Supply Chain Lab

Published research showcase - prioritised two-dose vaccine allocation & distribution, with real case-study figures.

Published research - interactive Gippsland evacuation network with a time-step scrubber and a contraflow toggle. Transportation Research Part A (2024).

Published research - interactive multi-agency coordination of aircraft, ground crews and water for the 2009 Black Saturday fires. IJDRR (2021).

Published research - interactive Greater Melbourne coverage map with a live dispatch drill; 22 new stations lift 5-minute coverage 81%→95%. Fire Safety Journal (2020).

Open Day live demo - run a 174-day two-dose vaccine campaign for Greater Melbourne under limited, ample or disrupted supply and watch coverage respond.

Slide-style mock-lecture story for Open Day - the vaccine DSS as an 8-step scroll story with the real day-by-day output, a live simulation and the Vaccine Rush phone game QR.

Open Day mobile game - two missions on one live city map: split 100k scarce doses across risk districts, then route cold-chain vans under a 60-minute clock. Beat the exact-optimal algorithm at 21,500 points.

Profit-maximising tanker scheduling and spring-water sourcing on Victoria's real road network - owned, leased and spot sources, PBS fleet limits, loading-bay queues and driver shifts, solved live in the browser, with source-outage and 200-day Monte-Carlo stress tests.

Solve capacitated VRP with PuLP - assign vehicles to customers and minimise total route distance on an interactive map.

Compute the optimal facility location that minimises weighted distance to demand points.

Ship goods from factories to stores at minimum cost, optionally routed through hubs, with flow lines weighted by volume.

One structural shock - earthquake, port closure, DC blackout - propagates downstream like a wave, burning buffer after buffer. Re-route and mitigate on a limited budget to keep the shelves alive.

Knock warehouses offline or drop flood and bushfire risk zones, then re-route supply to the nearest working site.

Map demand across suburbs as a live heatmap, then reveal the demand-weighted centre of gravity.

Drop demand points and candidate facilities, set a service radius, and open the best few sites to cover the most demand within reach.

Build an end-to-end chain - supplier to factory to distribution centre to retailer - and see the flow of goods.

Replenish the most urgent low-stock customers first with a capacity-limited truck fleet, drawn as depot loops.

Score global suppliers on cost, lead time, quality and geopolitical risk, and re-rank them live by sliding the weights.

Staff a DC for one week - 7 days × 3 shifts against a demand profile. Wages vs service penalties, weekend rates, and one deliberate trap: the shift that costs more than the gap it fills.

Size a distribution centre from product families and pallet footprints, and see utilisation and layout respond live.

Seven agent-based and discrete-event models on AnyLogic Cloud in one place: the Beer Game, bullwhip, supply-chain risk and disruption, drone delivery, AGV dispatching and reinforcement learning.

Compete across five mobile customer orders. Balance cost, delivery speed and stock health, then compare individual scores on the live class leaderboard.

Demand signals rise across Melbourne suburbs. The model forecasts demand, reallocates limited inventory to the hottest areas and flags which store to replenish.

Explore the Gaussian distribution interactively - adjust mean and standard deviation to see the curve respond.

Model discrete arrival events and rare occurrences with an interactive Poisson probability explorer.

Run Monte Carlo experiments - sample random distributions and observe convergence and uncertainty.

Five-tier mobile Beer Game with automatic seating, QR join, Chaos vs Collaboration rounds, team competition and Bullwhip awards.

Optimise 3D container packing - maximise space utilisation while respecting weight and stacking rules.

Up to 36 students, six five-tier chains, one market. Retailers bid a retail price every week - cheaper wins a bigger slice of demand but a thinner margin. Live multiplayer.

Stocks, flows, delays and feedback in one live Sterman-style model. Ignore the supply line and a 30% demand step becomes months of oscillation - the bullwhip, born.

Operate tier-2 suppliers, subcontractors, a factory, two DCs and three markets through border closures, plant shutdowns and demand shocks. Protect service and cash.

How Arena and AnyLogic really work: a future event list and a clock that jumps event to event. Step a truck dock one event at a time, then check M/M/c theory.

