Subassembly Stability

Nico Boneschanscher, Hans van der Drift, Stephen J. Buckley, Russell H. TaylorAAAI 1988

Abstract

This paper introduces an algorithmic approach to determine the static stability of a subassembly during the manufacturing process. By modeling parts as rigid polyhedra and linearizing 3D friction cones into a 6D system of linear equations, the authors provide a way to verify if a partial assembly will remain upright under the influence of gravity and the force of the next part being inserted.

Introduction

In automated manufacturing, planning the sequence of assembly is a massive combinatorial challenge. A critical constraint often overlooked by early planners is physical stability: at every single step of the process, the current "pile" of parts must remain stable on the worktable. If adding a new part causes the subassembly to tip over or slide due to gravity or insertion pressure, the assembly plan is a failure.

An insertion force applied to a two-dimensional block on a table.

The authors tackle the difficulty of calculating this stability for complex, three-dimensional objects. Previous methods were either computationally unproven, limited to 2D, or unable to handle external forces, such as a robot arm pushing a part into place.

Methodology

The proposed multi-stage algorithmic pipeline consists of the following steps:

  1. Geometric Modeling: Parts are represented as rigid polyhedra, such as cuboids or cylinders.
  2. Friction Linearization: To make the math tractable, the circular "friction cone" (Coulomb friction) at contact points is approximated using a set of linear inequalities in six dimensions (3 for force, 3 for torque).
  3. Network Representation: The subassembly is treated as a graph where nodes represent parts and edges represent contact relations.
Internal contact forces between parts A and B within a subassembly.
  1. Network Reduction: This graph is simplified into a single relation between the subassembly and the table using two main operations:
  2. Merging: Combining parallel contacts between two parts (mathematically, a Minkowski sum of their friction cones).
  3. Compounding: Reducing a chain of parts (e.g., A → B → C) into a single relation (A → C) by intersecting their stable force sets.
  4. Stability Testing: The final reduced system of equations is processed using a Simplex algorithm to check if a valid set of reaction forces exists to counter gravity and insertion forces.

Key Insights

The paper distinguishes between potential stability (there exists a set of forces that could keep it stable) and guaranteed stability (it will be stable regardless of internal force distribution). The algorithm primarily computes potential stability.

Furthermore, the authors address the concept of limited superposition. While "full superposition" assumes all contact points contribute to stability simultaneously, limited superposition is more realistic, accounting for scenarios where a force might break contact at one point while maintaining it at another.

Results and Discussion

The algorithm's computational efficiency was demonstrated to have an average-case complexity of $O(mn^3)$ for an assembly with $m$ parts and $n$ vertices. This efficiency made it practical for real-time assembly planning on contemporary hardware. Experimental validation in the AML/X programming language showed the algorithm correctly identifying unstable configurations where top-heavy insertion forces would cause a stack to collapse.

Limitations

The current network reduction method struggles with circular dependencies, such as three parts leaning against each other in a triangle. Additionally, while potential stability is solvable in polynomial time, guaranteed stability with friction is generally NP-hard. The authors suggest that most industrial assemblies can be solved with efficient heuristics despite these theoretical complexities.

Summary updated

BibTeX
@inproceedings{boneschanscher1988subassembly,
  author = {Nico Boneschanscher and Hans van der Drift and Stephen J. Buckley and Russell H. Taylor},
  title = {{Subassembly Stability}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {1988},
}

Loading related content...

Figure