Tag: supply chain

  • Engineering ToteSim: High-Fidelity Intralogistics Modelling in AnyLogic

    Engineering ToteSim: High-Fidelity Intralogistics Modelling in AnyLogic

    This September at the online AnyLogic Conference 2026 (#ALConf26), the Decision Lab and Gousto engineering teams dived seep the technical architecture behind ToteSim.

    ToteSim is a data-driven simulation twin of the automated replenishment and material-handling system at Gousto’s Warrington fulfilment facility. Delivered through an intensive five-month collaborative sprint comprising 24 epics and over 300 technical tasks, the model bridges industrial discrete-event kinematics with modern cloud data stacks and live Python decision services.

    Below is an engineering walkthrough of the system topology, technical challenges, and architecture presented by Dr Fabrice Durier and Will Seymour (Gousto), alongside Peter Riley, Jacob Whyte, and Andrew Nestor (Decision Lab).

    You can watch the full recorded presentation from the conference below:

    At a Glance: ToteSim Technical Summary

    • Project: ToteSim — a high-fidelity digital twin of Gousto’s Warrington fulfilment centre (fulfilling ~200k boxes/week).
    • Architecture: Developed with AnyLogic, coupled directly with Databricks for real-time factory state initialisation and live Python microservices (Dynamic Replenishment and Totewise).
    • Physical Scale: Simulates 18,000 reserve storage locations, 9,800 active pick slots, 20 vertical lifts, and 270 automated shuttles.
    • Production Impact: Used to engineer the “Robin Hood” flow-control algorithm, delivering a +4.3% throughput AUC gain, 24% lower p95 tail queueing, and zero operational disruption across 2M+ audited journeys.

    1. The Physical Modelling Challenge

    Gousto’s Warrington plant operates 24/7, fulfilling approximately 200,000 customer recipe boxes and 5 million ingredient picks weekly. While Gousto had already built an established AnyLogic simulation for red-box picking lines, that model historically relied on simplified assumptions about upstream blue-tote supply.

    ToteSim was developed to model the physical replenishment system with complete fidelity:

    • 18,000 reserve storage locations in a multi-level automated shuttle system
    • 9,800 active pick slots
    • 20 vertical lifts and 270 automated shuttles
    • Shared high-speed conveyor loops (spines) that transport full, empty, and reserve replenishment totes across multiple levels
    ToteSim Replenishment: The Golden Triangle Modular Decomposition of Automated Material Handling & Decision Loops Zone Eligibility Ambient vs chilled rules RESERVE ASRS High-Bay Storage 18,000 Storage Locations 20 Lifts • 270 Automated Shuttles DMS Inventory Bin-level location tracking DECANT Inventory Induction Dynamic Replenishment Pallet infeed & tote fill TRANSPORT SPINES Robin Hood Controller Dynamic Headroom Balancing Shared multilevel conveyor loop STATIONS Pick Faces 9,800 Active Pick Slots Totewise slotting & supply Stock & Intake Data Pallet arrivals & SKU specs Flow Constraints Conveyor merge quotas & limits Picking Demand (Red Box) Recipe mix & order schedules Full Totes Empty Totes Full Totes Empty Totes Infeed & Replen. Retrieved Totes Full Totes Flow Empty Tote Return Data / Operational Constraints Modelled Domain Boundary

    In a high-throughput network running at ~3 seconds per completed customer box, local conveyor merges, lift bottlenecks, and recirculation loops cannot be evaluated with static queueing formulas. Physical layout kinematics interact continuously with automated warehouse control software.

    2. System Architecture: Integrating AnyLogic with Databricks and Python

    One of the common traps in enterprise simulation is building monolithic desktop models fed by static Excel or CSV files. From day one, ToteSim was engineered as an enterprise software component embedded directly within Gousto’s data ecosystem.

    ToteSim Architecture & Data Pipeline Databricks Ingestion • AnyLogic Simulation Engine • Python Decision Services DATA PLATFORM Initial State Generation Historical WES/WMS telemetry move_tasks inventory positions tray_quantities SIMULATION ENGINE 1. Runs ToteSim Model 2. Executes Move Tasks Tray quantities & kinematics 3. Collects Simulation State from active simulated runs 4. Exports DR / Totewise Data CSV Intermediate State Files Store inputs for DR & Totewise DECISION SERVICESProduction Python Logic Reads telemetry states Runs DR & Totewise Returns updated decisions & tote replenishment targets Initial Input Write State Read Telemetry Updated Tasks Closed-loop integration: AnyLogic executes kinematics while production Python services provide dynamic replenishment instructions

    The Ingestion & State Initialisation Pipeline

    Run parameters, inventory layouts, and movement tasks originate in Databricks. Rather than “warming up” a model from empty over arbitrary hours, ToteSim initialises directly into a historically observed factory state—including the exact location and contents of all 27,800 storage and pick slots, active tote coordinates on the belts, and outstanding order queues.

    Coupling Live Decision Code via Python

    Rather than re-coding Gousto’s production allocation algorithms into Java heuristics inside AnyLogic, ToteSim invokes Gousto’s live Python decision services during runtime.

    At periodic runtime checkpoints, AnyLogic exports the simulated factory state. Gousto’s proprietary Python services (Dynamic Replenishment (DR) and Totewise) read this telemetry, execute their logic, and return updated tasks (which pallets to decant, how many ingredients to place in each tote, and where to route replenishment) directly back into the simulation.

    This architectural decoupling ensures that the simulation tests the actual software logic deployed in production, not a synthetic approximation.

    3. Modelling Depth: Kinematics, Storage Logic, and Asset Reliability

    To deliver actionable decision intelligence, the model implements granular logic across two critical subsystems:

    A. Intelligent Inventory Logic, Not Just Geometry

    A common mistake in warehouse simulation is treating storage bins as simple abstract sinks. In ToteSim, inventory allocation is explicitly modelled as an operational decision layer. Every SKU is tracked across aisle, level, side, bin, and pickslot locations.

    Because storage decisions determine lift cycle times, conveyor travel distances, and congestion around specific aisles, the model faithfully reflects Gousto’s dynamic slotting and zoning eligibility rules.

    B. Asset-by-Asset Mechanical Reliability

    Lifts and shuttles fail in the real world. In ToteSim, mechanical downtime is treated as an explicit experimental variable rather than random, uncalibrated background noise.

    Every single lift (20 units) and shuttle car (270 units) can be independently configured under three distinct execution modes:

    1. Historic Replay: Replaying exact logged machine breakdowns from physical telemetry to audit past shifts.
    2. Synthetic Stress-Testing: Forcing selected critical lifts or entire aisle groups offline to measure how the conveyor loop recirculates and recovers from catastrophic hardware failure.
    3. Stochastic Sampling: Drawing from parametric failure distributions to quantify variance under extended runs.

    4. The “Red + Blue = Purple” Simulation Modular Boundary

    To manage model complexity, the project employed a strict architectural pattern dubbed Red + Blue = Purple:

    • Blue-Tote Model (ToteSim): Focuses strictly on decant induction, Automated Storage and Retrieval (ASRS) reserve, conveyor spines, and pick-face replenishment.
    • Red-Box Model: The mature picking model tracking customer cartons, recipe assembly lines, packing, and dispatch.
    • Purple (Integrated Twin): Connecting both models together to test complete end-to-end dynamics.

    This decoupling is essential for simulation engineering. It allows data scientists to isolate causal mechanisms and so test, for instance, whether pick-line starvation is driven by physical conveyor merge constraints in Blue or inefficient pick-sequencing rules in Red.

    5. Automated Verification, Calibration, and Validation

    Building trust in an engineering model requires objective calibration before running counterfactual experiments.