Dispatch 1,290 real load and discharge jobs across nine quay cranes, four yard blocks and a pooled truck fleet. Compare AS-IS with two AnyLogic-inspired control policies.

Fold back a capacity decision under uncertain demand: EMV at chance nodes, options priced by backward induction, EVPI for what a perfect forecast is worth.

Simulated annealing, a genetic algorithm and ant colony race the same travelling-salesman map - temperature, generations and pheromone trails drawn live.

Lumpy demand breaks EOQ. Watch the exact dynamic program pick which months get an order, beat lot-for-lot and Silver-Meal, and show its whole cost matrix.

Eight warehouses, two inputs, one output: Data Envelopment Analysis builds the best-practice frontier, scores every site and names the peers to copy.

Goldratt’s dice game from The Goal: five dependent stations each average 3.5 a round, yet the line never ships 3.5. Find the constraint, then fix it with Drum-Buffer-Rope.

Every doubling of output cuts unit labour to a fixed percentage. Tune Wright’s curve, race a rival factory, and price against tomorrow’s cost, not today’s.

A machine wears toward failure. Value iteration derives a threshold repair policy - the maths under the RL agent, but with the model known. Checked against all 32 policies.

Book capacity now, demand lands later. Planning on average demand feels sensible and is quietly wrong - priced here as the Value of the Stochastic Solution.

No probabilities, only honest ranges. Turn the uncertainty budget Γ and watch the price of robustness: a few percent buys most of the protection.

Drivers to routes, one each, least cost. Step Kuhn’s algorithm through row reduce, covering lines and adjustment - O(n³) instead of n!, verified against brute force.

An LP has one objective; a manager has several that fight. Soft goals and deviation variables give you the least-bad plan instead of the word “infeasible”.

Weibull lifetimes, redundancy, optimal replacement age and Poisson spares. Turn β and watch preventive maintenance become completely pointless - you can’t prevent random.

Maximin, maximax, Hurwicz, Laplace, minimax regret: five defensible criteria on one payoff table that routinely pick five different answers. Your temperament is the model.

Optimise a supplier award across total cost, quality, service, resilience and Scope 3. Stress-test concentration, read the efficient frontier and defend the recommendation.

Allocate six supplier awards, then reveal shared demand, port, quality, carbon and geopolitical shocks. Host by QR and compete on the live class leaderboard.

Balance Melbourne fulfilment capacity, labour and last-mile couriers while live Australian orders move across the network.

Control ordering policy from Kwinana to Sydney and see delay, batching and information sharing amplify or suppress variability.

Operate coastal receipts, refineries, terminal inventories, road tankers and retail markets on a geographically aligned Australian map.

Run inbound docks, pallet storage, picking and truck despatch in a live Truganina distribution-centre operation.

Route intermodal trains through seven switch decisions, compare policies and train a reliable Dynon yard controller.

Tune robot cells, AGV fleet size and material replenishment in a Western Sydney palletising digital twin.

Solve max-flow across a logistics network on a real map. Line thickness shows optimal flow; drag facilities, edit rail/road capacities and compare train vs truck.

Staff an automated fulfilment centre shift by shift against a rising order wave - an M/M/c queue you can feel.

Open berths as a vessel surge hits; demurrage compounds on every ship left at anchor. Clear the queue before it buries you.

Dispatch a courier fleet across the city and find the density sweet spot before routes overrun the delivery window.

Design a reverse-logistics loop for returns, repair, remanufacture, resale and recycling without killing the margin.

Forecast demand period by period against an AI opponent. Capture the season, and do not chase the one-off anomaly.

Tune an alert threshold on a noisy stream to separate real disruptions from false alarms - precision vs recall, live.

Build an AI-ready data pipeline and test how freshness, coverage and quality drive forecast accuracy and trust.

Turn a planning problem into a model, then read the geometry - the best plan always sits on a corner of the feasible region.

Two live solvers in one - container loading (integer knapsack) and supplier-order assignment - with greedy-vs-optimal comparison, a sensitivity chart and a permutation explorer.

Graphical two-variable LP: watch the feasible region flex and the iso-profit line sweep to the optimal corner, with live shadow prices and RHS ranges.

Profit vs service on one feasible polygon - animate the simplex walk corner to corner while the weight slider traces the Pareto frontier in objective space.

