Computers can be used to quickly process and analyze large volumes of data in complex settings. Self-learning machines can recognize and remember patterns, and then react as we would when new situations with similar patterns arise, provided that we have defined the ideal result. This is facilitated by supervised learning. Machines can also learn without pre-defined results defining the outcome. In this case, machines can end up with a preferable, and sometimes surprising, result too.
With the help of digital assistants, the system can work through suggestions that are adopted, adjusted, or rejected in the control center. Those who trust in the capability of the system, can leave the control up to the automated system while letting humans monitor the situation, with the option of intervening at any point. It is important to note here that intervening too often will interfere with the self-regulating system and potentially throw it off track. We often find that automated systems that are left in control with only very few interventions deliver the best results.
Tailored algorithms or general machine learning processes via neural networks can be used to optimize a warehouse.
Automated Warehouse Control
Within an automated warehouse, a replica is created using a digital twin. The system can use this replica to learn without the impact of negative cycles. The digital twin is used to compute changing system parameters. The system can be trained with the digital twin and fed order variations from time to time, not to mention that the system also learns during live operation. This allows for the warehouse management and material flow systems to be adapted in line with changing requirements in the warehouse.
Dynamic Storage Allocation
Warehouse allocation is quite a skill. We have already compared manual and automated warehouses and looked at a different way of clustering goods categories. In a manual warehouse, the important thing is to keep clusters together, while in an automated warehouse it is all about even distribution, in order to avoid too much strain being placed on individual elements within a certain time frame thus provoking bottleneck situations.
One way of ensuring that there is plenty of scope for optimization is to use different machines and types of storage (e.g., pallets and totes) or transport devices – continuous conveyors (e.g., conveyor belts) or discontinuous conveyors (e.g., automated guided vehicles) – within one location.
Automated Order Management
Machine learning or tested algorithms can be used to work out the most suitable groups of orders from the wide range of potential order variations as well as their optimum sequence and starting point. The entire system can draw on its full power here, putting it well above the human mind when it comes to computing countless pieces of well-defined data. But humans on the other hand, are good at interpreting large volumes of fuzzy data and at drawing solid conclusions from it.
If a decentralized approach is applied, the individual objects in a warehouse can fulfill the tasks assigned to them automatically. The overall system is responsible for allocating tasks, while the individual physical objects automatically fulfill the tasks assigned to them. Similar objects then behave in a flock-like manner.
Interaction is comparable to the Internet of Things, with individual objects being engineered in such a way that they feature on-board intelligence. Alternatively, objects within the system are identified and they then call up their “agents,” which are responsible for dealing with tasks within the system. Agents are virtual replicas of objects within the system that have a certain degree of autonomy and on which machine intelligence can be implemented.
Peter Totz works as Director Business Consultancy at SSI SCHAEFER.
His career began as a project engineer, data analyst and simulation specialist in Graz. With intermediate steps in production planning and as a logistics consultant, he worked as a senior consultant and project manager for many years. Later, he was responsible for business development in Latin America before taking over the lead of the globally active Business Consultancy Group.