Engine States

In this section we will discuss the concept of an engine state. We will do so by going through the process of creating an engine state for the OdeEngine. The OdeEngine allows to simulate systems based on ordinary differential equations (ODEs).

For the OdeEngine we will create two engine states, i.e. the OdeEngineState and OdeParameters engine states. These engine states will allow to reset the state of objects and reset the parameters for the ODE integration, respectively.

Full code is available here.

In this section we will discuss the concept of an EngineState. In an engine state, we create an implementation of a state for a specific Engine.

OdeEngineState

The first engine state will will create is the OdeEngineState. This engine state will be responsible for resetting the states of objects in the OdeEngine during a reset of the environment. Engine states can be created using the EngineState base class. For creating an engine node, we need to implement three abstract methods:

• spec()

• initialize()

• reset()

spec

So, first we will implement the spec method. This method allows to specify default parameters, but also to add custom parameters. In our case, we do not need to specify parameters, so the implementation is fairly simple:

import numpy as np
from eagerx.core.entities import EngineState
import eagerx.core.register as register


class OdeEngineState(EngineState):
  @staticmethod
  @register.spec("OdeSimState", EngineState)
  def spec(spec):
      pass

Note

Mind the usage of the spec() decorator. This decorator is required to register the OdeEngineState. All entities within EAGERx have to be registered, such that their specification can be created based on their unique id. In this decorator we provide a unique id for the engine state (“OdeSimState”) and specify the type (EngineState).

initialize

The initialize() method allows to initialize the engine state. In our case, the only thing we need to do during initialization is to store the object name.

def initialize(self):
  self.obj_name = self.config["name"]

Note

Note that we have access to the config attribute. See config for more information.

reset

Finally, we will implement the reset() method. This method will be called during a reset and will reset the state of the object.

def reset(self, state, done):
  self.simulator[self.obj_name]["state"] = np.squeeze(state.data)

Note

Note that we have access to the simulator attribute, which is created in the OdeEngine class.

Similarly, we can create the OdeParameters EngineState by implementing the spec(), initialize() and reset() abstract methods.