InGen ProjectInGen is a subproject of the CoTeSys Cluster.
The objective of this project is to develop a hierarchical approach to knowledge representation and action planning which enable cognitive systems to compute goal attaining strategies efficiently even in presence of large variations of the environment. In detail, the project aims at:
- using the capabilities of context adaptation and especially the CAWAR-Framework to adapt the structure of cognitive systems to strongly and rapidly changing contexts
- introducing a layered representation of knowledge, models, and planning strategies to combine heterogenous decisions (i.e. adaptations, planning strategies, control trajectories) in an integrated fashion
- employing formal models to compute control strategies using predictive and adaptive algorithms which receive reference strategies from higher layers
- combine task decomposition with solution decomposition, meaning that the resulting subtasks after decomposition are assigned each a appropriate strategy depending on system context, knowledge and assessment of success of the strategies.
The hierachical decomposition of complex task into simpler subtasks, which resembles the approach to problem solving used by human brains, will push the complexity of problems that cognitive systems can solve autonomously.
We develop a layered architecture to represent different aspects of the system (adaptation, task planing and control) and make use of formal models on each leyer.