Autonomous Systems

Intralogistic material flow processes, such as transports, are becoming increasingly complex. Among other things, this is due to the wide variety of requirements that need to be considered, such as flexibility in terms of planning transports or increasing resiliency. With the advance of digitization and the interconnection of people, machines, sensors, and actuators, the number of data points worldwide will double by the end of the decade. This will push current systems to their computational limits when all available information will be used. We address this by distributing the intelligence from centralized instances to the individual participants, such as automated guided vehicles (AGVs) or autonomous mobile robots (AMR). This leads to an increase in the autonomy of the individual participants so that they can perform their tasks in a fully autonomous manner.

Large-scale AMR systems (e.g., warehouses with hundreds or thousands of interacting AMRs) will inevitably rely on new planning approaches. Therefore, in addition to auction-based task allocation, we are also investigating new methods and approaches. We deal with centralized and decentralized optimization models that are suitable for systems with a large number of participants and multi-objective problems.

Autonomous agents attempt to make independent decisions to achieve a set of goals in dynamic, unpredictable (usually complex) environments. The basis for decision-making can be the agent's own state (self-awareness) or the perception of the environment (context-awareness). Agents find application in cyberphysical production systems, the merging of the real world with the virtual world. While an AGV or an AMR are two examples of the real world, the standardized digital material flow and the transports it contains are examples of a virtual agent. Here, the autonomous material flow can negotiate the execution of the transports to the agents of the AGVs and AMRs, while the vehicles execute the transports.

It is important to look at the challenges holistically, from the application and its requirements to the individual participants and the interaction between them. The basis for interaction between agents, as with us humans, is communication. This must scale with the number of participants because without it, no active interaction is possible. In addition, dynamic networking is required at runtime to create a loose coupling of participants.

Simulations are an important part of the development process to understand the behavior of the agents. For this purpose, the application is first developed and tested in simulation before being deployed in the real world. We view the simulation as a digital representation of reality. However, simulations are only as good as their models. That is why we try to close the gap between the simulation and the real world. This means that simulations must approximate the real world as closely as possible. But a union or a symbiosis of two worlds is also conceivable. In the simulation, situations can be created that are not conceivable in the real world. These new situations serve as a basis for adapting the behavior of the agents.