Haptic Rapidly-Exploring Random Trees: A Sampling-based Planner for Quasi-static Manipulation Tasks

2025-05-31

AI Summary

Traditional motion planning algorithms fail for quasi-static manipulation tasks involving rich contact interactions—such as inserting a book into a crowded shelf—because they neglect active contact forces and mechanical equilibrium constraints. This paper proposes a sampling-based planning framework targeting implicit equilibrium manifolds. First, it formulates quasi-static manipulation as a constrained satisfaction problem defined on the equilibrium manifold. Second, it introduces a haptic metric derived from the Hessian of the manipulation potential function, explicitly linking geometric properties of haptic obstacles to multi-branch manipulation strategies. Third, it develops an enhanced RRT algorithm supporting implicit constraint sampling and active pushing. Evaluated on both a simulated inverted pendulum and real-world bookshelf insertion tasks, the method autonomously discovers wedge-based insertion and multi-branch pushing trajectories. Planning success rates increase significantly, demonstrating the approach’s effectiveness and generalizability in highly contact-rich environments.

Technology Category

• Intelligent Robots: Manipulation
• Search and Optimization: Sampling/Simulation-based Search
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

In this work, we explore how conventional motion planning algorithms can be reapplied to contact-rich manipulation tasks. Rather than focusing solely on efficiency, we investigate how manipulation aspects can be recast in terms of conventional motion-planning algorithms. Conventional motion planners, such as Rapidly-Exploring Random Trees (RRT), typically compute collision-free paths in configuration space. However, in manipulation tasks, intentional contact is often necessary. For example, when dealing with a crowded bookshelf, a robot must strategically push books aside before inserting a new one. In such scenarios, classical motion planners often fail because of insufficient space. As such, we presents Haptic Rapidly-Exploring Random Trees (HapticRRT), a planning algorithm that incorporates a recently proposed optimality measure in the context of extit{quasi-static} manipulation, based on the (squared) Hessian of manipulation potential. The key contributions are i) adapting classical RRT to a framework that re-frames quasi-static manipulation as a planning problem on an implicit equilibrium manifold; ii) discovering multiple manipulation strategies, corresponding to branches of the equilibrium manifold. iii) providing deeper insight to haptic obstacle and haptic metric, enhancing interpretability. We validate our approach on a simulated pendulum and a real-world crowded bookshelf task, demonstrating its ability to autonomously discover strategic wedging-in policies and multiple branches. The video can be found at https://youtu.be/D-zpI0RznZ4

Problem

• Adapting RRT for contact-rich quasi-static manipulation tasks
• Exploring manipulation strategies via equilibrium manifold branches
• Enhancing interpretability of haptic obstacles and metrics

Innovation

• Adapts RRT for quasi-static manipulation planning
• Incorporates Hessian-based optimality measure
• Discovers multiple equilibrium manifold branches