Pick-and-place Manipulation Across Grippers Without Retraining: A Learning-optimization Diffusion Policy Approach

2025-02-21

AI Summary

Existing robotic grasping and placement strategies rely on fixed gripper configurations; adapting to new end-effectors necessitates costly retraining—particularly limiting the generalizability of imitation learning approaches. This paper introduces the first diffusion-based policy framework that unifies learning and optimization, enabling zero-shot adaptation to unseen grippers. During the denoising process, it dynamically incorporates kinematic constraints, task-pose correction (TPC) offset modeling, adaptive gripper-width control, and collision-aware trajectory generation. Crucially, no new data collection or policy fine-tuning is required. The method achieves robust cross-gripper generalization across six heterogeneous grippers—including 3D-printed, silicone, and Robotiq models—on a Franka Panda platform, attaining a mean task success rate of 93.3%, substantially outperforming baselines (23.3–26.7%). To our knowledge, this is the first approach to enable generalizable, retraining-free grasping and placement control.

Technology Category

• Intelligent Robots: Manipulation
• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Current robotic pick-and-place policies typically require consistent gripper configurations across training and inference. This constraint imposes high retraining or fine-tuning costs, especially for imitation learning-based approaches, when adapting to new end-effectors. To mitigate this issue, we present a diffusion-based policy with a hybrid learning-optimization framework, enabling zero-shot adaptation to novel grippers without additional data collection for retraining policy. During training, the policy learns manipulation primitives from demonstrations collected using a base gripper. At inference, a diffusion-based optimization strategy dynamically enforces kinematic and safety constraints, ensuring that generated trajectories align with the physical properties of unseen grippers. This is achieved through a constrained denoising procedure that adapts trajectories to gripper-specific parameters (e.g., tool-center-point offsets, jaw widths) while preserving collision avoidance and task feasibility. We validate our method on a Franka Panda robot across six gripper configurations, including 3D-printed fingertips, flexible silicone gripper, and Robotiq 2F-85 gripper. Our approach achieves a 93.3% average task success rate across grippers (vs. 23.3-26.7% for diffusion policy baselines), supporting tool-center-point variations of 16-23.5 cm and jaw widths of 7.5-11.5 cm. The results demonstrate that constrained diffusion enables robust cross-gripper manipulation while maintaining the sample efficiency of imitation learning, eliminating the need for gripper-specific retraining. Video and code are available at https://github.com/yaoxt3/GADP.

Problem

• Zero-shot adaptation to novel grippers
• Eliminating gripper-specific retraining
• Ensuring kinematic and safety constraints

Innovation

• Diffusion-based policy adaptation
• Hybrid learning-optimization framework
• Zero-shot gripper adaptation