This library does 3D rectangular bin packing; it attempts to match a set of 3D items to one or more in a set of 3D containers. The result can be constrained to a maximum number of containers.
Projects using this library will benefit from:
- short and predictable calculation time,
- fairly good use of container space,
- brute-force support for low number of boxes (ideal for small orders)
Bugs, feature suggestions and help requests can be filed with the issue-tracker.
The project is implemented in Java and built using Maven. The project is available on the central Maven repository.
Maven coordinates
Add
<3d-bin-container-packing.version>3.0.1</3d-bin-container-packing.version>
and
<dependency>
<groupId>com.github.skjolber.3d-bin-container-packing</groupId>
<artifactId>core</artifactId>
<version>${3d-bin-container-packing.version}</version>
</dependency>
or
Gradle coordinates
For
ext {
containerBinPackingVersion = '3.0.1'
}
add
api("com.github.skjolber.3d-bin-container-packing:core:${containerBinPackingVersion}")
Java 11+ projects please use module com.github.skjolber.packing.core
.
The units of measure is out-of-scope, be they cm, mm or inches.
Obtain a Packager
instance, then then compose your container and product list:
List<StackableItem> products = new ArrayList<StackableItem>();
products.add(new StackableItem(Box.newBuilder().withId("Foot").withSize(6, 10, 2).withRotate3D().withWeight(25).build(), 1));
products.add(new StackableItem(Box.newBuilder().withId("Leg").withSize(4, 10, 1).withRotate3D().withWeight(25).build(), 1));
products.add(new StackableItem(Box.newBuilder().withId("Arm").withSize(4, 10, 2).withRotate3D().withWeight(50).build(), 1));
// add a single container type
Container container = Container.newBuilder()
.withDescription("1")
.withSize(10, 10, 3)
.withEmptyWeight(1)
.withMaxLoadWeight(100)
.build();
// with unlimited number of containers available
List<ContainerItem> containerItems = ContainerItem
.newListBuilder()
.withContainer(container)
.build();
Pack all in a single container:
PackagerResult result = packager
.newResultBuilder()
.withContainers(containerItems)
.withStackables(products)
.build();
if(result.isSuccess()) {
Container match = result.get(0);
// ...
}
Pack all in a maximum number of containers:
int maxContainers = ...; // maximum number of containers which can be used
PackagerResult result = packager
.newResultBuilder()
.withContainers(containerItems)
.withStackables(products)
.withMaxContainerCount(maxContainers)
.build();
Note that all packager
instances are thread-safe.
A simple packager
PlainPackager packager = PlainPackager
.newBuilder()
.build();
A packager using the LAFF algorithm
LargestAreaFitFirstPackager packager = LargestAreaFitFirstPackager
.newBuilder()
.build();
For a low number of packages (like <= 6) the brute force packager might be a good fit.
Packager packager = BruteForcePackager
.newBuilder()
.build();
Using a deadline is recommended whenever brute-forcing in a real-time application.
Algorithm details
The implementation is based on this paper, and is not a traditional bin packing problem solver.
The box which covers the largest ground area of the container is placed first; its height becomes the level height. Boxes which fill the full remaining height take priority. Subsequent boxes are stacked in the remaining space in at the same level, the boxes with the greatest volume first. If box height is lower than level height, the algorithm attempts to place some there as well.
When no more boxes fit in a level, the level is incremented and the process repeated. Boxes are rotated, containers not.
LargestAreaFitFirstPackager
stacks in 3D within each levelFastLargestAreaFitFirstPackager
stacks in 2D within each level
The algorithm runs reasonably fast, usually in milliseconds. Some customization is possible.
This algorithm selects the box with the biggest volume, fitting it where it is best supported.
This algorithm has no logic for selecting the best box or rotation; running through all permutations, for each permutation all rotations:
BruteForcePackager
attempts all box orders, rotations and placement positions.FastLargestAreaFitFirstPackager
selects all box orders and rotations, selecting the most appropriate placement position.
The complexity of this approach is exponential, and thus there is a limit to the feasible number of boxes which can be packaged within a reasonable time. However, for real-life applications, a healthy part of for example online shopping orders are within its grasp.
The worst case complexity can be estimated using the DefaultPermutationRotationIterator
before packaging is attempted.
The algorithm tries to skip combinations which will obviously not yield a (better) result:
- permutations
- two or more boxes have the same dimensions
- permutations which mutated at a previously unreachable index
- fewer rotations
- two or more sides have the same length
- rotations which mutated at a previously unreachable index
There is also a parallel version ParallelBruteForcePackager
of the brute-force packager, for those wishing to use it on a multi-core system.
Note that the algorithm is recursive on the number of boxes, so do not attempt this with many boxes (it will likely not complete in time anyhow).
There is a simple output visualizer included in this project, based of three.js. This visualizer is currently intended as a tool for developing better algorithms (not as stacking instructions).
To use the visualizer during development, make your unit tests write directly to a file in the project (see VisualizationTest
example).
The code has been structured so it is possible to extend and adapt to specialized needs. See AbstractPackager
class, the extreme-points
and test
artifacts.
If you have any questions, comments or improvement suggestions, please file an issue or submit a pull-request.
Note on bugs: Please follow shuairan's example and file a test case with a visualization.
Feel free to connect with me on LinkedIn, see also my Github page.
Apache 2.0. Social media preview by pch.vector on www.freepik.com.
- 3.0.1: Various performance improvements.
- 3.0.0: Support max number of containers (i.e. per container type). Use builders from now on. Various optimizations.
- 2.1.4: Fix issue #574
- 2.1.3: Fix nullpointer
- 2.1.2: Tidy up, i.e. remove warnings, nuke some dependencies.
- 2.1.1: Improve free space calculation performance
- 2.1.0: Improve brute force iterators, respect deadlines in brute for packagers.