    The engineering team implemented headless execution pipelines that ran automated test suites via GitHub without launching the graphical AnyLogic animation. Historical production replay windows were assessed across two quantitative dimensions:

    • Headline Network Accuracy (~80%): Assessed via cumulative journey-volume curves (“worms”) comparing simulated vs. recorded tote completions across all major pathways.
    • Granular Critical Pathway Accuracy (>95%): Key high-volume flows—specifically decant-to-reserve (94.51%) and reserve-to-pickstation (96.42%)—reproduced production physical timings with exceptional precision.
    • Transit Distributions & Percentiles: Transit-time distributions and p5-through-p95 profiles along complex conveyor routes aligned closely with observed sensor telemetry.

    6. The Production Output: The “Robin Hood” Headroom Controller

    The engineering value of ToteSim was validated in Q1–Q2 2026 when Gousto used it to design, test, and deploy a new spine flow-control algorithm dubbed Robin Hood.

    The previous controller enforced hard, static tote-category limits on the main spine conveyor loop. When decant induction totes hit their ceiling, they queued at the merge point—blocking upstream belts and starving pick stations, even while spare capacity existed on the belt for empty or replenishment totes.

    From 2,000+ Headless Runs to Live Deployment

    AnyLogic animation of totes on conveyors in ToteSim.

    Using ToteSim’s headless execution, the team executed over 2,000 simulation runs across varying stress levels (testing artificially compressed tray volumes to simulate heavy peak loads). Candidate control scripts were evaluated using Cumulative Spine Activity Area Under the Curve (AUC), tracking total entries and exits over time.

    The Robin Hood logic demonstrated a +4.3% throughput AUC gain under high stress (p<0.001p<0.001).

    Because the algorithm was coded as a drop-in Python module using the exact telemetry inputs and outputs established in ToteSim, Gousto deployed the code directly into Warrington’s production environment at the start of Q2 2026.

    A subsequent six-week production audit across more than 2 million live tote journeys confirmed the simulation findings:

    • Zero System Destabilisation: Normal operating throughput remained completely stable.
    • 24% Reduction in p95 Tail Queueing: The 95th-percentile queue time for heavily constrained decant induction dropped by nearly a quarter.
    • 10%–20% Lower Peak Congestion: Avoidable merge waiting was slashed during peak operational windows.

    Illustrative redraw of the Use Case 6 spine_activity metric: higher or earlier cumulative activity increases AUC.

    A higher cumulative activity curve means the spine transported more totes and/or transported them earlier.

    Looking Ahead: The Evolution Toward Real-Time Twin Control

    ToteSim demonstrates that building high-value digital twins does not require starting with a fully autonomous, live-connected IoT system. By treating simulation twinning as an integrated software discipline and combining AnyLogic’s discrete-event power with modern data lakes and live Python services, engineering teams can eliminate production risk and de-risk major operational software changes before touching live machinery.

    Current work on ToteSim includes pallet induction prioritisation, dynamic pick-face allocation, and reverse-engineering third-party equipment logic to further scale Gousto’s automated facilities.

  • OR68 and Gousto’s ToteSim Digital Twin

    OR68 and Gousto’s ToteSim Digital Twin

    From Data to Decisions at Scale

    Earlier this month, the UK and international Operational Research community descended on the University of Nottingham for OR68: From Data to Decisions.

    The event featured three intense days of plenaries, poster sessions, and valuable discussions at our stand. We were honoured to present and, in particular, answer the standout question from practitioners and business leaders alike: how do you build enough stakeholder trust in an analytical model to let it directly influence physical, high-risk operations?

    Decision Lab and Gousto tackled that question head-on in our joint session in the Simulation & Digital Twins stream. Presenting alongside Dr Fabrice Durier and Will Seymour from Gousto, Decision Lab’s Peter Riley unpacked the delivery of ToteSim, a data-driven simulation twin of the automated blue-tote replenishment system at Gousto’s Warrington fulfilment facility.

    Earlier this year, we published a high-level overview in our case study, How Simulation Became Gousto’s Recipe for Growth. At OR68, we took the conversation a level deeper into the applied decision science: the validation mechanics, the experimental rigour, and the physical factory evidence.

    Here are the key operational research insights shared during the conference.

    1. When Combinatorial Variety Meets Physical Bottlenecks

    As explored in our published case study, Gousto’s strategic mission to scale its menu from 50 to 200 weekly recipes created exponential combinatorial complexity behind the scenes. Fulfilling roughly 200,000 bespoke boxes and picking 5 million ingredients every week means that an ingredient pick must occur every three seconds across dynamic, non-repeating recipes.

    While Gousto already maintained a mature simulation model for customer “red-box” packing lines, that model assumed blue-tote ingredient replenishment was an infinite, frictionless feed. In reality, Warrington’s automated tote network, which houses 18,000 reserve bins, 9,800 pick slots, 20 vertical lifts, and 270 automated shuttles, was tightly constrained.

    Conveyor merges and shared transport spines meant that a micro-stoppage in tote replenishment could cascade across the entire facility. Yet with a 24/7 operating plant, conducting physical trial-and-error was out of the question: running live A/B experiments in a single facility carried unacceptable operational risk.

    The OR challenge was to determine whether intelligent algorithmic control could unlock latent transport capacity without committing to capital-intensive physical expansions.

    2. Bridging the Gap: What Makes “Simulation Twinning” Different?

    Much of the academic discussion at OR68 centred on the levels, or extent, of digital twin maturity. Rather than striving for an immediate, all-singing real-time twin, ToteSim was built around pragmatic Simulation Twinning:

    • The Golden Triangle Architecture: Decoupling the physical replenishment facility into three bounded operational domains (Decant Induction, Reserve Storage, and Pick Stations) connected by the shared transport conveyor loop. This modular design allowed subsystems to be isolated or recombined across 24 project epics during an intensive five-month collaborative sprint.
    • Executing the Live Decision Stack: Crucially, ToteSim is not an abstract approximation of rules. Built in AnyLogic and fed shift-level historical states via Databricks, the model directly invokes Gousto’s live production Python decision services (including Dynamic Replenishment and Totewise). The simulation twin evaluates the exact algorithmic logic deployed in the warehouse.
    • The “Red + Blue = Purple” Modularity: Analysts can isolate the replenishment twin (Blue) with synthetic pick rates, or couple it directly with the customer order picking model (Red) to evaluate holistic “Purple” system impacts.

    3. Decision-Specific Validation: Building Credibility for Capital Decisions

    A core lesson shared with the OR Society delegates was how validation must be structured to build operational trust. Rather than chasing an unachievable standard of “100% global fidelity,” ToteSim’s validation strategy followed Robinson’s and Sargent’s principles of operational validity for defined decisions:

    • ~80% weighted mean accuracy across broad, multi-flow network replays against historical shift data.
    • >95% accuracy on critical, high-volume pathways (such as decant-to-reserve and reserve-to-pick flows).

    This decision-specific calibration gave Gousto’s operations leadership the confidence to use ToteSim not just as a retrospective reporting dashboard, but as a proactive experimental testbed.

    4. The First Factory Proof Point: “Robin Hood” Headroom Balancing

    The first production deployment tested whether algorithmic optimisation could ease congestion on the shared conveyor spine feeding the pick tower.

    Previously, rigid category quotas caused totes to queue at the decant line whenever their specific category limit was reached, even when spare capacity existed elsewhere on the conveyor. The team developed a drop-in controller, internally dubbed Robin Hood, based on headroom balancing:

    1. When total spine occupancy is below target, spare headroom is equalised dynamically across tote streams, clearing queues before merges freeze.
    2. When the belt reaches saturation, category limits are pulled back toward a central balance to prevent any single category from monopolising throughput.
    3. Designed as a direct drop-in replacement, it utilised the exact same input/output telemetry, requiring zero changes to physical sensors or PLC controls.

