InHOSaS ('Integrated Hybrid Optimization of Autonomous Self-adaptive Systems‘) is a DFG-funded, research project between the Intelligent Systems Group at Kiel University and the Department of Food Informatics at the University of Hohenheim that aims to develop a hybrid Self-Adaptive and Self-Organizing (SASO) system.
SASO systems are often systems-of-systems that represent approaches to cope with the increasing dynamics of today’s software systems. They typically consist of productive units governed by an overarching self-adaptation layer. Control can thereby be fully centralized, initiated by a single central controller or fully decentralized, where each subsystem, i.e., agent, decides on its own actions. In the case of a fully centralized level of control, a central controller that potentially creates a bottleneck manages every autonomous entity and may be undermined by the egoistic decisions of the individuals prioritizing their own goals. Conversely, a fully decentralized level of control, where each entity independently determines its next actions, may lead to conflicting adaptations.
To address these challenges, this project aims to establish a hybrid architecture that combines system-wide, central planning with fully autonomous adaptation decisions of the local entities. This could be achieved by implementing a central control unit (CCU) that does not have full control over the agents. Instead, it just provides recommendations with corresponding incentives to them. Each agent can then decide how to implement a given plan or even disregard it entirely. In exchange, the agents provide the central planner with observations gathered from their environment.
Platooning, the coordinated driving of automated vehicles in convoys, serves as the project’s first use case and evaluation domain.
The contribution of the Department of Food Informatics (UHOH) focuses on the development of a central control unit that is equipped with a meta-adaptation logic and optimization techniques that focus on multiple-differing goals. At the same time, incentives and degrees of freedom will be used to incorporate local preferences.
