Fraunhofer Institute for Material Flow and Logistics

Topology-optimized parts are increasingly used in technical applications, as their production has been made feasible through modern additive manufacturing technologies, such as laser powder bed fusion. The post-processing of such parts to meet tolerance or surface requirements, for example by machining processes, is a challenge that requires an individual clamping solution. To address this issue, a methodology to automatically generate pareto-optimal clamping solutions is proposed, especially with regard to avoiding stress-induced deformation. Solutions are obtained by embedding a finite element software in a black-box optimization loop. Each solution consists of customized attachment jaws, which permit usage of a conventional vise.

Existing 6D pose estimation datasets primarily focus on small household objects typically handled by robot arm manipulators, limiting their relevance to mobile robotics. Mobile platforms often operate without manipulators, interact with larger objects, and face challenges such as long-range perception, heavy self-occlusion, and diverse camera perspectives. While recent models generalize well to unseen objects, evaluations remain confined to household-like settings that overlook these factors. We introduce MR6D, a dataset designed for 6D pose estimation for mobile robots in industrial environments. It includes 92 real-world scenes featuring 16 unique objects across static and dynamic interactions. MR6D captures the challenges specific to mobile platforms, including distant viewpoints, varied object configurations, larger object sizes, and complex occlusion/self-occlusion patterns. Initial experiments reveal that current 6D pipelines underperform in these settings, with 2D segmentation being another hurdle. MR6D establishes a foundation for developing and evaluating pose estimation methods tailored to the demands of mobile robotics. The dataset is available at https://huggingface.co/datasets/anas-gouda/mr6d.

This contribution presents the SuLo (Sustainable Logistics) framework developed through a systematic literature review of the drivers and barriers to implementing sustainable sustainability in logistics, with a focus on warehousing. As global trade expands, logistics operations face increasing pressure to balance economic efficiency with environmental and social responsibility. Warehousing plays a crucial role in supply chains, contributing significantly to energy consumption and greenhouse gas emissions. Despite growing interest in sustainable logistics, research on sustainable warehousing remains fragmented. This paper synthesizes the existing literature to identify key drivers, barriers, and best practices for sustainable warehousing. The findings reveal that regulatory policies, cost savings, and stakeholder expectations are primary drivers, while high implementation costs, lack of awareness, and operational complexity hinder progress. A structured framework is proposed to guide sustainability integration, incorporating renewable energy, waste reduction, automation, and employee well-being initiatives. The study highlights the necessity of a holistic approach that aligns environmental, economic, and social dimensions. While challenges remain, sustainable warehousing offers long-term benefits, including reduced environmental impact, improved operational efficiency, and enhanced corporate reputation. In conclusion, future research should explore empirical case studies and technological innovations, particularly artificial intelligence, to further advance sustainability in warehousing.

The performance of Large Language Models, such as ChatGPT, generally increases with every new model release. In this study, we investigated to what degree different GPT models were able to solve the exams of three different undergraduate courses on warehousing. We contribute to the discussion of ChatGPT’s existing logistics knowledge, particularly in the field of warehousing. Both the free version (GPT-4o mini) and the premium version (GPT-4o) completed three different warehousing exams using three different prompting techniques (with and without role assignments as logistics experts or students). The o1-preview model was also used (without a role assignment) for six runs. The tests were repeated three times. A total of 60 tests were conducted and compared with the in-class results of logistics students. The results show that the GPT models passed a total of 46 tests. The best run solved 93% of the exam correctly. Compared with the students from the respective semester, ChatGPT outperformed the students in one exam. In the other two exams, the students performed better on average than ChatGPT.

Semantic mapping has many use cases in mobile robotics. For example, in logistic environments, it is applied to find and transport loading goods or in mobile manipulation to identify objects to grasp, and it often relies on neural networks to accomplish image detection and segmentation for semantic mapping. Data sets are usually required to train those neural networks, which are difficult to generate for many organizations since their creation requires qualified personnel. This paper demonstrates how federated learning, combined with supervised neural networks, effectively reduces the effort for semantic map training while maintaining data sovereignty. It allows organizations to facilitate their individual training data generation efforts to build semantic maps by collaborating with other organizations without sharing their training data and, therefore, not losing their data sovereignty. At the same time, they can reduce their individual effort to generate and train data.

In logistic environments, diverse applications, reaching from robotics to monitoring and quality assess-ment, benefit from depth information processing. In addition, multiple depth image cameras have been introduced to the market which differ widely in the used imaging technology, quality and price. In this paper, an open source data set is presented, containing calibrated and timely synchronized real world data from seven different depth imaging sensors, including a quantitative evaluation of the quality of each used sensor, in multiple logistic scenarios.