Malfunctions

Stochastic events are common in railway networks. The initial plan often needs to be rescheduled during operations as minor events such as delayed departure from train stations, various malfunctions on trains or infrastructure, or even problematic weather lead to delayed trains.

Malfunctions are implemented using a Poisson process to simulate delays by stopping agents at random times for random durations. Train that malfunction can’t move for a random, but known, number of steps. They of course block the trains following them 😬

The parameters necessary for the stochastic events are provided as a NamedTuple called MalfunctionParameters:

malfunction_parameters = (
    malfunction_rate=env_params.malfunction_rate,
    min_duration=20,
    max_duration=50
)

The parameters are as follows:

  • malfunction_rate is the mean rate of the poisson process in number of environment steps.

  • min_duration and max_duration set the range of malfunction durations. They are sampled uniformly.

You can then introduce stochasticity in an environment by using the malfunction_generator_and_process_data parameter of the RailEnv constructor:

RailEnv(
    ...
    malfunction_generator_and_process_data=malfunction_from_params(malfunction_parameters),
    ...
)

In your controller, you can then check whether an agent is malfunctioning:

obs, rew, done, info = env.step(actions)
...
action_dict = dict()
for a in range(env.get_num_agents()):
    if info['malfunction'][a] > 0:
        # agent is malfunctioning and can't move!
        # info['malfunction'][a] contains the number of steps this agent will still be blocked

You will quickly realize that this will lead to unforeseen difficulties which means that your controller needs to observe the environment at all times to be able to react to the stochastic events!

Check out the starter kit for a complete example of how to train a model using malfunctions.