Periodic Reporting for period 1 - doks Innovation (Automated AI-based inventory management for logistics)
Période du rapport: 2019-08-01 au 2019-11-30
For any company along a supply chain, the ability to locate, quantify, and keep track of all goods that are typically stored on pallets or in crates is absolutely essential. However, this task is mainly a manual and labour-intensive task that is necessary but does not create any value and is moreover exceedingly prone to error. Hence, industry experts within the field of logistics predict a rapid increase in the need for automated and digitized solutions for inventory and master data management systems for logistics processes. The rapidly growing size and number of warehouses and stock areas will result in increasing demand for solutions that help logistics and industrial companies to automate and digitize warehouse and stocktaking processes.
For this reason, this project is aimed at developing DigiLOGIS, an automated inventory management system that uses artificial intelligence (AI) to track the location and condition of pallets and crates in real time. Proprietary algorithms for gathering and analysing different types of sensor data enable users to create a digital twin of the warehouse or storage yard with a wealth of actionable information that help to make better decisions. With the aid of machine-learning algorithms, data is processed to relevant information and thus offers advantages in the planning and design of a wide range of logistics processes. The DigiLOGIS system can count, find and verify the condition of goods stored. It replaces manual tracking as well as handheld scanner and manual measurements for master data gathering.
The project was carried out with partners such as freight forwarding and logistic service providers as well as automotive companies. The focus was on tracking the flow of goods from and into warehouses as well as developing algorithms for object detection. The latter can not only be used for the automatic generation of master data but also for automatically generating exact counting results of crates and pallets. This approach for inventory is extremely data-driven, which means that the underlying algorithm for machine learning optimizes itself continuously with increasing data volume.
The project was carried out with partners such as freight forwarding and logistic service providers as well as automotive companies. The focus was on tracking the flow of goods from and into warehouses as well as developing algorithms for object detection. The latter can not only be used for the automatic generation of master data but also for automatically generating exact counting results of crates and pallets. This approach for inventory is extremely data-driven, which means that the underlying algorithm for machine learning optimizes itself continuously with increasing data volume.
Although, the preparation of data for the training of so-called neural networks is extremly time-consuming and costly, doks. found ways to work with much smaller sets of data and managed to implement software features and supporting mechanisms that enable accuracy rates of up to 100% despite an actually insufficient data basis.