Abstract
Packing complex mechanical objects into containers is a significant challenge in logistics due to their sprawling geometries and internal joints. This paper introduces a "disassemble-and-pack" strategy that treats a mechanism's joints as optimization parameters. By hierarchically disassembling the object and folding sub-assemblies into their most compact forms, the system achieves superior space utilization compared to traditional methods.
Introduction
The 3D bin packing problem is a classic NP-complete computational challenge. For mechanical objects like robot arms or engines, the difficulty is compounded by awkward shapes that create substantial "dead air" when packed as static units. While previous methods have explored splitting objects, they often ignore the logical structure of mechanisms. This research proposes using the mechanism's own joints to break it down and fold it into a minimal volume for efficient storage and shipping.
Methodology
The proposed system follows a two-stage pipeline: Hierarchical Disassembly and Mechanical Group Packing.
Hierarchical Disassembly
The mechanism is modeled as a set of rigid bodies connected by various joint types (e.g., revolute, gear-to-gear). The algorithm generates a hierarchy of potential disassembly points. Crucially, for each sub-assembly, the system calculates the degrees of freedom allowed by the remaining joints and adjusts the parts to find the Minimum Oriented Bounding Box (MOBB). This "folding" step ensures each group is as compact as possible before packing.
Mechanical Group Packing
Once the object is decomposed into optimized groups, they are placed into the container using a heuristic search:
- Sorting: Groups are ordered by their maximum extension.
- Roto-translation: Quaternions and vectors are used to determine the optimal orientation and position.
- Hole-Filling: The algorithm actively searches for empty volumes (holes) created by previously placed parts to maximize density. If no suitable hole is found, it places the part to minimize the total height increase.
Experimental Results
The approach was evaluated on four representative mechanical models: a Robot-arm, a Crank-block, a Motor, and an Excavator.
The results demonstrate that the algorithm successfully condenses sprawling machinery into compact rectangular footprints. By exploiting joint disconnections and internal motion, the method produces group sets with minimal total volume, leading to tighter packing than static heuristics.
Discussion and Future Work
A key insight of this work is treating joints as variables rather than constraints. This allows the algorithm to find "good enough" solutions efficiently through breadth-first search. However, limitations remain. The current model assumes all parts are rigid, ignoring flexible components like hoses. Additionally, while qualitative results are strong, further quantitative benchmarking against industrial state-of-the-art heuristics is needed. Future research could integrate machine learning to predict optimal disassembly points, further accelerating the process for highly complex assemblies.
Summary updated