    From 2,000+ Runs to Measured Production Evidence

    Before factory rollout, ToteSim stress-tested the candidate algorithm across 2,000+ simulation runs, demonstrating a +4.3% improvement in cumulative spine activity area-under-the-curve under high stress ( p<0.001)p<0.001) ).

    Deployed into live production at Warrington at the start of Q2 2026, a six-week post-deployment evaluation across more than 2 million live tote journeys confirmed what the simulation predicted:

    • Baseline Stability: General transit times and empty/reserve movements remained completely stable, proving that the intervention was safe and non-disruptive.
    • High-Stress Tail Reduction: The 95th-percentile queue time for constrained decant totes dropped by 24%.
    • Peak Congestion Relief: Avoidable queueing during peak operational windows reduced by 10% to 20%.

    Moving from Event Models to an Enduring Decision Capability

    Our headline message at OR68 was simple: a digital twin creates value long before it becomes an autonomous, fully connected IoT system.

    By establishing a reliable loop—Observe  Initialise  Validate  Experiment  Deploy  Measure—ToteSim has evolved from a simulation model into an operational decision capability. Gousto is now utilising the twin to stress-test lift and shuttle failure modes, evaluate pallet induction strategies, and validate future automation logic before deploying capital.

  • Skin in the Game: An Antifragile Design Necessity

    Skin in the Game: An Antifragile Design Necessity

    Why shared-risk architecture, antifragile systems, and accountable AI are becoming essential to modern supply chain design

    Following the closure of the Strait of Hormuz, commercial maritime traffic in the region collapsed by over 90% almost immediately.

    This is no longer a localised disruption or a temporary energy shock; we are navigating a PolyCRISIS characterised by cascading failures across finance, agriculture, manufacturing, and energy. For clients and customers alike, the illusion of a stable, predictable global landscape has entirely shattered. The overarching philosophy of the past three decades focused on hyper-optimised, just-in-time (JIT) deliveries and it is dead.

    The Shift in Paradigm: Beyond Resilience

    Until recently, when disruptions occurred, organisations relied on resilience. They used static buffers like idle capacity or excess stock to absorb shocks and eventually bounce back to a baseline state.

    Now, thanks to developing techniques and technology, it is possible to move beyond the inefficiencies of traditional buffers. Recent global data indicates that traditional safety stocks tie up billions in stagnant capital only for them to fail when confronted with events of the modern world.

    We must abandon the pursuit of mere resilience. It is time to engineer genuine antifragility. Coined by Nassim Nicholas Taleb, antifragility describes a system that actively learns, adapts, and grows stronger as a direct result of volatility. While fragile systems break under stress, antifragile ones thrive because of it.

    The Design Necessity: Architecting Shared Risk

    Before deploying technology to harvest this volatility, there is a strict strategic prerequisite. If a business or a partner learns to profit from chaos, what stops them from deliberately manufacturing it? To prevent actors from exploiting systemic stress while offloading the cost of failure, supply chain leaders must engineer ‘skin in the game’ directly into their commercial and operational architectures.

    When actors face financial or operational ruin for creating destructive friction, the incentive to exploit volatility vanishes. Here is how this is structurally enforced in the real world:

    • Outcome-Based Servitisation: The definitive standard is shifting to outcome-based contracts, exemplified by Rolls-Royce’s ‘Power by the Hour’ programme. Instead of profiting from repairs, Rolls-Royce charges based exclusively on flying hours. By absorbing the repair costs, they assume absolute risk, incentivising them to pioneer predictive maintenance. For modern supply chains, this can mean paying logistics providers for network uptime rather than freight moved.
    • Vested Outsourcing: When risk is merely shifted, partners act defensively. The reverse logistics partnership between Dell and FedEx transitioned from rigid, cost-per-unit contracts into a shared economic model. By sharing the financial risk of delays and the rewards of efficiency gains, FedEx was structurally incentivised to proactively solve problems and adapt the network during crises, rather than hiding behind Service Level Agreements (SLAs).
    • Algorithmic Accountability: As organisations deploy mathematical optimisation and Agentic AI into critical infrastructure, skin in the game must extend to technology vendors. If an AI model is licensed to route freight, the vendor’s remuneration should be tied to verified, real-world operational outcomes.

    From Philosophy to Execution: The Decision Lab Approach

    At Decision Lab, we believe this algorithmic accountability is non-negotiable. We build mathematical frameworks that enforce shared-risk architectures. Once structural skin in the game is established, organisations can safely deploy autonomous technologies to harvest volatility.

    Using advanced optimisation techniques, we help organisations execute this strategic paradigm shift across four key operational vectors:

    1. AI-Driven Energy Orchestration The Strait of Hormuz closure has triggered extreme spot market volatility for energy. Our EcoSynth architecture uses an AI-powered orchestration engine to build a mixed-integer linear programming model of a facility. By intelligently shifting non-critical processes to cheaper off-peak windows, implementations have reduced total site energy consumption by 7% to 12%.

    2. Probabilistic Supply Chain Digital Twins Rerouting around the Cape of Good Hope adds a paralysing 10 to 15-day delay to global networks. We deploy Probabilistic Supply Chain Digital Twins to create a highly accurate, mathematical replica of an end-to-end logistics network, allowing planners to use Monte Carlo simulations to dynamically simulate mitigation strategies.

    3. Multi-Echelon Inventory Optimisation When lead times stretch unpredictably, single-echelon models fail, causing massive inventory swings. Decision Lab’s optimised capacity planning evaluates the network holistically, simultaneously analysing suppliers, maritime lanes, and distribution centres. Documented applications have proven to structurally reduce inventory costs by 10% to 35%.

    4. Autonomous Execution with Agentic AI Conflict-driven friction has spiked bunker fuel costs and war-risk premiums. Human planners cannot react fast enough, necessitating Agentic AI. These frameworks evaluate mitigation strategies and devise optimal responses.

    Securing the Future

    Deploying autonomous systems into critical infrastructure requires absolute trust. Because we believe in strict accountability, we maintain an unwavering commitment to AI TRiSM (AI Trust, Risk, and Security Management), ensuring our models operate securely within strictly defined guardrails of bounded autonomy.

    The mandate for the C-suite is clear: do not wait for the next macroeconomic shock to expose the limits of your operational architecture. By embedding predictive foresight into your value chain, you can mitigate today’s vulnerabilities and develop a game-changing competitive advantage.

    Read more in Navigating the PolyCRISIS: Why Supply Chain Resilience is Dead

  • Navigating the PolyCRISIS: Why Supply Chain Resilience is Dead

    Navigating the PolyCRISIS: Why Supply Chain Resilience is Dead

    …and what Supply Chain leaders must build instead

    The global macroeconomic environment has definitively transitioned from an era of manageable disruptions into one of permanent, systemic volatility. The 2026 closure of the Strait of Hormuz has severed one of the world’s most critical maritime and economic arteries, demonstrating exactly how fragile our interconnected systems have become. Following the blockades, commercial maritime traffic in the region has collapsed by over 90%.

    The war in the Middle East is creating the largest supply disruption in the history of the global oil market.

    The International Energy Agency (report)

    While the immediate reaction framed this purely as an energy shock, supply chain analysts rightly characterise the current operating environment as a ‘PolyCRISIS’. We are witnessing cascading failures across energy, manufacturing, agriculture, and finance.

    At Decision Lab, we recognise that for senior supply chain leaders, the illusion of a stable, predictable global landscape has entirely shattered. The overarching philosophy of the past three decades that was focused on hyper-optimised, just-in-time (JIT) manufacturing is no longer viable. It is time to abandon the pursuit of mere resilience and engineer genuine antifragility.

    The Antifragile Imperative

    Traditional resilience relies on static, defensive buffers: hoarding excess safety stock, contracting redundant suppliers, or building idle capacity. Resilience is designed to absorb a disruption and bounce back to a baseline state. However, these buffers are economically inefficient and inherently fragile when confronted with black-swan geopolitical events that exceed their design parameters.

