Cost-Optimal Foundation Model Deployment Portfolio for Transportation Management

2026-07-14

AI Summary

This study addresses the challenge of deploying multitask foundation models in traffic management centers under limited shared GPU resources. The problem is formally cast, for the first time, as a combinatorial optimization problem incorporating constraints on model quality, latency, and security, and is proven to be NP-hard. To minimize total cost of ownership, the authors formulate a mixed-integer programming model and propose a polynomial-time greedy heuristic algorithm that efficiently yields near-optimal solutions. Experimental results demonstrate that the proposed method generates a hybrid deployment strategy costing only \$34 per month—97% lower than an all-closed-source API alternative—and enables quantitative identification of the breakeven point for local GPU investment.

Abstract

Foundation models, including large language models (LLMs) and vision-language models (VLMs), are increasingly used for transportation management center (TMC) tasks such as anomaly detection, incident reporting, and traveler information. Deploying multiple such models across TMC functions raises a portfolio question: which model should serve each function, in which deployment mode, and under what shared hardware budget? We formulate this as the Foundation Model Deployment Portfolio (FMDP) problem, a mixed-integer program minimizing total cost of ownership (TCO) subject to per-function quality, latency, and safety constraints over shared GPU capacity. We prove the problem NP-hard by reduction from the 0-1 knapsack problem and propose a polynomial-time greedy heuristic. In an illustrative case study with five TMC functions and 19 candidate (model, mode) pairs, FMDP identifies a mixed portfolio costing $34/mo (97% below the cheapest feasible all-closed-API baseline) by routing four functions to open-source APIs and the one function whose quality floor no open-source model meets to a closed API. Break-even analysis shows that on-premise GPU investment becomes reasonable only above approximately 309 vision queries/hour or if API prices double.

Problem

• Foundation Model Deployment
• Transportation Management
• Cost Optimization
• GPU Resource Allocation
• Model Portfolio Selection

Innovation

• Foundation Model Deployment Portfolio
• Total Cost of Ownership
• Mixed-Integer Programming
• Greedy Heuristic
• Transportation Management Center