ISOMORPH: A Supply Chain Digital Twin for Simulation, Dataset Generation, and Forecasting Benchmarks

2026-05-14 · Source: stat.ML updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Emerging Technologies & Innovation · Depth: Expert, extended

Summary

Isomorph is an open-source digital twin of a multi-echelon logistics network designed to generate time-series forecasting (TSF) benchmarks. It addresses the lack of comprehensive supply chain datasets by simulating a 13-node U.S. network with three sources, nine intermediate warehouses, and a destination, supporting two catalogue scales ($C=50$ and $C=200$). The simulator models network dynamics as a Markov chain, tracking on-hand inventory, outstanding orders, in-transit shipments, and smoothed demand. It reproduces the bullwhip effect at empirically consistent magnitudes and structurally encodes three conservation laws for verification. The project releases datasets from baseline configurations, 30 scenario rollouts, and 20 Latin-hypercube perturbations. Zero-shot evaluation of four foundation models (Chronos, Moirai, TimesFM, Lag-Llama) on Isomorph data shows MASE values exceeding public GIFT-Eval references at horizons $h\geq 14$, indicating new challenges. The digital twin also enables forward uncertainty quantification (UQ) from parameter uncertainty, with foundation models serving as fast surrogates.

Key takeaway

For AI Scientists and Machine Learning Engineers developing forecasting solutions for supply chain logistics, Isomorph provides a critical resource. Your existing TSF foundation models may struggle with the complex, coupled dynamics present in Isomorph's datasets at longer horizons, suggesting a need for fine-tuning or incorporating this domain into your training. Leverage Isomorph's configurable parameters and conservation laws to generate tailored datasets, validate new control policies, and perform robust uncertainty quantification for your models.

Key insights

Isomorph is a digital twin generating supply chain TSF benchmarks with complex dynamics and parameter-driven uncertainty quantification.

Principles

• Digital twins can generate diverse, interpretable TSF datasets.
• Markov chain modeling ensures structural integrity and conservation laws.
• Parameter uncertainty can be propagated via zero-shot foundation models.

Method

Isomorph simulates a multi-echelon logistics network using a seven-stage per-step loop, incorporating Poisson demand, (s,S) replenishment, Dijkstra routing, and greedy first-fit packing to advance a high-dimensional state vector.

In practice

• Use Isomorph datasets to benchmark TSF models for logistics.
• Apply conservation laws to verify custom control rule implementations.
• Generate forecast confidence bands by perturbing demand-side parameters.

Topics

• ISOMORPH
• Supply Chain Digital Twin
• Time-Series Forecasting Benchmarks
• Multi-Echelon Logistics Networks
• Markov Chain Modeling

Code references

• tuhinsahai/ISOMORPH
• thuml/Time-Series-Library

Best for: AI Engineer, AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Learning to replenish: A hybrid deep reinforcement learning for dynamic inventory management in the pharmaceutical supply chains — A hybrid deep reinforcement learning approach (A3C DPPO) effectively optimizes pharmaceutical inventory replenishment under stochastic demand and variable lead times.
• TRUST-SCF: Transformer-based Risk Understanding and Scoring for Transactional Supply Chain Finance — TRUST-SCF uses transformers with financial context to predict transaction-level risk and generate dynamic credit scores without explicit labels.
• How Lenovo Built an AI-Powered Supply Chain — Effective AI in supply chain management fundamentally relies on a high-quality, integrated data foundation, not just advanced technology.
• Data-Driven Open-Loop Simulation for Digital-Twin Operator Decision Support in Wastewater Treatment — CCSS-RS provides a robust, data-driven simulator for WWTPs, enabling long-horizon scenario screening despite challenging data conditions.
• Designing Supply Chains for Volatility - with Dr. Gopalendu Pal of Target — Interconnected simulations and stable processes are crucial for navigating supply chain volatility and optimizing enterprise-wide outcomes.
• Graphical conditional generative modeling for digital twin modeling — A framework discovers essential variables for parsimonious stochastic surrogates by analyzing full conditional laws, not just means.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by stat.ML updates on arXiv.org.