    A diagram of the Uncertainty Spectrum, showing the progression from a Fragile state (suffers from uncertainty) to a Resilient state (resists uncertainty) to an Antifragile state (gains from uncertainty).

    Antifragility, a concept coined in a book of the same name by risk analyst Nassim Nicholas Taleb, describes systems that do not merely withstand volatility, but actively learn, adapt, and grow stronger as a direct result of it. If fragility implies breaking under stress, antifragility implies thriving because of it.

    Through mathematical optimisation, artificial intelligence, and probabilistic modelling, Decision Lab helps organisations map the full probability density function of all plausible future scenarios, executing a strategic paradigm shift toward antifragility. Here is how we are architecting this transformation across four critical operational vectors.

    1. AI-Driven Energy Orchestration

    The Strait of Hormuz closure has triggered severe spot market volatility in energy. This is particularly damaging in the UK, where energy constitutes one of the largest single operational expenses for the FMCG and Pharmaceutical sectors, with the UK Pharmaceutical sector spending over £1 billion annually.

    To survive, manufacturers must transition into dynamic energy orchestrators. Decision Lab’s EcoSynth architecture provides an AI-powered orchestration engine that constructs a highly detailed mixed-integer linear programming model of a facility.

    • By ingesting historical data, grid pricing, and weather forecasts, the system precisely anticipates the cost and carbon footprint of every kilowatt-hour required.
    • It intelligently identifies non-critical processes and autonomously shifts them to cheaper, low-carbon off-peak windows.
    • Implementations of EcoSynth have demonstrated the ability to deliver a 7% to 12% reduction in total site energy consumption.

    To see this principle applied, explore our work on building an antifragile pharmaceutical production facility and creating an antifragile future for medicines manufacturing.

    2. Probabilistic Supply Chain Digital Twins

    With the Red Sea rendered highly dangerous, global shipping lines are rerouting around the Cape of Good Hope, adding 3,500 to 11,000 nautical miles to Asia-to-Europe voyages. This introduces a paralysing 10 to 15-day delay into global networks.

    Decision Lab addresses this operational blindness through Probabilistic Supply Chain Digital Twins.

    • A Digital Twin is a highly accurate, mathematical replica of an organisation’s entire end-to-end logistics network.
    • The system uses Monte Carlo simulations to inject realistic randomness into the model, generating thousands of potential future states based on variables like port congestion and weather.
    • Instead of waiting for delays to manifest as stockouts, planners can dynamically simulate mitigation strategies, transforming unmanageable chaos into a quantifiable mathematical equation.

    For an in-depth look at how this operates at scale, review our case study on strategic capacity planning for a revolutionary pharmaceutical development, developed in collaboration with AstraZeneca.

    3. Multi-Echelon Inventory Optimisation (MEIO)

    The global supply of Active Pharmaceutical Ingredients (APIs) is dangerously concentrated, with approximately 60% to 70% of global production situated in Asia. When supplier lead times stretch unpredictably, traditional single-echelon inventory models inevitably fail, leading to the massive inventory swings known as the “bullwhip effect”.

    Decision Lab resolves this through Multi-Echelon Inventory Optimisation (MEIO).

    • MEIO simultaneously analyses Asian API suppliers, maritime lanes, European formulation facilities, and distribution centres to evaluate the entire network holistically.
    • The system identifies the precise mathematical optimum for inventory positioning, balancing the high cost of holding working capital against the catastrophic risk of a medical stockout.
    • Documented industry applications of MEIO methodologies have consistently demonstrated structural inventory cost reductions ranging from 10% to 35%.

    Discover more about how we implement this via optimised production and sustainable capacity planning.

    4. Autonomous Execution with Agentic AI

    The friction generated by the Middle East conflict has manifested as severe financial strain, with war-risk premiums and escalating bunker fuel costs adding between £150 and £600 per container. In this volatile environment, human supply chain planners simply cannot move fast enough to secure optimal routing.

    The solution is Agentic AI.

    • Unlike traditional AI designed to passively answer questions, Agentic AI is engineered to take goal-oriented, autonomous action.
    • When the agent detects a massive rate spike or capacity crunch, it evaluates mitigation strategies and can autonomously execute the optimal response—booking freight space and rerouting shipments directly within the enterprise software environment.
    • Agentic frameworks promise to reduce overall logistics costs by up to 15% and optimise inventory levels by 35%.

    Moving an organisation along the spectrum from a fragile to an antifragile state requires a new cognitive architecture for decision-making. For a detailed guide on the four pillars of this transformation—from de-risking capital decisions to building a sentient factory—download our complete white paper.

    Securing the Future: Our Commitment to AI-TRiSM

    As we integrate these highly autonomous, self-governing systems into critical supply chains, trust and security are paramount. Decision Lab maintains an unwavering commitment to AI-TRiSM (AI Trust, Risk, and Security Management). We ensure that every algorithm deployed, from our Digital Twins to our Agentic AI frameworks, operates within strictly defined guardrails of “bounded autonomy,” ensuring models are rigorously transparent, compliant, and secure.

    Our advanced tooling is specifically engineered for mid-to-large global manufacturers seeking to dominate complex markets through mathematical superiority. The mandate for the C-suite is clear: do not wait for the next macroeconomic shock to expose the limits of your operational architecture.

    By partnering with Decision Lab to embed predictive foresight and intelligent automation into your value chain, you can secure antifragile pharma production and transform today’s vulnerabilities into tomorrow’s insurmountable competitive advantage.

  • Introducing anyLogistix Sandbox

    Introducing anyLogistix Sandbox

    Try Supply Chain Optimisation in Your Browser

    Our technology partners at The AnyLogic Company have recently launched the anyLogistix Sandbox, offering a completely new way to experience advanced supply chain optimisation directly from your web browser.

    In today’s fast-paced logistics environment, speed is essential when evaluating new strategies. The Sandbox is designed for supply chain professionals, managers, and consultants who want to see how advanced supply chain optimisation works in practice, with zero installation, download, or technical setup required.

    Key Capabilities to Explore:

    • Greenfield Analysis: Design a supply chain network from scratch to determine the optimal number and locations of your facilities.
    • Network Optimisation: Balance demand, throughput, and storage to define the most efficient strategic structure for your network.
    • Simulation & Risk Analysis: Run dynamic what-if scenarios to evaluate how your supply chain performs over time and assess its resilience against disruptions or demand changes.
    • Last-Mile Optimisation: Plan efficient delivery routes and minimise total drive times.

    From Sandbox to Enterprise Solutions

    The Sandbox is a fantastic tool for fast exploration and learning. Because it is a demonstration environment, it does have a few intentional limitations—such as restricted model sizes, a one-project limit, and no custom data import.

    When you are ready to move beyond the Sandbox, into business-scale or enterprise projects, Decision Lab is here to help. We can provide comprehensive demonstrations of the full anyLogistix software and develop bespoke, end-to-end supply chain solutions tailored to your specific data. We have a proven track record of successfully delivering these advanced logistics models for major companies operating across the UK and Europe.

    anyLogistix Sandbox

    We run anyLogistix courses each quarter, live online. They are a great way to get started quickly and get input on your project: anyLogistix course.

  • Decision Intelligence in Route Optimisation

    Decision Intelligence in Route Optimisation

    A 6-Week PoC with FedEx European Linehaul

    Executive Summary

    Decision Intelligence moves an organisation beyond the fixed-plan trap toward proactive, automated resilience in route optimisation. By evaluating the strategic trade-offs between explainable Stochastic Optimisation and scalable Reinforcement Learning, we proved that move-level agility is the key to maintaining flow in a high-uncertainty environment.

