Software solutions for the digital position investigation of geometries

The RASAND laboratory aims at the research and development of software solutions for the digital position analysis of geometries. In this field, research and industrial practice are to be combined. On the one hand, this means that the issues dealt with are oriented towards industrial practice. All algorithms and data structures developed by us will be integrated into our own software library RASAND, which is to be used in industry in the long term. A further goal of RASAND is the education of doctoral students in this field. High-quality research work is to be carried out in the laboratory, which will lead the employees to a doctorate.

Designing vehicles virtually - save costs for building real prototypes

Software solutions for digital prototyping are researched and developed in the Rasand laboratory. When planning a new vehicle in the automotive industry, engineers first design all components on the computer. In the so-called Digital MockUp, they are then checked for their functional and spatial requirements, both individually and in combination. Geometrically, each component is given by a set of triangles that approximate its surface. The verification is done virtually by computer simulations, since the construction of real prototypes is very expensive.

Collisions between components - Detecting errors in the design

An important aspect of the Digital MockUp is to check that the components do not collide with each other, i.e. that they do not occupy the same installation space. The collisions between components that occur frequently in practice sometimes show the engineers relevant errors that require a design modification of the components. However, the majority are irrelevant for the engineers. A frequent example are collisions involving screws, bolts or clips. Their task is to attach components to each other, and the resulting conflicts are intentional.

 

Neural networks for 3D CAD data - qualified evaluation of collisions

In the GeoCADUp project, research is being conducted on AI-based static classification and evaluation of collisions between components. For this purpose, neural networks based on point clouds or images of the 3D CAD data are suitable. The neural network LocALNet developed in the project achieves one of the world's best results for the classification of 3D CAD data based on point clouds in an academic competition on the ModelNet40 dataset of the University of Princeton. Image-based nets have already been successfully trained to classify small parts such as screws, nuts and clips, which can have very different characteristics.

Digital assembly planning - avoid collisions between components

The avoidance of collisions between components also plays a major role in digital assembly planning. First of all, an assembly sequence of the components must be determined. In so-called path planning, a collision-free assembly path is then found for each component from its starting position outside the vehicle to its installed end position in the vehicle. In the past, Rasand has developed and implemented geometric algorithms for fast collision and tolerance calculation between components and, based on this, for path planning for 3D CAD data. The algorithms for path planning are specialized on narrow spaces, which are often found in a densely packed installation space.

Path planning in the near and far range - Consider fastening elements

The path planning for real components is currently being researched in the Assembly Planning project. Near their installed position, in the so-called close range, components are often unavoidably in collision with their surrounding geometry due to their fastening elements, e.g. cast-on movable clips. In addition, the fastening elements are often responsible for rotational movements during assembly. The software developed in the project considers cast-on fasteners. No other academic or commercial software is currently able to do so. For the path planning in the distant area following the planning in the close range, suitable assembly paths to the body address stations are found.

Assembly planning - part of the Discover Industry Truck

Since autumn 2019, our assembly planning has been part of the Discover Industry Truck, which drives to schools in Baden-Württemberg to get pupils interested in the MINT sector. In the dismantled truck, the working world of MINT professions is shown on 100m2 exhibition space. In the converted station on the subject of digital assistants, the pupils solve a nail puzzle in a race against the computer.

Research

In the RASAND laboratory, research is conducted in the three areas of registration, validation and planning.

Registration Research and development in the field of registration was started in 2009 in a MINT project with Autoform. The motivation for the project stems from quality assurance. For the quality analysis of industrially manufactured components, solutions have been and are being developed that position two triangulated geometries as congruently as possible. The registration software has been continuously expanded to include penetration-free registration, feature-based registration, registration of point clouds on triangular meshes and registration of spatial scans.

Validation The second research area of validation was started in 2010 as part of an FHProfUnt project together with the working group of Prof. Dr. Elmar Schömer at the University of Mainz and Teraport GmbH. The motivation for this project in the field of validation comes from the field of complete vehicle design. Efficient algorithms and software solutions, including parallelization mechanisms on the CPU and the GPU, were developed and implemented to validate compliance with safety distances between moving components.

Planning These fast collision and distance tests were used further to solve problems in the field of planning. In the development and manufacture of products, the question arises as to whether and how components and tools can be moved from a starting position to their installed position in a complicated and very compact installation space. Here, the focus was initially placed on the problem of bottlenecks that a component must pass through on its way from the start to the target position. Based on this, a feasibility study, commissioned by Daimler AG, was carried out, in which the feasibility of an automated buildability test for an entire automobile was examined. Currently, we include the deformability of components in our research. Components are usually fixed in their installed position, e.g. by clips. These fixings must be detected and, if necessary, deformed to simulate removal.