Live forward/backward pass on a ten-activity project - earliest/latest times, slack, the red critical route, a Gantt strip and a crashing cost curve.

Race naive, moving average, exponential smoothing, Holt and Holt-Winters on one demand series with a live MAPE / MAE / bias leaderboard.

Newsvendor double marginalization: integrated optimum vs wholesale, buyback and revenue-sharing contracts, with a live chain-efficiency gauge that hits 100% when coordinated.

2D bin packing on an AU-standard pallet - the FFDH shelf heuristic races a guillotine search with rotation, with live utilisation, a dock strip for what won't fit, and mix sensitivity.

Assign nine precedence-linked tasks to workstations against a demand-driven takt time, then race the RPW heuristic on efficiency and smoothness.

Twelve months of demand, three workforce plans - Level, Chase and yours - simulated on identical costs with a live leaderboard and a one-line diagnosis of why the winner won.

Six custody blocks sealed with real SHA-256 hashes - fake a temperature reading, watch the chain break downstream, then pay the proof-of-work price of hiding it.

Six organisations, one vaccine consignment, one shared tamper-evident ledger. A DSS scores the risk from the chain - but immutable data can still be wrong data. You make the call.

FSTSP-lite: a 2-opt truck tour plus one exactly-enumerated drone sortie. Drag the customers, tune range and speed, and read the completion-time saving live.

Paste or upload YOUR own CSV of demand points and watch a live twin run it - stochastic demand ticks, a weighted-centroid hub, live service and utilisation. Right-size the hub to your own dataset.

An AI agent runs your replenishment for 26 weeks - you only set its policy: autonomy threshold, caution and reactivity. Survive the promo without drowning in human approvals.

Run the whole company for 8 weeks - source, produce, ship and sell into one cumulative profit score. Play turn-based (relay) or commit whole weeks, survive the promo spike, and beat the class leaderboard.

Tabular Q-learning on a Poisson-demand inventory problem - watch the Q-table sharpen, the cost curve converge to a tuned (s,S) benchmark, and quiz the agent mid-training.

Edit a master schedule and watch gross-to-net requirements cascade down a four-item bill of materials - lot-for-lot vs FOQ, lead-time offsets and past-due warnings.

Push vs kanban vs CONWIP on a four-station line: watch WIP explode or stay capped, and verify Little's Law live while throughput barely moves.

A live bottling line sampled onto X̄ and R charts - inject a mean shift or extra variance, see which chart catches it, and learn why control ≠ capability.

Assign six inbound trucks to three dock doors against arrival times - your schedule vs the greedy rule vs the exact enumerated optimum, on a live Gantt.

One DC feeding three stores, simulated over 1,500 days - split the same total stock between echelons and find where inventory should live.

Find the cheapest route through a logistics network step by step with Dijkstra's algorithm - the engine inside every routing tool.

Score candidate facility sites on weighted criteria and compare locations side by side - the judgement step before the optimisation models.

Bid to win supply contracts in a sealed reverse auction. Shade your bid: win probability against margin.

A retro platformer quiz on Weeks 7, 8 and 9. Bump ? blocks for generative AI, process mining and warehouse automation, beat the Disruption Dragon, then screenshot the evidence card for A2 and A3.

Classify spend into the Kraljic matrix, choose sourcing plays, then stress-test the board with executive shock cards.

Choose Incoterms, payment terms and clauses, then watch cost, risk transfer and cash exposure move across a trade lane.

Balance demand, capacity, expedite and inventory across a six-month plan as role pressure and service/cash KPIs update live.

Train a supplier ranking model, detect proxy bias, apply mitigation, and compare value, fairness, compliance and trust.

Keep reefer shipments in-band through heat events and door-opens. Trade energy against spoilage; sensor sampling catches excursions before product is lost.

Set the delivery promise your storefront advertises and balance the conversion it wins against the on-time performance and penalties it demands.

Deploy AI agents to run logistics autonomously. Tune autonomy and confidence gates, keep humans in the loop, and capture savings without over-automation incidents.

Build traceability so a recall implicates the smallest slice of product. Capture more tiers, raise granularity and enable a ledger to narrow recall scope.