    Key Takeaways:

    • Beyond Rigid Scheduling: Shifting from historical templates to dynamic, operational-time decision-making to maximise capacity utilisation.
    • The Technical Showdown: Comparing the audit trails of Stochastic Programming against the autonomous adaptability of Reinforcement Learning.
    • Predictive Simulation: Utilising a road-based “digital sandbox” to test courses of action and mitigate risks before committing resources.
    • Tangible ROI: Delivering financial returns by improving linehaul utilisation and significantly reducing the need for costly ad-hoc transport.

    The Challenge: The Friction of Fixed Planning

    In the high-stakes corridors of European logistics, fixed plans are often the first casualty of reality. For an Operations Director managing a distribution network across the UK and EU, the daily friction is visceral. You are constantly forced to ask: “Should I delay this trailer, so it leaves full, or stick to the schedule? Do I need to commission an expensive ad-hoc truck to cover this surge, or will the bottleneck clear itself?”

    When package volumes fluctuate unpredictably at major hubs, static schedules become more than just an inconvenience—they become a drain on margins and a threat to service levels. At Decision Lab, we operate under a foundational truth: the success of an organisation is nothing but the sum of all its decisions. To help global leaders move beyond the fixed-plans trap, we conducted a six-week Proof of Concept (PoC) with FedEx. This project tackled real-world complexity head-on, proving that Decision Intelligence is the key to transitioning from reactive firefighting to proactive, automated resilience.

    Static Schedules are the Enemy of Efficiency

    The core challenge identified within FedEx’s European network was the inherent limitation of pre-defined linehaul schedules. These schedules were designed for averages, whereas logistics are often defined by exceptions. When incoming and outgoing package volumes at European hubs diverged from the forecast, rigid plans could not adapt.

    A dynamic approach, powered by operational-time decision-making, is the only way to maintain flow in a volatile environment. By rerouting assets and scheduling departures based on real-time parcel traffic rather than historical templates, an organisation can achieve step-changing improvements in capacity utilisation.

    Our expertise in AI, ML, simulation, and mathematical optimisation helps organisations cut through complexities in strategic, tactical and operational processes.

    The Solution: Bridging the Gap with Decision Intelligence

    The choice between technical approaches is rarely straightforward; in this case it was a strategic balancing act between Explainability and Scalability. During our PoC, we evaluated two competing methodologies: explainable Stochastic Programming and scalable Reinforcement Learning (RL).

    FeatureStochastic ProgrammingReinforcement Learning (RL)
    Primary StrengthFast solving speed; mathematically explainable and provable.Reacts to high uncertainty using World Models and Graph Neural Networks.
    Logic BasisLocates the best strategy to optimise expected outcomes over uncertainty.Uses a dynamics model to predict the optimum next action.
    AdaptabilityMulti-objective handling: Uniquely suited for balancing cost vs. customer service levels.Observation-size invariant: Handles environments with variable data lengths and network nodes.
    Strategic RiskConsulting intensive: Very sensitive to human-built heuristics, which are expensive and time-consuming to develop.Compute intensive: Requires significant hardware resources for training the World Model.

    While Stochastic Programming offers a clear audit trail for every decision, RL provides the adaptability required for massive, interconnected networks. The right choice depends on whether your organisation prioritises a provably optimal solution or a highly performant best-effort that can autonomously learn the shifting dynamics of global markets.

    Don’t Just Predict—Simulate the Impact

    One of the most powerful tools developed for FedEx was a road-based, hub-to-hub package movement simulator. This provides a digital sandbox where controllers can explore alternative COAs (Courses of Action) before committing resources.

    Our completely data-driven deployment method allows us to build these simulations without the months of manual coding traditionally required. By accessing relevant operational and transport data directly, we can simulate supply chain environments to predict the ripple effects of a delay or reroute.

    This tool predicts the impact of different actions, helping to mitigate risks and optimise routes.

    Data Maturity is the Ultimate Competitive Moat

    For large-scale firms with a £200M+ turnover, the transactional backbone—usually an ERP or MRP system like SAP or Oracle—is necessary but insufficient. To achieve true antifragility, you must layer Decision Intelligence over these systems.

    Antifragility is the ability to not just survive volatility, but to actually improve because of it. By utilising a World Model within an RL framework, the system treats every fluctuation in package volume as a learning opportunity, refining its dynamics model to better anticipate future shocks. This requires three layers of data maturity:

    • Strategic Level: Long-term high-level routes, fleet capacity, and cost-per-mile data.
    • Operational Level: Real-time visibility into items currently loaded or waiting at the depot.
    • Historical Level: Deep archives of how volumes fluctuated in similar time slots in the past.

    The Result: Antifragility and Bottom-Line Returns

    In the C-suite, the value of AI is measured by the bottom line. The FedEx project was not an academic exercise; it was focused on delivering the financial returns demanded by an industry with tight margins. The PoC demonstrated that an autonomous planning agent directly impacts:

    • Improved Linehaul Utilisation: Driving higher Overall Equipment Effectiveness (OEE) across the fleet.
    • Reduced Rescheduling: Eliminating the administrative friction and cost of mid-stream plan changes.
    • Minimised Ad-hoc Linehauls: Directly de-risking Operational Expenditure (OPEX) and informing more accurate Capital Expenditure (CAPEX) by reducing the need for emergency transport.

    Conclusion: Toward the Global Digital Twin

    The ultimate evolution of this journey is an advanced road-based package movement digital twin. By connecting multiple hubs in real-time, organisations can create a living model of their entire network that learns, adapts, and optimises itself.

    What is the sum of your organisation’s decisions? How many of your current logistics choices are being left to a fixed plan that no longer fits your reality? In a world of increasing volatility, the goal is no longer just to have a plan—it is to have a system that provides decision clarity and reliable value.

    Transform your logistics operations today. Reach out directly via out contact page, or connect with us on LinkedIn to start a conversation about de-risking your future.

  • Why LLMs Aren’t Enough: Engineering Antifragile Operations with Composable Decision Intelligence

    Why LLMs Aren’t Enough: Engineering Antifragile Operations with Composable Decision Intelligence

    As we navigate the technological landscape of 2026, Generative AI has undoubtedly transformed the way we interact with information. Chatbots and Large Language Models (LLMs) have proliferated across enterprise software, streamlining communication and automating basic workflows. However, for operations and supply chain leaders in complex, capital-intensive industries like FMCG, Automotive, and Retail manufacturing, a stark reality is emerging: LLMs are not a silver bullet.

    While language models excel at processing text, they cannot single-handedly optimise a global supply chain network, nor can they provide the quantitative assurance needed to de-risk a £50m factory expansion. When dealing with physical realities, extreme market volatility, and fragmented legacy systems, text prediction is insufficient.

    By 2026, 75% of Global 500 companies will apply decision intelligence practices

    Gartner

    The definitive competitive edge in 2026 belongs to those looking beyond Generative AI toward Decision AI. It belongs to organisations thoughtfully advancing their tech stack and building on established capabilities to embrace the architecture of a Composable Decision Intelligence Platform (DIP).

    The Industrial Reality: High Stakes and High Volatility

    Traditional planning systems struggle to account for agile and accelerated business and the consequent hazards. Today’s supply chain and operations leaders are caught in a crossfire of overlapping challenges, two of the most critical being:

    • Demand & Supply Volatility: SKU proliferation, shifting consumer behaviours, and frequent supply chain disruptions are breaking static planning models. The inability of legacy systems to cope with this extreme volatility inevitably results in poor service levels, excess inventory, and spiralling costs.
    • High-Stakes CAPEX Uncertainty: Securing funding for major capital investments—whether a new automated line, a facility expansion, or rationalising a post-merger manufacturing network—requires robust, data-driven justification. Without quantitative assurance, it is incredibly difficult to de-risk these investments and guarantee ROI.

