constrained-mdp

The agent loop, embodied. Tool-using runtime + fail-closed safety pipeline + sim-first execution + statistically-rigorous bench harness + post-hoc analysis (counterfactual replay, causal attribution, LLM judge, property mining) for embodied AI. Sister project to GhostLM.

Safe RL for Autonomous Driving: A project using Constrained MDPs to train a vehicle agent for lane-centering, speed adherence, and collision avoidance. Features kinematic modeling, reward shaping, and safety constraints.

Implementation of a Hybrid Shielded-CMDP framework for safe smart grid voltage control

Reference simulator and stored results for Lyapunov-guided safe RL with risk-budget feasibility in energy-aware O-RAN scheduling: CVaR tail constraint, LCB safety filter, PID dual, and a swept classical drift-plus-penalty frontier.