Home | Accessing Simulation Outputs | Model Parameters | Sampling Custom Data

Spatial Sampling Multi-Robot System Simulation

A 2D grid-based simulation of a spatial sampling Multi-Robot System (MRS). Created using the MESA agent-based modelling framework for Python. This simulation is developed as a platform for simulated experiments to compare the performance of different combinations of task allocation and sampling methods for multi-robot mapping of spatial data. Its use in modelling a soil compaction sampling MRS is described in detail in Roberts-Elliott et al.'s TAROS 2022 conference paper: 'Agent-Based Simulation of Multi-Robot Soil Compaction Mapping'. Its abstract representation of a MRS allows many experiments to be conducted quickly, with low computational complexity. It also includes interactive visualisation with easy control of parameters. By default, the simulated MRS samples surface soil compaction data, recorded using an outdoor Thorvald robot at the University of Lincoln's Riseholme campus by Jaime Fentanes et al. Further details of their dataset and their single-robot work on soil compaction sampling can be found in '3D Soil Compaction Mapping through Kriging-based Exploration with a Mobile Robot'. This data is measured in Kpa, and is interpolated by this simulation using kriging, so that samples can be taken from ground-truth data at arbitrary locations.

You can provide the simulation with your own geolocated data for the robots to sample, following the instructions in Sampling Custom Data.

Dependencies

The simulation runs entirely in Python 3, requiring the following packages to be installed with: pip install mesa numpy astar-python PyKrige matplotlib. The simulation has been tested to work on Ubuntu 18.04, Windows 10, and Windows 11. It may work on other operating systems, but these have not been tested.

Running the simulation

To run the simulation, use the bash command mesa runserver with your working directory set to that of mesa_spatial_sampling_MRS. If the mesa command is not found, and you have installed mesa in a virtual environment, ensure that you have activated this environment before running this command.

Alternatively, you can run the Python script run.py.

If all has been set up correctly, running the simulation should open a web page in your default internet browser for visualisation. It should resemble this screenshot:

Instructions on using this grid visualisation web UI can be found in the Grid Visualisation section of MESA's advanced tutorial.

Running Batches of Simulated Experiments

An example script for running headless batches of experiments is provided in BatchRunner_RR_and_SSI.py. This script reproduces the simulated trials detailed in Roberts-Elliott et al.'s TAROS 2022 paper 'Agent-Based Simulation of Multi-Robot Soil Compaction Mapping'. The automation of batches of experiments is powered by Mesa's BatchRunnerMP class. Although this supports multiprocessing, the script defaults to running trials sequentially on a single core, as some features of the simulated MRS use multiprocessing, e.g. PyKrige kriging interpolation. The Accessing Simulation Outputs section of this documentation details how to access the data and figures generated from batches of simulated experiments.