    Solving these multi-dimensional problems requires more than just analysing past data; it requires a platform capable of simulating the future and discovering the optimal path forward.

    The Path to Implementation: A Composable Architecture

    At Decision Lab, we deliver Decision Intelligence to help leaders master this uncertainty. We achieve this not through a rigid, black-box AI model, but by building a Composable Decision Intelligence Platform based on responsible AI TRiSM principles.

    Composability is the principle that enables businesses to be agile. Rather than relying on a single vendor’s inflexible suite, a composable DIP orchestrates best-in-class, modular capabilities that ingest data from fragmented ERP, MES, and WMS systems. This creates a unified, dynamic view—an AI Simulation Twin.

    The Strategic Advantage of the AI Simulation Twin

    Instead of waiting years for a fully instrumented, hardware-dependent Digital Twin, leading organisations are accelerating their time-to-value by deploying an AI Simulation Twin.

    Traditional Digital Twin programmes often stall in pilot purgatory due to immense IoT integration challenges, prohibitive hardware costs, and fragmented legacy data pipelines. A Simulation Twin, while still ingesting real data, fundamentally bypasses these immediate infrastructure hurdles. It delivers the core predictive and prescriptive advantages now—providing a high-fidelity virtual environment to solve urgent CAPEX and operational bottlenecks—while your physical IoT maturity can be developed as a separate, parallel track. This decoupling ensures you realise ROI in months, rather than years, before moving into the four pillars of the platform:

    Infographic of the four pillars of a composable decision intelligence platform.

    1. The Cognitive Engine: Autonomous AI Agents

    Agentic AI serves as the reasoning layer of the platform. These agents can interpret complex scenarios, model market volatility, and process multi-tiered supply chain dynamics, translating raw data into actionable context.

    2. The Virtual Sandbox: Simulation

    To understand a complex physical network, you must be able to interrogate it and test it. Practically, that means replicating it. We use simulation to build a high-fidelity digital twin environment, employing appropriate technologies, such as AnyLogic’s multi-method capabilities. A simulation maps constraints, machines, and distribution nodes, providing the holistic view necessary to test what-if scenarios safely. It answers critical CAPEX questions before money is spent.

    3. The Mathematical Engine: Optimisation

    Where simulation shows you what could happen, optimisation dictates what should happen. For us, that means employing mathematical optimisation, such as Gurobi’s world-class mathematical solver, to cut through millions of potential permutations. It discovers the mathematically perfect production schedules and inventory policies—maximising throughput and service levels while minimising duplicated costs. The key is being timely—it is no good getting the answer after it was needed. Gurobi’s speed is key here (Gurobi white paper on solver speed).

    4. The Continuous Learning Loop: Reinforcement Learning

    This is where the platform moves from a passive analytical tool to an active operational asset. By applying Reinforcement Learning, specifically leveraging AgileRL, a platform can learn from real-time feedback. It continually experiments within the simulation, discovering new strategies to navigate supply shocks or demand spikes as they happen.

    Engineering the Antifragile Supply Chain

    The ultimate goal of implementing a Composable Decision Intelligence Platform is to shift operations from a state of fragility to one of Antifragility.

    A robust system merely survives a shock. An antifragile operational system improves when exposed to volatility. When a sudden supply chain disruption occurs, the reinforcement learning algorithms immediately assess the new reality within the simulation, trigger the optimisation engine to recalculate the best path, and deploy autonomous agents to orchestrate a self-adapting response. Relying on singular AI models or monolithic ERPs to solve complex physical problems is being consigned to the past.

    For leaders navigating constant disruption, true agility requires an adaptable, composable ecosystem. By implementing a Decision Intelligence Platform, you gain the foresight not just to predict the future, but to engineer your position—a compelling competitive advantage now and for the future.

    To find out more, check out our case studies or contact us!

  • Bold Choices in an Uncertain World

    Bold Choices in an Uncertain World

    Why Gurobi’s Latest Keynote is a Blueprint for Antifragility

    In the current landscape of supply chain and operational planning, the search for certainty is a losing battle. Yet, many organisations still build their strategies around static forecasts, hoping the world will comply with their spreadsheets.

    At the recent EMEA Gurobi Summit in Vienna, a different path was illuminated. In their keynote Bold Choices, Proven Methods, Gurobi’s Dr. Kostja Siefen and Ronald van der Velden outlined a comprehensive framework for empowering confident decision-making. They effectively demonstrated that optimisation is not only a tactical tool for efficiency; it is a core method underpinning bold leadership. Their presentation encapsulated what we at Decision Lab recognise as Antifragility and the ability to not just withstand volatility, but to use it to gain a competitive edge.

    While resilience is about surviving a shock, antifragility is about improving because of it. Siefen and van der Velden’s keynote perfectly articulated the mechanics required to build such a system. They argued that to move from tentative planning to bold action, organisations must master three specific areas: challenging the status quo, mastering the projects, and achieving business value.

    Here is how Gurobi’s technical roadmap aligns with the strategic imperative of antifragility.

    Resilience Over Prediction

    One of the most dangerous traps in decision-making is the ‘illusion of certainty’. As noted in the keynote, “failures in planning and strategy often stem from misplaced confidence… rather than from making the ‘wrong’ decision”.

    A fragile system assumes the forecast is correct. An antifragile system, however, relies on a Robustness Strategy. As their presentation articulated, the goal is not to eliminate uncertainty but to design processes that endure when reality diverges from expectations. By moving away from a single, unrealistic scenario and embracing uncertainty-aware optimisation, businesses can design plans that remain effective even when the unexpected happens.

    Evolution Through Pacing

    True antifragility is not achieved through a single ‘big bang’ implementation. It requires an Evolution Strategy, viewing the past and future not as opponents, but as partners.

    The keynote speakers emphasised the importance of a Pacing Strategy, where change is introduced through focused stages that create momentum. Rather than demanding instant perfection, successful optimisation projects ‘start small, learn fast, and adapt’. This iterative approach allows an organisation to absorb stressors, such as data quality issues or stakeholder resistance, and use them to refine the model, making the final solution stronger and more fit for purpose.

    The Art of Adaptive Focus

    In the age of Digital Twins, there is a temptation to model every atom of a supply chain. However, complexity without clarity leads to paralysis. The Gurobi keynote introduced the concept of Adaptive Focus: the practice of purposeful abstraction to create robust output.

    By keeping the level of detail adjustable (by balancing relevance, detail, and time) decision-makers can utilise a ‘zoom lens’ to focus on what truly matters in a crisis. This capability is essential for antifragility; it allows leaders to filter out the noise and make rapid, high-quality decisions based on the relevant constraints of the moment.

    Trust as the Currency of Change

    Perhaps the most critical insight for leadership was that trust drives adoption. No matter how advanced the mathematics, a solution will fail if the humans at the helm do not trust it.

    Gurobi’s Trust Strategy suggests that trust should be an informed stance rather than an emotional leap. By utilising what-if scenarios, explainability, and human-in-the-loop validation, organisations can transform compliance into commitment. When teams trust the ‘black box’, they are empowered to make the bold choices required to navigate volatility.

    AI trust and AI TRiSM, a woman printing to a node network featuring a padlock and a shield with a check mark.

    Building trust in complex systems is essential. See how Decision Lab builds trust into projects from the start using the AI TRiSM framework.

    The Decision Lab Perspective

    The methods outlined by Siefen and van der Velden at the Gurobi Summit confirm that the technology to build antifragile systems is already here. We see the future of supply chain planning as range-based experiment-driven. Indeed, Gartner recently recognised Decision Lab as a representative provider in an Innovation Insight report.

    On the journey from fragile, through resilient, to antifragile, we recognise the speed and power of Gurobi’s world-class solver, leveraging its capabilities for solutions to supply chain and defence related challenges. By combining Gurobi’s proven methods with a strategic focus on antifragility, we help our partners stop fearing uncertainty and start using it to their advantage.

