Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

H-MaP: An Iterative and Hybrid Sequential Manipulation Planner

Berk Cicek, Arda Sarp Yenicesu, Cankut Bora Tuncer, Kutay Demiray, Ozgur S. Oguz

TL;DR

Experimental results across seven diverse tasks demonstrate H-MaP's superior performance compared to existing methods, particularly in highly constrained environments where traditional approaches fail due to local minima or scalability issues.

Abstract

This paper introduces H-MaP, a hybrid sequential manipulation planner that addresses complex tasks requiring both sequential actions and dynamic contact mode switches. Our approach reduces configuration space dimensionality by decoupling object trajectory planning from manipulation planning through object-based waypoint generation, informed contact sampling, and optimization-based motion planning. This architecture enables handling of challenging scenarios involving tool use, auxiliary object manipulation, and bimanual coordination. Experimental results across seven diverse tasks demonstrate H-MaP's superior performance compared to existing methods, particularly in highly constrained environments where traditional approaches fail due to local minima or scalability issues. The planner's effectiveness is validated through both simulation and real-robot experiments.

H-MaP: An Iterative and Hybrid Sequential Manipulation Planner

TL;DR

Experimental results across seven diverse tasks demonstrate H-MaP's superior performance compared to existing methods, particularly in highly constrained environments where traditional approaches fail due to local minima or scalability issues.

Abstract

This paper introduces H-MaP, a hybrid sequential manipulation planner that addresses complex tasks requiring both sequential actions and dynamic contact mode switches. Our approach reduces configuration space dimensionality by decoupling object trajectory planning from manipulation planning through object-based waypoint generation, informed contact sampling, and optimization-based motion planning. This architecture enables handling of challenging scenarios involving tool use, auxiliary object manipulation, and bimanual coordination. Experimental results across seven diverse tasks demonstrate H-MaP's superior performance compared to existing methods, particularly in highly constrained environments where traditional approaches fail due to local minima or scalability issues. The planner's effectiveness is validated through both simulation and real-robot experiments.
Paper Structure (23 sections, 3 equations, 5 figures, 1 table, 3 algorithms)

This paper contains 23 sections, 3 equations, 5 figures, 1 table, 3 algorithms.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. Optimization-Based Manipulation Planning
  4. Sampling-Based Manipulation Planning
  5. Constrained Sequential Manipulation Planning
  6. PRELIMINARIES
  7. Optimization Based Motion Planning with KOMO
  8. Problem Formulation
  9. ITERATIVE HYBRID MANIPULATION PLANNER
  10. Overview
  11. Waypoint Generation
  12. Contact Point Generation
  13. Optimization-Based Motion Planning
  14. Extension to Dynamic and Bimanual Scenarios
  15. Experiments
  16. ...and 8 more sections

Figures (5)

  • Figure 1: (Top) Examples of tool manipulation in nature: a crow using a tool to obtain foodmccoy2019new and manipulating a sliding bolt latch, alongside a Franka Panda robot performing an analogous obstacle removal task. (Bottom) Complex manipulation scenarios solved by our proposed approach: tool-assisted object retrieval through pushing and picking, latch mechanism manipulation, and tool-based obstacle removal.
  • Figure 2: System flowchart of H-MaP's three-phase architecture: (I) Bi-RRT-based waypoint generation with obstacle handling, (II) hybrid contact point determination through learning and sampling, and (III) optimization-based motion planning. The system supports both standard end-effector ($manipulator_0$) and tool-based manipulations ($manipulator_n$), enabling dynamic replanning and recursive obstacle handling for single and multi-tool scenarios.
  • Figure 3: From left to right: pushing an object through the tunnel; manipulating an object through the tunnel using a tool; operating a sliding latch lock; navigating an object with a tool around a fixed obstacle; and maneuvering an object with a tool while clearing movable obstacles. The top and bottom rows show initial and final configurations, respectively. The middle row displays generated waypoints (red spheres), contact points (yellow spheres), and intermediate object states (gray silhouettes) during task execution. For a complete demonstration: https://sites.google.com/view/h-map/
  • Figure 4: Real robot execution of bookshelf task: obstacle removal and tool-assisted book placement.
  • Figure 5: Performance comparison between learning-informed and sampling-only approaches, normalized to the sampling-only baseline (1.0x). Lower bars indicate better performance, with improvements shown as percentages. Tasks are ordered by relative improvement, highlighting the effectiveness of the learning-informed approach across manipulation scenarios.