Skip to content

asprilo (Robotic Intra-Logistics Benchmark Suite)

Description

A benchmark framework to study typical scenarios in intra-logistics and warehouse automation. It models complex multi-agent pathfinding (MAPF) and task assignment problems where mobile robots move shelves to picking stations.
The Scripts directory contains scripts to generate structured and random benchmark sets from the A,B,C and M domain using the asprilo generator (see below).

Usage

All asprilo tools can be installed from the asprilo repository.
The tools are also available via the potassco conda channel, but these are most likely outdated. To install the tools simply use or create a conda environment with python=>3.9. 

conda create -n <env> python=3.13
conda activate <env>
git clone https://github.com/potassco/asprilo.git
Install the generator: 
cd asprilo/generator
pip install .
gen -h
Install the visualizer: 
cd asprilo/visualizer
pip install .
viz -h
A detailed description on how to use these tools can be found here

Examples

Structured (real-world-like) instance:

  • 9x6 floor grid (-x 9 -y 6)
  • 3x2 shelf cluster dimensions (-X 3 -Y 2)
  • 6 shelves (-s 6)
  • 1 picking station (-p 1)
  • 2 robots (-r 2)
  • highway layout (-H)
  • 4 products (-P 4)
  • 20 product units (-u 20)
  • 4 orders (-o 4)
  • guarantee that all products are ordered at least once (--oap)
gen -x 9 -y 6 -X 3 -Y 2 -s 6 -p 1 -r 2 -H -P 4 -u 20 -o 4 --oap

Random instance:

  • 10x10 floor grid (-x 10 -y 10)
  • 50 shelves (-s 50)
  • 10 picking station (-p 10)
  • 30 robots (-r 30)
  • 10 products (-P 10)
  • 100 product units (-u 100)
  • 8 orders (-o 8)
  • 2 products per shelf (--prs 2)
  • 4 clingo threads (-t 4)
  • verbose output (-V)
  • incremental generation (-I)
gen -x 10 -y 10 -s 50 -p 10 -r 30 -P 10 -u 100 -o 8 --prs 2 -t 4 -V -I

Batch generation: 

gen --batch ./Scripts/batch/abc/structured.yml -I -t 4 -V

Generated instances can be found in the generatedInstances directory.

Encodings

The encodings are in three directories with regard to their problem domains:

  • ./abc contains encodings for asprilo domains A, B and C
  • ./m contains encodings for asprilo domain M
  • ./control contains encodings for supplementary features such as task assignment, highway constrains, etc. for all (some) asprilo domains

A detailed description of the encodings and their naming conventions can be found here

It is recommended to use encodings with the encoding* prefix, which provide a shorthand to call all necessary encodings. A time horizon should also be set, e.g.: 

clingo ./Encodings/m/encoding.lp <instance.lp> -c horizon=8

Metadata

Technical Details

  • Type: Optimization, Multi-shot, Planning
  • Format: ASP-Core-2, clingo (utilizes incremental solving and Python scripting)
  • Tested With: clingo >= 5.4.0

Instances

  • Generation: Automatic (via the integrated asprilo-generator)
  • Details:
    Includes structured and random sets (e.g., domains A, B, C, and M). Scale ranges from small 9x6 grids with 2 robots to large-scale warehouse layouts with dozens of robots and hundreds of shelf locations.

Data & Access (Confidentiality)

  • Status: Public
  • License: MIT
  • Sensitivity: None (Synthetic/Generated data)

Source & Literature

Contact

Miscellaneous

  • Complexity: NP-Hard
  • Metadata:
    Comprehensive documentation, a solution checker, and a visualizer are available at https://asprilo.github.io/.
  • Other Notes:
    The benchmark is designed to be highly modular, supporting various sub-problems like movement-only (M), shelf-to-station assignment (A), and full warehouse scenarios.

Download

In the future you will be able to download the benchmark set here.