How AI Is Driving Warehouse Automation

Implementing AI holds a promise for organisations looking to transform their distribution centres operations, but there are gaps in knowledge and expertise

A lot has been written about how artificial intelligence (AI) is changing how business is done. As a result, it wasn’t long before companies started to figure out how to utilise AI-enabled warehouse technology to get ahead of the competition.

The warehouse automation market has been steadily on an upward trend — forecasted to be worth $30 billion by 2026. The pandemic accelerated the pressure on supply chains and, in particular, the distribution centre. Virtually all companies are trying to improve their distribution effectiveness driven by growth in eCommerce, the high cost and scarcity of warehouse labour, and rising customer expectations for accuracy and service levels.

AI is a warehouse game changer, but organisations are struggling to use it optimally.

According to a study by Lucas System, executives are optimistic about AI — counting on quick and generous returns from their investment and expecting an average ROI of more than 60 per cent within five years — increasing productivity by 75-90 per cent, accuracy by 65-90 per cent and responsiveness by 60-90 per cent

Companies that are not actively investing in AI technologies will be missing out on these gains. ROI expectations from AI software are certainly high, but even more interesting is that senior executives are the most bullish on expected returns vs middle management.

But where in the distribution centres are companies looking to drive these kinds of returns using AI? Roughly half of companies rank reducing errors, reducing costs, making better decisions, and improving customer experience within the top three biggest benefits when implementing AI based software.

The fascination with AI technology and the large expected payoff is driving a great deal of enthusiasm with many companies. Over 90 per cent of respondents want to use advanced technologies such as AI and Machine Learning to drive warehouse and distribution centres performance improvements.

Knowledge and expertise gaps exist

Despite this optimism, 99 per cent of organisations say they face challenges to use AI more effectively, admitting that gaps exist in the knowledge and expertise required to achieve the full potential of AI. While many companies believe they are quite mature in warehouse automation, there are gaps in perception vs the reality of automation maturity.

Two-thirds of C-level executives and 39 per cent of senior management surveyed say they are using AI in their distribution centres, while only 30 per cent of middle management say they are using AI in the distribution centre.

About 80 per cent of respondents agree that there is a general lack of understanding of how AI-based software can be put into practice to enhance warehouses and distribution centres’ operations across multiple industries.

This gap is due to a few reasons: Perceptions of high costs compared to benefits; concerns about risks and control of operations decisions; cost and time for training and a lack of understanding for implementation. Nearly 90 per cent of respondents, regardless of industry, admitted to needing more expertise and information when it comes to implementation and use.

Implementing AI-based systems can have a profound effect on management effectiveness; safety and ergonomics; picking accuracy; labour costs; employee satisfaction; and throughput.

Dynamic Slotting

Optimal placement of products within the warehouse, or slotting, has a significant impact on all the warehouse key performance indicators – productivity, shipping and inventory accuracy, warehouse order cycle time, and storage density. Slotting is a difficult problem to solve with traditional approaches.

On top of that, there are typically thousands of products and slots involved. But AI can bring more to the table than better slotting results. It can lower implementation costs as it does not require a detailed CAD drawing of the warehouse, as is the case with traditional slotting software. Instead, with AI-based slotting solutions the spatial characteristics and travel time predictions can be automatically learned based on machine learning and activity-level data generated by modern work execution systems.

Workflow Orchestration

Warehouse operations consist of an assortment of systems and human labour, and therefore opportunities arise to optimise the orchestration of workflows between systems and labour. As an example, there is a real need for the orchestration of robots and people within warehouses.

Depending on the type of robot, there can be different use cases. Let’s consider autonomous mobile robots that serve as a multi-position cart for order picking. Without orchestration, a work execution system simply directs a robot to a location, then a nearby user logs into that robot and delivers one or more picks to the robot. After completing those picks, the user must find the next closest robot or perhaps query the system to be directed to a robot. As one can imagine, this process is not ideal.

With an AI approach, the system can automatically perform orchestration to optimise both the robot’s and picker’s time simultaneously. This is accomplished in part by machine learning based predictions about where the robots and pickers will be located in the future, as well as by a process called deep reinforcement learning whereby the system learns by experience the best policies to obtain the desired benefits – in this case how to make the best use of the picker’s time.

Workforce Planning

As labour is typically the largest operational cost of a distribution centre, workforce management plays a key role in cost-effective operations. Optimal allocation of workers to meet anticipated demand is essential to eliminate overstaffing and understaffing and reduce overtime while ensuring that orders get out on time.

For example, during a shift, a supervisor may need to make a series of decisions related to shifting personnel from area to area to meet shipping deadlines associated with a wave of orders. Good decisions along these lines mean the supervisor will need a variety of near-real-time data in order to be alerted to and be able to process that data in a way that leads to good decisions.

This is where an AI-based solution to workforce planning comes in. Machine learning can be applied at multiple levels, including predicting travel times between locations and predicting when waves of work will be completed by area – given the remaining work and current staffing levels in each area.

With those predictions, the AI solution can then run near-real time simulations to determine how to best complete the remaining work to ensure shipping deadlines are met.

Also Read: Adding AI to Supply ChAIn

Performance Management

Engineered Labour Standards (ELS) and labour management systems have been around for some time, but an AI approach to performance management is better because it is easier to implement and produces better results.

To implement ELS, industrial engineers typically go through a labour-intensive process that includes a combination of on-site observations, software calculations, benchmarking, and validations through actual practice. Implementing the AI approach is easier because it automatically learns the normal pace to complete tasks, including taking into account details such as expected travel time between point A and point B.

This learning is conducted on real-world performance data collected from within the operation, taking into account many variables. As a result, machine learning produced standards are more accurate than the ELS. And since ML models can be continuously updated, the ML approach will automatically adjust when operational changes are introduced.

Also Read: 5 Ways to Harness the Power of Website Personalisation

In-Warehouse Travel Optimisation

The time spent navigating a warehouse can account for more than 30 per cent of the total cost of labour. This makes optimising that travel time a critical component to running a successful warehouse or distribution centre. Complicating the matter is that minimising total travel time isn’t the only priority that needs to be taken into consideration.

The priority of each pick based on shipping deadlines can compete with travel minimisation and also evolves in its level of importance throughout the day. This problem has been solved using heuristic-based approaches for both batching and navigating the warehouse, namely FIFO and S-shaped pathing, respectively.

AI and machine learning based approaches can exploit the large amounts of data collected about this process in almost every warehouse management system. The system learns how to balance priorities to ensure we optimise for efficiency while still meeting all shipping deadlines with high confidence.

Additionally, by leveraging historical picking data, you can take into consideration common congestion areas and slow-moving routes to efficiently calculate the fastest pick path. In cases where customers leverage AI systems, the study found an average of 50 per cent labour cost savings while simultaneously increasing the probability of meeting outbound shipping deadlines.

It is clear that AI will be a key driver in the future of warehouse automation as there is massive potential for gains in increasing productivity, reducing labour costs, and improving customer metrics like accuracy and responsiveness.