D.I.D.O

ELSA: Machine Learning & Simulation Classification Platform

ELSA (Everything Lives Somewhere Appropriate) is a Machine Learning Platform that’s role is to classify SKUs into the most appropriate types of pick-face apertures. We do this by adopting a banding system, ranging from Band 01 (very small, volumetrically slow moving) to Band 06 (very large, volumetrically fast moving).

CLARA: Ultra-Quick, Artificially Intelligent Design Bot

CLARA (Creates Layouts Algorithmically And Redesigns Appropriately) is an articial intelligence platform that has been trained to create warehouse designs following the same methods and logic as human designers. However, CLARA is able to create complex layouts in milliseconds, and can design 1000s of layouts in the same time it takes a human designer to create 1. This supports the concept of the multiverse.

• Average Design creates in <<1s
• Trained to follow same logic as human designers
• Higher levels of Accuracy and Consistency
• Will try 1000s of iterations to produce optimal result

ELISA: Simulation & Comparison Tool

ELISA ensures that the layout is sufficiently agile. Its role is to provide feedback to DIDO.

• Simulates the layout created in CLARA
• Scores the design based upon strategic data input
• Provides feedback to DIDO, who then makes suggestions on changes to ELSA
• Process then repeats

The Process

In the first phase, Data Capture, we ‘train’ DIDO using a variety of data to achieve two critical objectives. Firstly, a “big” operational dataset provides raw information that allows DIDO to understand the operation in detail. Secondly, answers to strategic questions are converted into a numerical system that prioritises and ‘weights’ the strategic objectives of the warehouse.

In phase two, Conceptual Design Iteration, DIDO orchestrates three daughter AI platforms to analyse, design and evaluate thousands of potential warehouse layouts.

The first platform, ELSA, uses machine learning to classify every SKU into six bands of products that have similar picking location needs. You may think of this as a very sophisticated ABC analysis that ensures that we place fast-moving products into larger apertures and slow-moving ones into smaller apertures.

DIDO then passes the data to CLARA, an AI design bot trained to quickly and accurately design warehouses. CLARA develops a layout appropriate to ELSA’s analysis before giving that design to ELISA, who simulates it.

Once the simulation is complete, ELISA calculates performance metrics and combines this with strategic objective ‘weightings’ completed earlier to create a “Design Score” out of 100, indicating this system’s overall success.

Then the process starts again, albeit this time a different layout is created by CLARA, which generates a new score, and so on, until the highest scoring system – the optimal design – has been identified. This optimal design is then turned into a precise virtual twin, which we can visualise in Virtual, Augmented or Rendered Reality using SEC XR, our Extended Reality engine.

Ultimately, through the combination of Machine Learning, AI Simulation and Extended Reality tech

nology, we can ensure that we deliver THE optimal design, ideally suited to an operation’s specific strategic and operational requirements.