Advancing AS/RS with Digital Twins and Artificial Intelligence – webinar insights

Monday, January 20th, 2025

On 14 January 2025, ALICE hosted the first session of the TG3 webinar series on Digital Twins and Artificial Intelligence. The webinar “A Digital Twin for Automated Storage and Retrieval Systems (AS/RS): implementation, Digital Modelling and the role of Artificial Intelligence” explored solutions to improve the efficiency and adaptability of AS/RS in the face of growing e-commerce demands and supply chain complexity.

Welcomed by Giuseppe Luppino, led by Andrea Ferrari from Politecnico di Torino and moderated by Anike Murrenhoff (Fraunhofer IML) and David Cipres (ITAINNOVA), the session provided a detailed overview of how Digital Twins (DTs), integrated with AI, can optimise warehouse operations, reduce lead times, and improve overall system performance.

Here are some of the takeaways

Digital Twins in AS/RS: a new approach to logistics optimisation

Digital Twins are digital replicas of physical systems that use real-time data exchange to simulate, predict, and optimise operations. They can have different levels of automated data flows (Digital model, Digital shadow or Digital twin) and bring particular value to AS/RS by enabling:

Real-time decision making: DTs process real-time data from Warehouse Management Systems (WMS) and Warehouse Control Systems (WCS), enabling instant feedback and action.
Scenario analysis: They provide a risk-free environment to simulate operational scenarios, such as peak demand days or unexpected disruptions.
Order sequencing: Optimise the order picking sequence to minimise lead times and improve efficiency.

By integrating conceptual models, simulation tools, and optimisation algorithms, DTs can address key operational challenges, particularly in fast-paced environments such as e-commerce and omni-channel logistics.

AI and Digital Twins: a complementary relationship

The scope of the webinar was to highlight the synergies between Artificial Intelligence and Digital Twins:

AI for speed: Machine learning models such as neural networks can quickly predict outcomes (e.g., order processing times), enabling fast decision making.
Digital Twins for precision: While AI provides speed, DTs provide a higher level of accuracy by simulating complex operational scenarios in detail.
Integration benefits: Using AI to enhance DTs results in a scalable, adaptive system that balances speed and accuracy.

A specific use case shared during the webinar demonstrated the effectiveness of this approach in solving the order sequencing problem. An Artificial neural network (layers of interconnected nodes that process data, enabling tasks such as prediction and classification) trained with synthetic data from the DT enabled rapid evaluation of order sequences, drastically reducing computational time compared to traditional simulation methods, which, from the other side, is more accurate.

Practical lessons from research and implementation

Andrea Ferrari’s presentation offered actionable insights for companies looking to implement DTs:

Development process:

System mapping: Before building a DT, it’s critical to have a deep understanding of the physical system, including key parameters, behaviours and constraints.
Modelling levels: Select an appropriate level of model abstraction. Over-modelling can result in unnecessary complexity and slow processing times.
Data management: Clean, structured data is vital for training machine learning models and ensuring effective feedback loops.

Operational benefits:

Efficiency gains: DTs enable continuous optimisation of AS/RS, improving utilisation rates, reducing energy consumption, and increasing throughput.
Adaptability: By simulating potential disruptions, DTs empower organisations to adapt quickly to unforeseen changes.
Scalability: These systems can be scaled across multiple warehouses, improving network-wide efficiency.

Challenges:

Data integration: Establishing seamless data flows between physical systems (e.g., WMS, WCS) and the DT requires significant initial investment and collaboration.
Training AI models: Training and validating neural networks is a time-intensive process but essential for accuracy and reliability.
Industry adoption: While DTs are increasingly popular, many implementations remain confined to laboratory environments, with scalability to industrial settings still a challenge.

The potential of DTs extends beyond individual warehouses to the wider supply chain. When combined with data sharing among stakeholders, DTs can:

Predict and mitigate supply chain disruptions.
Support collaborative decision making by providing a common operational view.
Enable dynamic optimisation across interconnected systems, improving overall supply chain performance.

However, effective collaboration requires overcoming barriers to data exchange and trust among partners, as well as standardising data formats and interfaces.

Back to overview