    As a Trusted Partner, we offer Gurobi Compass training. It allows teams to integrate solutions and make the most of the solver quickly, whatever challenges they seek to resolve.

    Our Antifragility C-Suite Roadmap for supply chain from Decision Lab CEO David Buxton. Get your copy

  • Logistics Optimisation: LOGOS+

    Logistics Optimisation: LOGOS+

    Solving the Interdependent Challenges of Packing Fitness and Delivery Distance

    In global distribution, operational efficiency is a balancing act between conflicting physical realities. To maximise margins, supply chain leaders are constantly caught between two critical metrics: packing fitness (how densely items are secured within a vehicle) and delivery distance (the total mileage of the transportation route).

    For any commercial fleet, the core objective is simple: deliver inventory from a central hub to multiple destinations at the absolute lowest total cost. Yet, when executed on an enterprise scale, material costs (pallets, vehicle overheads) and transport costs (fuel, driver hours) frequently pull operations in opposite directions.

    To achieve true capital efficiency, decision-makers can no longer afford to evaluate packing and routing in isolation. They must be solved simultaneously.

    The Costly Reality of Fragmented Models

    Traditionally, logistics architecture has relied on the Separated Packing and Routing (SPR) model. This legacy approach treats the Bin Packing Problem (BPP) and the Vehicle Routing Problem (VRP) as two entirely independent workflows managed by separate teams.

    While the SPR model offers administrative simplicity, it introduces two critical operational flaws:

    • Conflicting Corporate Objectives: Packing teams focus entirely on minimising fixed asset and pallet usage, while dispatch teams focus solely on reducing mileage. Without a unified framework, these goals directly undermine each other.
    • The 20% Cost Penalty: Because packing configurations are finalised before routing algorithms begin, vital destination and drop-off sequence data are completely excluded from the initial loading phase. This structural blind spot creates systemic delivery bottlenecks and highly sub-optimal routes.

    The Computational Bottleneck

    The obvious alternative is an integrated model powered by Mixed Integer Programming (MIP). When paired with advanced mathematical solvers like Gurobi, MIP models can guarantee a mathematically flawless, optimal solution.

    However, exact mathematical modeling suffers from exponential computational scaling. For an enterprise fleet managing 1,000 boxes across 100 destinations, calculating a perfect MIP solution could literally take years of computing time. In live logistics environments where windows are tight, this approach is commercially unfeasible.

    Introducing LOGOS+: Next-Generation Hyper-Heuristics

    To bridge the gap between mathematical idealism and real-world operational velocity, Decision Lab developed LOGOS+ (Logistics Optimisation System — Two-Level Capacitated Vehicle Routing Problem, or LOGOS-2S-CVRP).

    LOGOS+ is a proprietary hyper-heuristic framework that bypasses exponential computing delays. By combining two distinct algorithmic classes, it delivers near-optimal operational plans in seconds.

    1. Constructive Heuristics

    This layer builds a highly viable, baseline operational solution from scratch using two distinct, configurable strategies:

    • Volume-Driven (VD) Heuristics: Prioritises the minimisation of material and pallet costs. Items are systematically sorted and packed into vehicles based on customisable, high-scoring geometric constraints.
    A three-part scientific diagram demonstrating a step-by-step 3D bin packing problem solution, showing an open bounding box grid sequentially filled over three stages with colourful, tightly packed cuboid items to maximize spatial utility.
    LOGOS+ 3D Bin Packing Volumetric Space Mapping
    • Destination-Driven (DD) Heuristics: Prioritises transportation efficiency. The engine automatically clusters items by geographic proximity or pre-loads containers to perfectly align with the intended multi-drop sequence.
    A technical line chart plotted on an X-Y axis demonstrating vehicle routing clusters for three separate vehicles, labeled truck_0, truck_1, and truck_2, showing how delivery destinations are partitioned into distinct, optimised spatial zones to minimise transit distances.
    LOGOS+ Geographic Clustering and Vehicle Routing

    2. Perturbative Heuristics

    Once the initial solution is constructed, LOGOS+ deploys advanced local search and hill-climbing algorithms to eliminate hidden inefficiencies. The engine continuously optimises the fleet layout via four core operators:

    Packing Swap: Exchanging assets between two vehicles to achieve a superior volumetric fit.

    Packing Insert: Shifting an individual item to an alternative vehicle to maximise space utilisation.

    Degrading: Completely unloading a under-utilised vehicle and resetting its contents across the remaining fleet.

    Destination Swap: Altering the drop-off sequence within a single vehicle’s manifest to uncover a shorter, faster transit path.

    A technical workflow diagram demonstrating a local search optimization technique where a multi-drop delivery route originating from a central depot has its sequence modified by a destination swap between two stops to refine transit efficiency.

    The system iterates automatically, accepting a new layout only if it actively reduces total operational expenditure, until a refined local optimum is achieved.

    Proven Performance and Speed

    Through rigorous benchmarking against randomised enterprise datasets scaling up to 80 variables and 50 destinations, LOGOS+ demonstrated a massive leap forward in computational and financial performance.

    Optimisation MethodSolution QualityComputation TimeOperational Viability
    Separated Packing & Routing (SPR)Poor (Up to a 20% cost penalty vs integrated models)Very Fast (Seconds)High administrative ease, but poor financial efficiency.
    Mixed Integer Programming (MIP)Perfect (100% mathematically optimal)Extremely Slow (30+ mins for just 10 boxes)Commercially unfeasible for live, large-scale operations.
    LOGOS+ (Hyper-Heuristic)Excellent (Averages within 10% of true mathematical optimum)Instantaneous (Processes 1,000 boxes across 100 drops in 6.52 seconds)Ideal for real-time, large-scale enterprise logistics.

    Strategic Business Impact

    LOGOS+ represents a significant paradigm shift for enterprise supply chains. Rather than breaking logistics challenges apart (packing individual pallets, loading vehicles, and planning routes in silos) our engine processes them within a single, unified pipeline.

    By eliminating the traditional 20% cost penalty associated with fragmented planning, LOGOS+ allows organisations to tailor their optimisation core to match specific corporate KPIs:

    • Decarbonisation & Sustainability: Configure the system to explicitly minimise total CO2 emissions, automatically influencing optimal vehicle selection and routing configurations.
    • Bottom-Line Maximisation: Dynamically balance fuel consumption variables against fluctuating driver labor rates, clean-air zone fees, and service level agreements (SLAs).

    What’s Next: The AI Frontier

    While we are actively expanding our suite of heuristics, our next frontier is total intelligent automation. Decision Lab is currently developing an Artificial Neural Network (ANN) layer designed for automatic algorithm selection. Once deployed, this AI capability will instantly evaluate incoming logistics profiles and automatically select the absolute best heuristic combination for that specific dataset, unlocking true hyper-optimality without human intervention.

    LOGOS+ is architected to serve as the high-performance optimization engine powering modern supply chain and enterprise resource planning software.

    To learn how to integrate this capability into your logistics ecosystem, contact us.

  • Logistics Optimisation System: LOGOS

    Logistics Optimisation System: LOGOS

    Solving the 3D Bin Packing Problem with AI and MILP

    The relatable anxiety of trying to fit an entire holiday wardrobe into a single 23kg suitcase mirrors one of the most financially demanding challenges in global supply chain logistics: the 3D Bin Packing Problem (3D BPP).

    While a consumer might face an extra airline fee for poor packing, the stakes for commercial shipping are exponentially higher. A single poorly allocated item on a pallet can force the use of an additional container, instantly doubling transport costs for that consignment.

