Learning Dynamic Non-Prehensile Object Reorientation via Reinforcement Learning

Abdullah Mustafa *
Ryo Hanai *
Ixchel Ramirez *
Floris Erich *
Ryoichi Nakajo *
Yukiyasu Domae *
Tetsuya Ogata *,†

IEEE/SICE International Symposium on System Integration (SII 2026)

*National Institute of Advanced Industrial Science and Technology, Waseda University

Abstract

This work proposes a learning-based approach to dynamic non-prehensile object reorientation, enabling fast reorientation of large, grasp-infeasible objects within a constrained task space using uni-manual manipulation. Our policy is trained in simulation via reinforcement learning, utilizing a carefully designed observation space, action space, and reward function to reorient randomly sized cuboids with varied physical properties. Given an object model and a target rotation direction, the policy plans offline trajectories suitable for both simulation and real-world deployment. Although the policy is sensitive to modeling uncertainties, accurate modeling enables successful sim-to-real transfer across different objects and rotation directions.

--------------- (ChocoBox) ---------------

Y-Axis|CCW|90 [7/10]
X-Axis|CCW|90 [10/10]
Z-Axis|CCW|90 [10/10]
Y-Axis|CCW|180 [10/10]
X-Axis|CCW|180 [10/10]
Z-Axis|CW|90 [10/10]

--------------- (ChocoBox-Rotated) ---------------

Y-Axis|CCW|90 [10/10]
X-Axis|CCW|90 [10/10]
Z-Axis|CCW|90 [10/10]
Y-Axis|CCW|180 [2/10]
X-Axis|CCW|180 [4/10]
Z-Axis|CW|90 [10/10]

--------------- (TissueBox) ---------------

Y-Axis|CCW|90 [7/10]
X-Axis|CCW|90 [6/10]
Z-Axis|CCW|90 [4/10]
Y-Axis|CCW|180 [10/10]
X-Axis|CCW|180 [9/10]
Z-Axis|CW|90 [9/10]

--------------- (WoodBlocks) ---------------

Y-Axis|CCW|90 [6/10]
X-Axis|CCW|90 [8/10]
Z-Axis|CCW|90 [1/10]
Y-Axis|CCW|180 [0/10]
X-Axis|CCW|180 [0/10]
Z-Axis|CW|90 [3/10]

--------------- (TeaBox-40g) ---------------

Y-Axis|CCW|90 [10/10]
Y-Axis|CCW|180 [5/10]
X-Axis|CCW|90 [1/10]

--------------- (TeaBox-200g) ---------------

Y-Axis|CCW|90 [8/10]
Y-Axis|CCW|180 [2/10]
X-Axis|CCW|90 [10/10]

BibTeX citation

    @misc{DNP_Reorientation_SII2026,
  author = "{Abdullah Mustafa and Ryo Hanai and Ixchel Ramirez and Floris Erich and Ryoichi Nakajo and Yukiyasu Domae and Tetsuya Ogata}",
  title = "Learning Dynamic Non-Prehensile Object Reorientation via Reinforcement Learning",
  year = "2026",
}