Speaker
Description
The advancement of artificial intelligence (AI) in intralogistics critically depends on the availability of realistic and diverse datasets. However, existing datasets in this domain often focus on narrow tasks such as object detection or activity recognition, lacking comprehensive three-dimensional (3D) representations of entire intralogistics systems. This paper addresses this gap by proposing a methodology for generating and formally describing synthetic 3D scenes of intralogistics environments, particularly warehouses, using a procedural content generation (PCG) approach. The proposed method constructs warehouse layouts by randomly placing and populating functional areas such as goods in, goods out, storage, packaging, and charging zones with realistic assets selected from domain specific libraries. Each generated scene is accompanied by a structured matrix describing asset positions, orientations, and types, facilitating downstream AI applications without needing to parse complex 3D model files. The approach was implemented using Nvidia’s Isaac Sim, producing a dataset of 100 diverse warehouse scenes comprising over 19,000 assets. The dataset's variability is confirmed through statistical metrics of capacity, workstations, and charging infrastructure. This foundational dataset aims to support a wide range of AI applications, from robot navigation and vision-based scene understanding to digital twin modeling and system simulation. Future work will include finding metrics to measure realism, finding learning-based approaches for the generation of the dataset, as well as incorporating further logistics processes and assets.