All developed procedures in the three areas of registration, validation and planning have been bundled in a separate software library called RASAND. This makes it easy for employees and students to try out new ideas and develop solutions based on them.

  • R. Hegewald, N. Wolpert, E. Schömer: "Simulation and Optimization of Industrial Disassembly Paths Using Position Based Dynamics”, accepted for 26th International Conference on Methods and Models in Automation and Robotics (MMAR), 2022
     
  • S. Dorn, N. Wolpert, E. Schömer:  "An Assembly Sequence Planning Framework for Complex Data using General Voronoi Diagram", accepted for IEEE International Conference on Robotics and Automation (ICRA), 2022
     
  • R. Hegewald, N. Wolpert, E. Schömer: "Iterative Mesh Modification Planning: A new Method for Automatic Disassembly Planning of Complex Industrial Components", accepted for IEEE International Conference on Robotics and Automation (ICRA), 2022
     
  • M. Adesso, R. Hegewald, N. Wolpert, E. Schömer, B. Maier, B. Epple: "Automatic Classification and Disassembly of Fasteners in Industrial 3D CAD-Scenarios", accepted for IEEE International Conference on Robotics and Automation (ICRA), 2022
     
  • M. Adesso, N. Wolpert, E. Schömer: ConfusionTree-Pattern: “A Hierarchical Design for an Efficient and Performant Multi-Class Pattern”, IEEE International Conference on Machine Learning and Applications (ICMLA), pp. 754-759 (2021)
     
  • R. Hegewald, N. Wolpert, E. Schömer: “Saliency Features for 3D CAD-Data in the Context of Sampling-Based Motion Planning”, IEEE International Conference on Robotics and Automation (ICRA), pp. 7858-7864 (2021)
     
  • S. Dorn, N. Wolpert, E. Schömer: “Expansive Voronoi Tree: A Motion Planner for Assembly Sequence Planning”, IEEE International Conference on Robotics and Automation (ICRA), pp. 7880-7886 (2021)
     
  • Q. Bytyqi, N. Wolpert, E. Schömer: “Local-Area-Learning Network: Meaningful Local Areas for Efcient Point Cloud Analysis”, CoRRabs/2006.07226 (2020) 
     
  • S. Dorn, N. Wolpert, E. Schömer: “Voxel-based General Voronoi Diagram for Complex Data with Application on Motion Planning”, IEEE International Conference on Robotics and Automation (ICRA), pp. 137-143 (2020) 
     
  • D. Schneider: “Rigid and Deformable Motion and Disassembly Planning with a Focus on the Digital Mockup Process in the Automotive Industry”, Promotion an der Johannes Gutenberg-Universität Mainz (2017) 
     
  • D. Schneider, E. Schömer, N. Wolpert: “Collision Detection for 3D Rigid Body Motion Planning with Narrow Passages”, IEEE International Conference on Robotics and Automation (ICRA), pp. 4365-4370 (2017) 
     
  • D. Schneider, E. Schömer, N. Wolpert: “A Motion Planning Algorithm for the Invalid Initial State Disassembly Problem”, International Conference on Methods and Models in Automation and Robotics (MMAR), pp. 35-40 (2015) 
     
  • D. Schneider, E. Schömer, N. Wolpert: “Completely randomized RRT-connect: A case study on 3D rigid body motion planning”, IEEE International Conference on Robotics and Automation (ICRA), pp. 2944-2950 (2015) 
     
  • A. Mantel: “Dynamic Distance Analysis: New Geometric Data Structures and Algorithms for the Real-Time Calculation of Tolerance Violating Regions in the Digital MockUp Process”, Promotion an der Johannes Gutenberg-Universität Mainz (2015) 
     
  • R. Erbes: “Efficient Parallel Proximity Queries and an Application to Highly Complex Motion Planning Problems with Many Narrow Passages”, Promotion an der Johannes Gutenberg-Universität Mainz (2013) 
     
  • R. Erbes, A. Mantel, E. Schömer, N. Wolpert, “Parallel Collision Queries on the GPU - A Comparative Study of Different CUDA Implementations”, Facing the Multicore Challenge, Springer, volume 7686, pp 84-95 (2012)
     
  • R. Erbes, A. Mantel, E. Schömer, N. Wolpert, “Triangle-Triangle Tolerance Tests”, European Workshop on Computational Geometry, pp 105 – 108 (2012)

Contact person

Nicola Wolpert
Nicola Wolpert nicola.wolpert@hft-stuttgart.de +49 711 8926 2697