Switch on to digital advances for faster warehousing

How AI and edge computing are speeding up warehouse logistics

In a follow-up to his article on visibility and artificial intelligence (AI) in the broader logistics chain, Ruari McCallion takes a closer look at the growing role of AI and ‘edge computing’ within the warehouse walls.

(Executive Summary)

Warehouse and logistics operations are under simultaneous pressure from a range of issues: labour shortages, rising energy costs, space constraints and increasingly volatile demand. At the same time, digital technologies that were once experimental, such as artificial intelligence (AI), machine learning and edge computing, are now mature enough to sit directly in the flow of materials handling.

This doesn’t mean that all warehouses and inventory operations, of whatever size, are suddenly going to become ultra-automated and shiny-floored supermarkets of supply, with robot platforms gliding around while small messenger robots scurry about like terriers. This isn’t science fiction; it’s the real world.

“AI-driven warehouse automation: A comprehensive review of systems” – a report published by GSC Online Press in February 2024 – identified four technological developments that are bringing about a structural shift in warehouse operations and management:

• AI integration in warehouse management
• Machine learning and computer vision
• AI-driven automation systems
• The rapid emergence of edge computing

Stand‑alone vehicles and fixed mechanisation are being replaced with connected, sensor‑rich, software‑orchestrated systems in which lift trucks, people, robots and storage all interact, in real time.

It’s happening at different speeds, depending on the size and intensity of the operation. Not everyone is ready to go full ‘I, Robot’ right now – and some may never be. But it is happening, with inevitability and growing momentum, as the value of the novel technologies becomes clearer and (crucially) commercially attractive.

Potential gains in warehouse productivity and throughput

Productivity gains in warehouses have, historically, come primarily from mechanisation: better racking, more capable lift trucks, and the implementation of conveyor systems and automated storage and retrieval systems (AS/RS). Control systems have been rule‑based and relatively static. Once designed and installed, flows were stable and changes were expensive. As a result, gains have been incremental and gradual, for decades.

Two things have shifted that model.

First, the data now available from operations has exploded. Trucks, conveyors, scanners, sensors, cameras and warehouse management systems all generate continuous streams of information. Second, AI and edge computing make it possible to process that data locally and centrally, enabling decisions based on real-time data instead of historical reports. It’s the Fourth Industrial Revolution (4IR) in action.

What are edge computing and AI?

Edge computing is a reversal of the established trend to take processing functions away from frontline operations, typically to the cloud.

Large computing power offered by offsite servers is a huge advantage in strategic planning, for data storage and for coordinating multi-site operations, across countries and internationally. In the automotive industry, for example, it has cut new vehicle development time from years to just a few months. However, it comes at a cost and that is slower reaction time – ‘latency’, in the technical jargon.

Data moves very fast along cables and even through the air – at light speed, in fact – but it still takes time. It takes time to be sorted, as well, as even the most powerful server farm will prioritise the tasks it’s going to perform. At the user’s end, that latency manifests in the frustration of a screen featuring a rolling beach ball, or an apparent freeze, or a delayed response to a routine enquiry.

But computer chip capability has mushroomed enormously over the past 20 years and edge computing delivers processing power to where work is actually done. Instead of sending all data to a remote cloud for analysis, critical algorithms run on-vehicle controllers and embedded processors and intelligent devices, such as onboard and fixed cameras, sensors and gateways, as well as local servers.

Click here for a summary of previous Eureka articles on warehouse automation.

Processing very close to the point of activity means lower latency: raw data streams, from cameras, LiDAR (light detection and ranging) and sensors, can be handled locally. This means that functions that cannot tolerate network delays, such as navigation, braking, obstacle detection and collision avoidance, become more reactive, reliable and resilient. On-site operations can continue safely, even if external connections are compromised.

By processing data at the edge, systems can forward off-site only those insights needed for fleet‑wide optimisation and long‑term analytics. Everything that can be, is handled locally.

Intelligent materials handling enables warehouse management and control systems that optimise routes, priorities and labour allocation. New technologies enable dynamic adaptation in real time, rather than adherence to fixed rules based on historic data.

Forklift trucks are no longer just vehicles that move pallets. They are mobile, networked assets that feed information into – and take instructions from – a wider optimisation layer. They become intelligent nodal points in a larger system.

Autonomous mobile robots (AMRs) and automated guided vehicles (AGVs) that navigate freely rather than following fixed paths are being seen in the shape of automated storage and handling equipment that can rebalance workloads, change routing and adapt to congestion, instantaneously. Human operators, rather than being distracted by bottlenecks, blockages and misplaced inventory, are supported by assistance systems that respond to their behaviour and environment in real time.

Monetising technological gains in warehouse management

Cat electric forklifts, particularly the latest, Li-ion battery-powered generation, already incorporate a lot of onboard intelligence. Battery management systems (BMS), for example, continuously monitor cell voltages, temperatures and currents, as well as the battery’s charge level and state of health generally. Other monitoring systems deal with driver profiles, acceleration and braking patterns, for instance. These are all key indicators that many people will be familiar with from OEE (overall/operational equipment effectiveness) toolsets. AI makes OEE management better, clearer and more integrated.

AI in operational optimisation (deciding which pallet to move next; which truck will move it; along which route) takes into account congestion, operator availability, forklift condition and other priorities that fixed rules struggle to match.

AI enables better forecasting and inventory management by using signals such as seasonality, promotions and customer behaviour, rather than relying primarily on historical averages. Computer vision systems can read labels and barcodes at high speed, detect damage on products and/or packaging, check consignments for accuracy and support automated exception handling.

The integration of AI and edge computing into warehouse management systems delivers a more adaptive flow of goods. Decisions are based on data in real time, rather than on historical data that’s out of date even before it’s presented to a planning meeting. It means actions, reactions and control happen in genuine real time.