Skip to content

Latest commit

 

History

History
52 lines (45 loc) · 5.91 KB

model-parameters.md

File metadata and controls

52 lines (45 loc) · 5.91 KB
Raw

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

Model Parameters

Changing Parameters in the Web UI

The web UI presented when running mesa runserver or 'run.py' allows you to change some parameters of the model using interactive elements such as drop-down selections and sliders. For these changes to take effect, you must restart the simulation by clicking the 'Reset' button in the menu bar. Some parameters can only be changed by modifying code directly in 'server.py', or 'BatchRunner_RR_and_SSI.py', for effecting the web-connected simulation, or headless batches of simulations respectively.

The following parameters are exposed via the web UI:

Parameter Description Default Value
Maximum Time Steps for Sampling (max_steps) Sets a time step at which to stop the simulation and save performance metrics. 0 (run indefinitely)
Number of Robots (num_robots) The number of robots to be deployed for the MRS. 3
Task Allocation Algorithm (task_allocation) The algorithm to be used to allocate sampling tasks to robots. Can be "Sequential Single Item (SSI) auction" or "Round Robin". "Sequential Single Item (SSI) auction"
Sampling Strategy (sampling_strategy) The strategy to be used to select locations for sampling. Can be "Random" or "Dynamic". The dynamic sampling methodlogy is described in the TAROS '22 paper. "Dynamic"

Changing Parameters in Code

The above parameters accessible via the web UI can also be changed within the 'server.py', or 'BatchRunner_RR_and_SSI.py' scripts. In both scripts, the values of model parameters, including some not exposed by the web UI, are defined within a dictionary. In server.py this dictionary is called model_params. An example of defining its values can be seen in server.py:

model_params = {
    "max_steps": max_steps_slider,
    "height": height,
    "width": width,
    "num_robots": num_robots_slider,
    "vis_freq": 5,
    "task_allocation": task_allocation_choice,
    "sampling_strategy": sampling_choice
}

In BatchRunner_RR_and_SSI.py parameters that are constant across a batch of trials are set in the fixed_parameters argument of instances of BatchRunnerMP.

The following parameters can only be changed within the code:

Parameter Description Default Value
Height (height) Defines the height (in cells) of the grid-world. In server.py this is set to the height of the data provided for the robots to sample. In BatchRunner_RR_and_SSI.py you will need to manually set this to the height of your data, if sampling custom data. 20
Width (width) Defines the width (in cells) of the grid-world. In server.py this is set to the width of the data provided for the robots to sample. In BatchRunner_RR_and_SSI.py you will need to manually set this to the width of your data, if sampling custom data. 20
Visualisation Frequency (vis_freq) The frequency, in simulation time steps, at which to generate visual outputs. Lower frequencies can significantly degrade the simulation's performance. 5
Results Directory (results_dir) The path to a directory in which data and visual outputs will be saved. "./results/default"
Verbose (verbose) If True, the simulation prints additional information to the terminal. True