eta-ctrl

Abstract

A modular framework for rolling-horizon operational control using a Gym-compatible agent–environment abstraction. The framework enables integration of optimization-based agents (Pyomo), reinforcement learning agents (Stable-Baselines3), heuristic search methods, and rule-based controllers. Environments can be instantiated from FMI-compliant simulations, mathematical models, or real-world system interfaces. 0.4.0 (2026-09-09) The focus of this release is making run file layout configurable via the [paths] config section, along with the API cleanup that goes with it. ConfigRun becomes RunInfo and the prior hardcoded path values are now defined in ConfigPaths. The csv_export function plus the monitor-path helpers are now deprecated. Alongside this, BaseEnv can export its state_log as a pandas DataFrame, the Pyomo export script and documentation improve, and several bugs are fixed (MPC solver time limits, CSV date-parsing warnings, sim_steps_per_sample).A hotfix adds provisional live-connection retry logic to LiveEnv and routes eta_nexus/asyncua output into the run log, with further refinement to follow. CI gains parallel testing, automatic Docker builds, and local mypy.