Wrapper¶

BOLeRo is able to wrap C++ implementations of all components easily with the wrappers that are provided in this package.

The class CppBLLoader provides an interface to load C++ components. It has the functions

• acquire_behavior()
• acquire_behavior_search()
• acquire_contextual_environment()
• acquire_environment()
• acquire_optimizer()

that require a library name as an argument and return a wrapped C++ object. These libraries have to be registered previously with CppBLLoader.load_config_file() which takes the name of a configuration file as an argument. The configuration file contains the name or location of one library per line, for example

pso_optimizer
mountain_car

C++ components usually need to be configured during acquisition. The configuration is provided via configuration files. As a convention, the configuration file learning_config.yml is used for this purpose, however, that depends on the implementation of the component. Another convention is that, for example an environment only uses the section Environment from the configuration file, an optimizer uses the section Optimizer, etc. For example, the following configuration file configures a MARS-based environment.

Environment:
    calc_ms: 10
    stepTimeMs: 10
    graphicsUpdateTime: 10
    velocityControl: True
    enableGUI: True

We can load an environment in Python with

bll = CppBLLoader()
bll.load_config_file(LIBRARY_CONFIG_FILE)
env = bll.acquire_environment("mountain_car")