Project design and usage

This section explains how EnactiveAgents can be used. It will describe in detail how the various modules can be used and adapted to specific needs.

High-level architecture

First it is useful to know the high-level design of the project. The project is split into four main packages: model, view, controller, and experiment. As their names imply, the model, view, and controller packackes are designed with a model-view-controller (MVC) pattern. In essence this simply means that the model logic is kept separate from the means of manipulating the experiment and the means of displaying the experiment to the user.

Model

The model package contains modules implementing the entities available (structures, agents), the various agent components within the enactivist cognitive architecture, and additionally implements the means for agents to interact with a world. The world model implements basic world logic, such as whether two entities are at the same location.

View

The view package contains modules to represent the model in various ways. For example, the main view module (view.py) is used to render the model to a graphical representation on screen.

Controller

The controller package contains modules implementing the ways with which a user can interact with the experiment. Currently, it only impements keyboard controls.

Experiment

The experiment package contains modules to set up various experiments using the existing model. So, the model package implements the agents and the basic world logic, and the experiment package is used to set up experiments using this model. An experiment defines a world layout, defines the agents that interact in the world, and uses the basic world logic to define the interaction logic. Additionally, an experiment can define when the simulation should automatically end, and can calculate metrics at each simulation tick. For example, an experiment can define a “step” and “bump” interaction, and can then define whether an agent that is attempting to step successfully steps or bumps depending on whether there is a structure in front of the agent.

Usage

• Creating experiments
  • What do experiments consist of?
  • Defining the world
    • Automatic world generation
    • Defining agents
    • Defining primitive interactions
    • Defining agent-world interaction logic
      • Reusable agent-world interaction logic
    • Defining complex agent-world interaction logic
  • Automated experimentation
  • Automatically stopping simulations
  • Keeping track of metrics
• Running simulations
• Interactive simulation control
  • Experiment control
• Agent vision
  • Perception handler
• Agent boredom
• Programmable agents
• Custom structures