Abstract
Traditional robotic grasp metrics prioritize force closure on rigid bodies, often neglecting the structural integrity of fragile objects. We present a Stress-Minimizing (SM) metric, $Q_{SM}$, which incorporates brittle fracture theory and material geometry to identify grasps that minimize internal stress. By utilizing the Boundary Element Method (BEM) and a progressive constraint detection algorithm, our approach achieves computational efficiency comparable to traditional metrics while significantly improving grasp safety for delicate items.
Introduction
Robotic manipulation has historically relied on the assumption that objects are infinitely stiff. Metrics like the widely used $Q_1$ evaluate grasp quality based on the ability to resist external wrenches through force closure. While effective for robust industrial parts, this "material-blind" approach fails when handling fragile objects where excessive contact force or poor placement can lead to structural failure. We propose a shift toward stress-aware grasping by introducing a metric that considers the object's internal mechanical response.
The Stress-Minimizing Metric
The core of our contribution is $Q_{SM}$, a metric derived from the maximum tensile stress within an object under a given grasp. To calculate this efficiently, we employ the Boundary Element Method (BEM). Unlike the Finite Element Method (FEM), which requires complex volumetric meshing, BEM only requires a surface triangle mesh, making it highly compatible with standard robotic vision outputs.
To maintain real-time feasibility, we leverage the sparsity of stress "hot spots." A progressive algorithm identifies active constraints during optimization, focusing computational resources only on areas prone to fracture. This allows the system to ignore the vast majority of the geometry where stress is negligible.
Experimental Evaluation
We evaluated $Q_{SM}$ against the traditional $Q_1$ metric using several benchmarks, including the Stanford Bunny and a U-shaped tuning fork. In the bunny test, $Q_1$ frequently suggested grasping the ears—a high-leverage, fragile point that satisfies force closure but risks breakage. In contrast, $Q_{SM}$ correctly identified the ears as high-stress zones and prioritized the sturdier base of the head.
Computational Efficiency
A critical concern for physics-based metrics is the overhead. By moving heavy BEM matrix assembly to a precomputation step, the online planning phase remains efficient.
As shown in the figure above, the evaluation time for $Q_{SM}$ is remarkably similar to $Q_1$. Global grasp planning for 100 potential points took roughly 1.7 hours for $Q_{SM}$ compared to 0.6 hours for $Q_1$, a manageable difference for offline planning scenarios.
Conclusion
The $Q_{SM}$ metric provides a robust framework for grasping fragile objects by integrating material physics into the planning loop. It demonstrates that shape-awareness is key to safe manipulation. Future work will focus on handling non-watertight meshes from real-world sensors and extending the model to multi-material objects.
Summary updated