile manipulators, focusing on navigation-manipulation tasks in environments with both rigid and articulated objects. Base placement is particularly critical in such environments, where improper positioning can severely hinder task execution if the object's kinematics are not adequately accounted for. MoMa-Pos selectively reconstructs the environment by prioritizing task-relevant key objects, enhancing computational efficiency and ensuring that only essential kinematic details are processed. The framework leverages a graph-based neural network to predict object importance, allowing for focused modeling while minimizing unnecessary computations. Additionally, MoMa-Pos integrates inverse reachability maps with environmental kinematic properties to identify feasible base positions tailored to the specific robot model. Extensive evaluations demonstrate that MoMa-Pos outperforms existing methods in both real and simulated environments, offering improved efficiency, precision, and adaptability across diverse settings and robot models. Supplementary material can be found at https://yding25.com/MoMa-Pos

Figures (5)

• Figure 1: Framework of MoMa-Pos. The framework is composed of four key phases. Phases 1 and 2 form the task-specific kinematic perceptual modeling, while Phases 3 and 4 focus on base placement optimization and guided navigation. In Phase 1, key objects in the scene are prioritized to enable efficient kinematic modeling without processing every object. In Phase 2, kinematic modeling is conducted for these prioritized objects, supporting the subsequent object-kinematic-aware base placement optimization. In Phase 3, potential base placement areas are identified by considering both robot-specific constraints and environmental kinematics. Finally, in Phase 4, the robot navigates to the optimal position, ensuring physical feasibility and adaptability for task execution.
• Figure 2: Our mobile manipulator includes: a base, a body, and an arm. The symbol $\Delta r$ represents the horizontal distance between the base's placement and the end-effector in the x-y plane. $r^b$ represents the base dimensions, $r^h$ is the body height, and $r^l$ is the arm extension length. For explanations of additional symbols, see Section \ref{['sec:problem_setup']}.
• Figure 3: The upper-left figure provides an overview of the robots' operating environment, while the three subfigures on the right focus on key task-relevant objects in simulation. Green circles indicate potential robot standing positions and red diamonds represent targets, such as the fridge door handle. In example (1), when the door handle changes position, the corresponding feasible placement map is also altered, consistent with intuitive expectations. This is best viewed in the enlarged digital version.
• Figure 4: Real Robot Demonstration. We implemented MoMa-Pos on a real robotic system comprising an AgileX mobile platform and a Realman-63F robotic arm. The robot's task in this setup is to open a fridge door. In the first figure, MoMa-Pos predicts key objects in the environment, with these objects highlighted by colored outlines. Meanwhile, the manipulation waypoints are predicted and indicated by yellow circles. Next, MoMa-Pos performs kinematic modeling, representing rigid objects as boxes and articulated objects using URDFormer. A feasible base placement is then selected, taking into account robot-specific constraints and the kinematic properties of the environment. The final two figures illustrate the robot successfully completing navigation, where the trajectory is indicated in a dashed black line, and fixed-base manipulation tasks using the predicted waypoints.
• Figure 5: The modeling time comparison between MoMa-Pos and its variant, which lacks the object importance prediction, with the x-axis denoting the location of the target object.

Theorems & Definitions (1)

• proof