Run eight uncertain shifts. Allocate Barcode, QR, RFID and manual capacity under a budget, absorb disruptions, adapt from operational results and compete on cumulative profit.

Compete across ten rapid business situations. Match Barcode, QR, RFID or manual entry to the operational need, receive instant teaching feedback and climb a separate live class leaderboard.

Eight live team rounds mix rapid choices, calculation, debate, confidence, disruption and an animated readiness simulation. One device per team, one connected score.

Two teams, the same five disruptions. One lets the optimiser decide alone, the other keeps a human who can overrule it. Service, resilience and trust move together, and the flood round asks whether the cheapest answer is the one you would defend.

Pick a Supply Chain 5.0 module and the engine writes a fresh case, then examines you Socratically and marks you against five criteria. Every case is generated on request, so there is nothing to copy from a classmate or a chatbot.

Street-arcade combat quiz. Defeat ten IoT supply-chain bosses with ten hard questions on MQTT, LPWAN, RFID physics, edge computing, EPCIS and device security - answer right to land the knockout, answer wrong and take the hit.

Ten street fights, ten levels of IoT. Beat each opponent by answering on what the Internet of Things actually is, the information value loop, enabling technologies, cloud and analytics, standards, privacy and responsible design. Two questions per fight, wrong answers cost health.

Complete the five-scene campaign, then turn the result into an A2-ready data-to-decision design linking evidence, an A1 KPI, analytics, platform, action, value, risk and human control.

Weeks 7 to 9 in one seeded game: choose analytics, ML, DL or GenAI cards, set rules and approval gates, then commit printers, cobots and AGVs. Every period freezes a twin prediction before execution and prints a screenshot-ready result screen. How to play: sc5-twin-run-guide.html

Run a six-week 3D printing cell. Match FDM, SLA and SLS capacity to uncertain orders while managing materials, labour, quality, maintenance and cash.

Coordinate a swarm of autonomous mobile robots. Grow the fleet for throughput until congestion and charging eat the gains and the marginal robot hurts flow.

Score alternatives on cost, quality, lead time and sustainability. Set weights, switch between weighted-sum, TOPSIS and PROMETHEE, and watch ranking and rank reversal respond.

Sell a fixed inventory over a season under price-sensitive, uncertain demand. Let dynamic repricing react to what is left and maximise revenue without stocking out early.

Tune an e-commerce recommender across popularity, item-CF and hybrid. Balance relevance against catalogue coverage and novelty, handle cold-start, and avoid the filter bubble.

Move freight across air, road, rail and sea. Balance cost and service against Scope 3 emissions under a carbon price and an emissions cap.

Compare criteria two at a time on Saaty’s 1–9 scale; derive priority weights and a consistency ratio (CR), then feed them into the MCDM Studio.

Multi-issue integrative bargaining (ZOPA / BATNA) against an AI supplier - find the zone where both sides say yes, then claim your share.

Machine learning (ML) toolkit - regression, logistic regression, forecasting, k-means clustering, decision trees and random-forest / boosting ensembles on supply-chain data.

Run a 12-day Melbourne vaccine cold chain. Compare exact dynamic programming, approximate dynamic programming and a myopic policy as the value function learns.

Form Melbourne–Sydney–Brisbane resilience coalitions, allocate savings with Shapley values, and keep every partner inside the cooperative-game core.

Learn reinforcement learning by moving a forklift through a live fulfilment cell. See state, action, reward, Q-values, exploration and policy improvement happen together.

Match stochastic, fuzzy, robust and scenario methods to eight Australian evidence cases, then compare buffer, service and cost under the same hidden outcomes.

Operate a cold-chain risk model for 60 days. Distinguish noise from drift, inspect accuracy, recall and PSI, then investigate or retrain with a governed audit trail.

Five live teams run procurement, demand, production, logistics and finance/ESG across an eight-week Australian network. Decisions combine, disruptions hit, and one shared score reveals the trade-offs.

Watch a roaster's morning automation run node by node: read stock, sales and the supplier message, let a simulated assistant recommend an order, then act and pay with a person approving. Inject a supplier delay, a stale feed or a duplicate invoice, repair it and compare with manual work.