Basic API Usage

This tutorial explains the basics of using the EvoGym API. Completed code can be found here.

Page Outline

1. Creating an EvoWorld environment
2. Adding a robot
3. Creating a Simulation
4. Stepping the simulation

Creating an EvoWorld environment

Create an EvoWorld object by loading a simple environment created with the Evolution Gym Design Tool.

from evogym import EvoWorld, EvoSim, EvoViewer, sample_robot
world = EvoWorld.from_json(os.path.join('world_data', 'simple_environment.json'))

The best way to visualize EvoWorld environments is within the Evolution Gym Design Tool. However, we can also visualize them with this handy function.

world.pretty_print()

Adding a robot

Let’s create a randomly sampled 5x5 robot.

robot_structure, robot_connections = sample_robot((5, 5))

We can add the robot to our existing EvoWorld environment and specify its initialization position in 2D space.

world.add_from_array(
	name='robot',
	structure=robot_structure,
	x=3,
	y=1,
	connections=robot_connections
)

Creating a Simulation

We can create a simulation using from our EvoWorld environment. This will initialize the simulation with the locations of all objects.

sim = EvoSim(world)
sim.reset()

A viewer object will allow us to visualize our simulation. We’ve set the viewer to track both the robot and box objects in the simulation, so it will always keep them in frame.

viewer = EvoViewer(sim)
viewer.track_objects('robot', 'box')

Stepping the simulation

To run the simulation, we use the following loop. In each iteration, we sample a random action vector for our simulation, step the simulation, and render it.

for i in range(500):
	sim.set_action(
		'robot',
		np.random.uniform(
			low = 0.6,
			high = 1.6,
			size=(sim.get_dim_action_space('robot'),)
		)
	)
	sim.step()
	viewer.render('screen')