Data for "Auditing Single-Agent Reinforcement Learning for EV Charging Assignment: A Protocol-Amended Comparison of Trained, Untrained, and Heuristic Policies"

Abstract

Version 2 of the data deposit for the manuscript "Auditing Single-Agent Reinforcement Learning for EV Charging Assignment: A Protocol-Amended Comparison of Trained, Untrained, and Heuristic Policies" (Algorithms, MDPI, manuscript ID algorithms-4572374). Version 2 replaces the preliminary data-only content of version 1 (30 August 2026). It contains the per-episode run records underlying every table and figure of the revised manuscript and its Supplementary Material:- final analytical dataset, version 2 (240 run files, sealed on 2026-09-18), identified by scenario, condition (trained, untrained, baseline, MECT), phase and cohort, with its manifest and SHA-256 digests;- legacy benchmark dataset (210 runs);- run collections: PPO (40 jobs), R2-WAIT (84), isolated ablation arms (168), R3 quad-off (14), R5 prioritised experience replay (14) and Tangier sensitivity (28);- campaign specifications (manifests). The archive is organised as final_dataset_v2/, legacy_v3_full/, collections/ and campaign_manifests/, with a README.md and a global MANIFEST.sha256 (SHA-256 of every file). Per-timestep SUMO event logs, trained model weights, source code and scripts are not included; event logs and model weights are available from the corresponding author on request.