Show description

In the current implementation we rely on an evolutionary optimization algorithm that has been successfully used for similar problems [TCGK02]. For the selection, we use the well known evolutionary multi-objective optimizer SPEA2 [ETZ00, ZLT01]. 5 Exploration Results In this section, we focus on the detailed discussion of a proof-of-concept case study. Although the considered system is not very complex, it represents an important class of wearable systems. We emphasize on the most important features of the simulation environment and provide first insights into the trade-offs involved in the design of wearable computing systems.

The maximal amount of data to be transmitted. Based on this information, we calculate the power consumption. It can be seen that all deadlines are satisfied if the period of the channel is τ∗ = min1≤k≤n (δk /2). We get the worst case delay for transmitted data packets Twc = min (δk ) + Td . 8) Considering the case that all tasks start sending their data at the same time, we get the maximal burst size n ∗ l = lk k=1 τ∗ τk . 9) As we switch between two bursts from transmitting/receiving to standby and back to transmitting/receiving, we require that l∗ /Bmax + Ti ≤ τ∗ for 30 Chapter 2.

Column 6 lists the derived computing power requirements 22 Chapter 2. Performance Estimation for Design Space Exploration and one output task. This representation allows to model the computation/communication trade-offs involved in distributing the execution of an application onto different devices. For this purpose, the input and output tasks are treated in a special way: their computational loads are defined to be zero. In this manner, input and output tasks may be assigned to the wearable’s IO devices while the computational tasks are assigned to the computing devices.