SpringGrasp: Synthesizing Compliant, Dexterous Grasps under Shape Uncertainty

TL;DR

SpringGrasp addresses dexterous grasping under object shape uncertainty by optimizing a compliant, dynamic grasp using a differentiable SpringGrasp metric and a GPIS-based surface model. It jointly optimizes a pregrasp pose and per-finger impedance gains to drive the object toward a force-closure equilibrium, with uncertainty-aware terms and collision-avoidance constraints. Real-robot experiments show high grasp success (89% from two views, 84% from one) and consistent improvements over force-closure baselines by 18–27%, highlighting robustness to partial observations. The approach enables uncertainty-aware grasp planning, adapts contact forces and motion through a Wahba-based equilibrium, and offers practical implications for dexterous manipulation with imperfect perception.

Abstract

Generating stable and robust grasps on arbitrary objects is critical for dexterous robotic hands, marking a significant step towards advanced dexterous manipulation. Previous studies have mostly focused on improving differentiable grasping metrics with the assumption of precisely known object geometry. However, shape uncertainty is ubiquitous due to noisy and partial shape observations, which introduce challenges in grasp planning. We propose, SpringGrasp planner, a planner that considers uncertain observations of the object surface for synthesizing compliant dexterous grasps. A compliant dexterous grasp could minimize the effect of unexpected contact with the object, leading to more stable grasp with shape-uncertain objects. We introduce an analytical and differentiable metric, SpringGrasp metric, that evaluates the dynamic behavior of the entire compliant grasping process. Planning with SpringGrasp planner, our method achieves a grasp success rate of 89% from two viewpoints and 84% from a single viewpoints in experiment with a real robot on 14 common objects. Compared with a force-closure based planner, our method achieves at least 18% higher grasp success rate.

Figures (17)

• Figure 1: Compliant grasp as virtual spring attached from finger to object
• Figure 2: Process for generating grasps from a partial point cloud.
• Figure 3: A compliant grasp $\mathcal{G}$ on a 2D triangle at $t_0$ and at $t_\text{eq}$. The black solid line shows initial fingertip positions and the object pose, the blue dashed line shows fingertip positions and the object pose at equilibrium. $\bm{p}_i$ is the fingertip contact position and $\bm{o}_i$ is the target position.
• Figure 4: Shape of p.d.f. for different line trajectory, we prefer (c) instead of (a) and (b) as contact is mostly likely to happen at the expected time and location.
• Figure 5: Real robot setup showing the Kuka iiwa arm equipped with a left Allegro hand and the configuration of the three RGB-D cameras.