    Despite these high stakes, human operators fill only 50% to 60% of available space on average, often requiring multiple manual attempts to achieve higher efficiency. To solve this, Decision Lab developed LOGOS (Logistics Optimisation System). By leveraging two distinct computational paradigms (mathematical optimisation and deep reinforcement learning) LOGOS consistently outperforms human capabilities in both speed and spatial utility.

    A 3D model visualisation of the LOGOS simulation environment solving a 3D Bin Packing Problem. The graphic features a transparent 100x100x100 grid container holding several semi-transparent, multi-coloured cuboid blocks (orange, green, brown, red, and blue) to demonstrate volumetric integrity, spatial boundaries, and item allocation.

    Defining the 3D Bin Packing Challenge

    To evaluate the effectiveness of different digital solutions, we established a standardised testing environment. The core objective is to pack a varied sequence of cuboid items into the minimum number of containers while maximising spatial utility.

    For our baseline study, we used a 5x5x5 container environment packed with items ranging in dimensions from 1x1x1 to 3x3x3. Container efficiency is precisely calculated using the following utility formula:

    Utility=∑Volume of items packedVolume of container\text{Utility} = \frac{\sum \text{Volume of items packed}}{\text{Volume of container}}

    To find the optimal operational approach, we developed and compared three distinct computational methods:

    • Mathematical Optimisation (LOGOS-OPT): A deterministic approach that searches the entire problem space to calculate an absolute mathematical optimum.
    • Deep Reinforcement Learning (LOGOS-RL): An AI-driven approach where an autonomous agent learns highly adaptive packing strategies through continuous trial and error.
    • Rules-Based Algorithm: A standard procedural heuristic used as a comparative baseline for traditional industry software.

    Approach 1: Mathematical Optimisation (LOGOS-OPT)

    Our mathematical framework formulates the 3D BPP as a Mixed Integer Linear Programming (MILP) problem. This method assumes full information is available upfront, allowing the system to map out the entire packing sequence simultaneously.

    Driven by the commercial Gurobi solver, LOGOS-OPT maximises total container utility while strictly enforcing real-world physical constraints:

    • Spatial Boundaries: Items must remain entirely within the physical dimensions of the container.
    • Volumetric Integrity: No physical overlap between items is permitted.
    • Rotational Logic: Items can be orthogonally rotated, though specific orientation restrictions can be applied to individual assets.
    • Vertical Stability: To prevent damage or structural collapse, the bottom face of every item must be structurally supported by the container floor or the flat surface of items beneath it, verified across all four base vertices.

    Approach 2: Deep Reinforcement Learning (LOGOS-RL)

    In real-world logistics hubs, operations are rarely static. Items arrive sequentially on a conveyor belt, demanding immediate, real-time placement decisions with highly imperfect future data. This dynamic, stochastic environment is where Deep Reinforcement Learning (DRL) excels.

    Unlike LOGOS-OPT, LOGOS-RL does not have an upfront list of items; it evaluates items sequentially, holding visibility of only one step ahead on the conveyor line.

    A 3D digital simulation of a logistics conveyor belt used to train a deep reinforcement learning model. Various multi-coloured cuboid items of differing heights and dimensions, including green, blue, orange, teal, light blue, pink, and purple boxes, are spaced sequentially along the dark grey conveyor track moving toward an empty packing bin at the end. The visual demonstrates the stochastic, item-by-item queueing system handled by the LOGOS-RL agent.

    The AI Training Architecture

    To build a highly capable neural network, we paired advanced simulation with automated cloud training:

    • Simulation Environment: Built using AnyLogic, the model tracks advanced states, including a dynamic height map matrix, container utility trends, and real-time feasibility maps that serve as an action mask.
    • AI Training Engine: We utilised machine teaching hosted on Azure to scale training across millions of parallel iterations. The system automatically selected the optimal deep reinforcement learning algorithms (such as SAC, Apex DQN, or PPO) based on environmental feedback.
    • Reward Function: Rather than relying on a delayed end-of-game reward, the agent received immediate positive reinforcement proportional to the change in container utility, alongside a steep penalty and immediate episode termination for invalid placements.
    A technical flowchart mapping the standard Reinforcement Learning (RL) feedback mechanism. Two main dark blue nodes are labelled 'Agent' and 'Environment', linked by pink directional paths. An arrow flows from the Agent to the Environment representing a specific action, labelled A sub t. Two sequential feedback loops return from the Environment to the Agent, representing the environment's current state, labelled S sub t, advancing to the next state, S sub t plus one, and the generated reward, R sub t, advancing to R sub t plus one. This illustrates how the AI continuously evaluates and adjusts its spatial decisions.

    Continuous curriculum learning allowed the agent to master basic 1x1x1 packing across 318,000 iterations before successfully mastering highly variable multi-dimensional item packing over 7 million training iterations. Once fully trained, the brain is exported into a lightweight Docker container for instant, on-site deployment.

    Benchmarking Performance: Human vs. Machine

    To validate LOGOS in practice, we executed two series of rigorous live trials comparing manual human efforts against our automated systems.

    Trial 1: Standard Pallet Allocation (17 Items)

    • LOGOS-OPT: Generated an optimal packing configuration in 13 seconds, achieving a 75.8% packing density. Using the LOGOS Visual Assistant (a digital guidance system that projects exact placement instructions), the physical packing took just 40 seconds.
    • Unassisted Human: Averaged 90 seconds across multiple attempts, only managing to load 12 to 15 of the items, resulting in a significantly lower spatial density between 55% and 70%.

    Trial 2: High-Complexity Pallet Allocation (19 Items)

    • LOGOS-OPT: Calculated a complex packing arrangement in 105 seconds, yielding a 73.5% packing density that allowed the physical operator to safely pack all 19 items in 50 seconds.
    • Unassisted Humans: The increased complexity highlighted a severe performance gap. One experienced packer gave up entirely after 5 minutes. A second achieved only 60% density, leaving 5 critical items completely unpacked. The final human packer managed to fit all items but required 10 full minutes, which is four times longer than the combined computing and packing time of LOGOS.

    Comparative Results: Finding the Operational Sweet Spot

    When evaluating the three algorithmic models across multiple randomised arrival sequences, Deep Reinforcement Learning proved to be an incredibly formidable alternative to perfect mathematical models.

    MetricMathematical Optimisation
    (LOGOS-OPT)    		Deep Reinforcement Learning (LOGOS-RL)Standard Rules-Based Heuristic
    Exp 1: Packing Density93.5%83.2%75.0%
    Exp 1: % of Theoretical Max100%89.0%80.0%
    Exp 2: Packing Density88.35%80.4%74.8%
    Exp 2: % of Theoretical Max100%91.0%84.0%
    Exp 3: Packing Density93.6%80.5%71.8%
    Exp 3: % of Theoretical Max100%86.0%77.0%

    Key Strategic Insights

    1. The DRL Advantage: Despite having no knowledge of future item arrivals, LOGOS-RL consistently achieved 86% to 91% of the mathematical gold standard, comfortably outperforming standard rules-based algorithms. In fact, during Experiment 2, the RL agent matched the absolute mathematical optimum a staggering 55% of the time.
    2. Infinite Scalability: Traditional mathematical optimisation complexity scales exponentially, often requiring complex problem decomposition to avoid computation time explosion when managing large fleets. Conversely, LOGOS-RL delivers near-instantaneous decision-making responses and scales linearly across an infinite number of containers without any added computational overhead.
    3. Real-Time Resilience: Because the DRL framework processes decisions item-by-item, it is uniquely suited for live logistics environments where orders change mid-route, or stock availability suddenly shifts.

    Transforming Fleet Economics

    By bridging the gap between theoretical data science and warehouse floor execution, the LOGOS framework provides logistics enterprises with a tangible path toward maximising fleet utilisation, reducing fuel consumption, and eliminating systemic supply chain waste.

    To learn more about deploying LOGOS within your logistics infrastructure, contact us.