Speedup Estimation on Heterogenous Hardware
Modern computer architectures provide developers with large computational power, they consist of multi-core CPUs and accelerators such as General Purpose GPUs (GPGPUs). The main differences of the computing hardware components are scalability and performance. While multi-core CPUs can only handle a small amount of threads, GPGPUs are designed to handle thousands of threads concurrently. There are three main reasons for engineers to not use the offered computing capabilities efficiently:
Systems designed according to the Organic Computing paradigm, can quickly adapt their behaviour and even their architecture to changing situations. This reorganisation can be triggered by environmental conditions or other constraints that have influence on these systems. The adaption of such a system can improve the system, as it can become more applicable for the environment it is used in or more robust to certain failures. However, as the adaption can change the system fundamentally, a new approach is needed to cover for the possible loss of trustworthiness of the system. In the project OC-Trust, we at SRA focus on system-to-system Trust. We are especially interested in Trust relations between entites of the system, expressed as credibility of these entities.
Selbst-organisierte robuste kooperative Exploration (SoREx)
The project studies methods for the description, planning and execution of complex procedures which require a larger number of agents in different roles. We will try to shift the process of the planning from the design time to the run time under constraints. In addition, we distribute the planning from the centralized server towards the agents. The planning at run time allows the quick response to special situations like obstacles or conflicts. These special situations would result in a local re-planning as long as the agent is capable of resolving these problems locally. Otherwise the agent should escalate to the next higher layer. To fulfill the replanning under limited resources, we will investigate to reduce the complexity of the replanning. The evaluation will be performed in the scope of multi-agent-simulation together with a workbench of twelve Khepera mobile robots.