Fleet Intelligence

Context

Despite the hype around Artificial Intelligence (AI), successful AI applications in real industrial environments are still scarce. This especially applies to a fundamental component in many Supply Chains – the Warehouse. The need for AI solutions to support warehouse operations becomes even more pressing considering the emergence of “hybrid” warehouse models where manual pickers and automated guided vehicles (AGVs) or autonomous mobile robots (AMRs) co-work in a shared floorspace.

Idea

The co-ordination of both these hybrid fleets within the warehouse relies exclusively on the experience of the shift Supervisor or Warehouse Manager, often utilising commercial Warehouse Management Systems (WMS). Co-ordination issues involve different aspects of order picking operations which range from order batching and sorting, vehicle dispatching and traffic management. Communication issues also exist as multi-vendor AGVs and forklifts fleet management systems must interface with the WMS. Through ongoing digitization of warehouses, there are now readily available communication channels between the vehicle elements and the warehouse management systems. These channels, however, are often exclusive by vendor and are often bespoke to a given facility application. Bespoke middleware solutions are usually required to make warehouse systems work.

Aim

This project will develop interconnected intralogistics vehicle fleet technology coupled with machine learning to automatically optimise and manage mixed fleets of both manually operated forklifts and material handling AGVs into existing digital and physical infrastructures. The ultimate target of FleetInt is to create an overarching picking management system that is vendor-agnostic and can easily integrate with different FMSs and WMSs while including AI elements for intelligent traffic management. While the data communication objectives will aim at standardising communication protocols between various warehouse management systems, the optimisation objectives will focus on selecting the best possible candidate route for a pick task so as to minimise the risk of traffic congestion and vehicle collisions or deadlocks. This will also impact alternative KPIs, such as order pick time, deadlock time and energy per pick.

Challenge

Challenges in this project reflect the challenges deriving from the mixed fleet warehouse model, which are multi-faceted:

1. The coordination of multi-vendor fleets that generally use distinct management systems;
2. The communication between warehouse management systems and multi-vendor fleet management systems;
3. The optimisation of picking operations across mixed fleets;
4. The unpredictability of the human drivers’ behaviour.

Outcome

The main break-through capabilities of the foreseen system include:

• Flexibility to work with multiple commercial WMS and AGV vendors while ensuring seamless data communication across distinct management systems.
• Possibility of managing various vehicles from multiple vendors in a single unified fleet through the development of an integrated fleet manager.
• Ability to leverage historical traffic information and data flow available through the WMS and FMS to make optimal routing decisions.
• Ability to learn human behaviour and improve its working internal model to make the most out of the individual characteristics of every forklift driver (i.e. incorporate predictive driving models to generate robust routing decisions).

Partners

• Combilift (IE)
• Mabo Engineering and Automation (BE)
• Irish Manufacturing Research (IE)