11. Controllers

Sinergym includes a section for implementing your own controllers. At present, we’ve developed a random agent and a rule-based agent for 5Zone and Datacenter buildings. You can locate this code in sinergym/sinergym/utils/controllers.py. This is particularly useful for performing benchmarks as a reference point for studying DRL algorithms.

The functionality is straightforward; given an environment observation, these instances return an action to interact with the environment. You can create your own controllers or modify existing rules, for instance. An example of usage can be found in the section Rule Controller example.

Warning

You have to make sure that the variables used particularly for your controller are part of the observation space of the configured environment.