FlipItRight: Stable Pose-Targeted Throw-Flip Across Diverse Objects

2026-06-01

AI Summary

This work addresses the challenge of enabling high-degree-of-freedom robotic arms to perform stable and target-pose-controlled toss-and-catch maneuvers on diverse objects. The authors propose a two-layer planning architecture: an object-level planner generates candidate release states that satisfy desired landing poses, while a robot-level planner evaluates their feasibility and constructs executable swing trajectories. By explicitly representing release states as an intermediate abstraction, the approach enables principled candidate filtering, adaptive configuration selection, and structured motion design near release—all without requiring prior data or learned models, thereby generalizing to novel objects and targets. Combining analytical planning, a constant end-effector velocity swing strategy, and a model-free deployment mechanism, the method achieves a 90% success rate over 120 real-world trials across objects of varying shape, size, and mass, with ablation studies confirming the contribution of each component.

Abstract

We propose FlipItRight, a framework for stable planar pose-targeted throw-flip with a high-DoF manipulator. The task is decomposed into an object-level planner, which generates candidate release states satisfying the desired landing pose, and a robot-level planner, which evaluates executability and constructs a feasible swing motion. Treating the release state as an explicit intermediate representation enables principled candidate filtering, adaptive selection of release and pre-swing configurations, and structured near-release motion design -- in particular, approximately constant end-effector velocities during the final swing phase to improve robustness to release-timing uncertainty. We validate on a real platform across objects of varying shape, size, and mass, achieving a 90% success rate across 120 trials. Ablation studies confirm that each design choice contributes to throwing performance, and the framework requires no prior data or learned model, enabling direct deployment on new objects and targets without environment-specific calibration or data collection.

Problem

• pose-targeted throw-flip
• stable manipulation
• high-DoF manipulator
• release state
• object flipping

Innovation

• pose-targeted throw-flip
• release state planning
• high-DoF manipulation
• model-free robotic throwing
• robust motion design