User guide

1. Try with demo

Start server with server/app.py

Execute cells in client/demo.ipynb

2. The possibilities

• Create agents with different characteristics
  • radius
  • texture: color or spg compatible texture
  • initial_coordinates: ((x, y), orientation) with x, y between 0 and 1
  • agent_type: string usefull for behaviors
• Create the playground
  • size: (width, height)
  • interactions: list of interactions between agents and environnement. Only one is implemented: BIG_EAT_SMALL if an agent is hit by a larger one it is instantly moved to another random position.
• Start (and stop) simulator with playground and agents
  The simulator_interface supplies two functions:
  • open_session(nb_agents, agents, playground_params)
    This function open simple_playground with characteristics defined in agents and playground_params
    If nb_agents is greater than the number of agents then default agents will be created and return a Simulator instance
  • close_session(simulator)
    Close the simple_playground window linked to the simulator instance
• Create and attach behaviors
  • Agents behaviors Agent class has 3 functions:
    • attach_behavior(behavior_func, freq)
    • start_behavior(behavior_func)
    • stop_behavior(behavior_func)
      The behavior function has one parameter the agent and must return 2 float between 0 and 1 respectively for the activation of right and left wheel
  • Sphere apparitions
    It is possible to make objects appear randomly on the field at a certain frequency. These objects can be aborbed by the agents or not. To do it you can use the start_sphere_apparition function from the simulator instance return from open_session with 3 parameters: period, eatable and radius. You can stop this behavior with the method stop_sphere_apparition

3. Structure

Server

• app.py contains the Flask API.
• services.py contains the model. Manage the simulator. (used by app.py)
• interactions.py contains some function transmitted to simple_playground to managed interactions between element of the playgrounds. Currently there is only one interaction: BIG_EAT_SMALL. (used by services.py)
• controllers/remote_controller.py is an extensions of simple_playgrounds to control agents. (used by services.py)

Initialy the folders was created in anticipation of the implementation of other controllers.

Client
Functions and class are the same or almost the same as the one used in Pyvrep-epuck

• agents.pycontains Agent class,
• simulator.py contains Simulator class,
• behaviors.py contains Routine, Behavior, BehaviorMixer and RoutineManager class,
• simulator_interface.py supplies functions to start and stop simulator.
• demo.ipynb is an example of what it's possible to do.

4. Tests

A part of the server is tested with pytest (test/test_services.py):

• start / stop simulator
• add agents
• change speed agents
• attach and retrieve agents type
• consistent sensor values according agents positions

Installation

Library and dependencies via poetry

poetry install