Agentic-Aggregator: replication package for "Agentic Aggregation for Electric Bus Depot Coordination: Operational Trade-Offs, Contextual Pricing, and Deployment Implications"

Résumé fourni par la source

Replication package for the manuscript "Agentic Aggregation for Electric Bus Depot Coordination: Operational Trade-Offs, Contextual Pricing, and Deployment Implications" (Transportation Research Part C, manuscript TRC-26-02380, under review). This archive is a snapshot of the public repository https://github.com/AlienEslami/Agentic-Aggregator at commit 1eb9d9f. It contains: the self-contained Python package implementing the day-ahead and real-time agentic workflows (Trigger, Pricing, and Evaluator agents as typed structured-output calls, with deterministic rule-based counterparts implementing the same interfaces); the mixed-integer optimization core (Pyomo/Gurobi); the frozen experiment protocols and per-run provenance manifests; the complete experiment inputs, including the day-ahead case-study workbooks and the real-time operational workbooks (48-timestep state and intraday-price series); the published result matrices behind every table of the revised manuscript, including the component ablation, the matched non-agentic baseline, the revenue-matched two-stage stochastic recourse benchmark, the solver-tolerance check, the prompt and threshold sensitivity study, the extended disturbance patterns, the chained multi-day experiment, and the fleet-scaling study; and the frozen day-ahead prompt-paradigm texts. Every deterministic experiment replicates from this archive alone; re-running the agentic arms additionally requires access to the hosted language model identified in the run manifests (gpt-5.6-luna). Funding: Government of Canada (Environmental Damages Fund, Climate Action and Awareness Fund) and Fonds de recherche du Québec – Nature et technologies (FRQNT), Programme de recherche en partenariat – Réduction des GES – Mobilité Durable.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.