Motivation
The aim of the project is to develop intelligent software services that allow robot systems to independently generate and execute assembly and disassembly plans based on CAD and environmental information.
Assembly, which is the joining of subsystems to form a functional overall system, is one of the most cost-intensive processes in production technology. However, the German economy is heavily dependent on production technology and could benefit from automated assembly in various trades. Disassembly is the reverse operation, i.e. the disassembly of the individual starting elements. Due to increasing resource problems and the ever more important energy balance in the overall life cycle of products, an automated solution is also required here. One possibility would be the use of two or more robot systems that perform this task with suitable tools and cooperative interaction. However, there are currently no suitable algorithms that allow a cooperative robot program to be generated fully automatically, meaning that a robot programmer is required. However, this is not acceptable for economic use.
Overall goal
This research project therefore aims to develop methods for planning disassembly and assembly sequences based on CAD and camera data. The generated planning sequences can then be translated into a robot program that is executed automatically. The individual algorithms are to be integrated into a software architecture so that the complete consistency and autonomy of the process can be demonstrated.
Methodology
The planning algorithms to be developed are based on a combination of symbolic AI methods and data-driven approaches. The use of symbolic methods allows for transparency in finding solutions as well as the embedding of kinematic and dynamic substitute models. However, these methods fail for certain assembly combinations due to the complexity of the problem, which is why data-driven approaches are used to reduce the solution space. The translation of the discrete planning steps into a program that can be executed by the robot is carried out using a sub-symbolic task description.
Project funding
The ADeMoRos project is funded as part of the Carl Zeiss Foundation 's Forschungsstart program.