dpct-env 0.0.3

Environment-specific plugins for the DPCT framework

dpct-env

Environment-specific plugins for the DPCT framework.

This package provides:

• Fitness functions — class-based fitness computation for specific environments
• Results processors — Comet ML logging for environment-specific metrics
• Environment evaluations — documentation on environment suitability for DPCT

Installation

pip install dpct-env

Usage

Fitness Functions

from dpct_env.fitness import get_fitness_functions, ControlEffortFitness

# Get all fitness functions registered for CartPole-v1
fns = get_fitness_functions("CartPole-v1")

# Compute a fitness value
fitness = fns["stability_ratio"](observation_history=obs, action_history=acts)

# Use a fitness function directly
effort_fn = ControlEffortFitness(mode="changes")
fitness = effort_fn(observation_history=obs, action_history=acts)

Results Processors

from dpct_env.processors import create_environment_results_processor

processor = create_environment_results_processor("CartPole-v1")
processor.log_results(experiment=exp, individual=ind, history=history)

Registered Environments

Environment	Fitness Functions
CartPole-v1	control_effort, observation_variance, stability_ratio
MountainCar-v0	control_effort, height_achieved
Acrobot-v1	control_effort, height_achieved
Pendulum-v1	control_effort, observation_variance

Architecture

dpct-env  ──depends on──▶  dpct
     │                        │
     └── provides plugins ───┘

dpct-env is a pure plugin package — it imports from dpct but dpct never imports from dpct-env. This keeps the core framework clean while allowing environment-specific functionality to evolve independently.