Design Space Exploration Sample Clauses
The Design Space Exploration clause defines the process and parameters for systematically evaluating multiple design alternatives within a project. It typically outlines the criteria, methods, and tools to be used for comparing different design options, such as performance, cost, and feasibility, and may specify the roles responsible for conducting the analysis. By formalizing this process, the clause ensures that all viable solutions are considered before finalizing a design, thereby optimizing outcomes and reducing the risk of overlooking better alternatives.
Design Space Exploration. To define a design space to explore, initialisation attributes in CoHLA are very useful. As Situations and ClassInitialisa- tions affect the initialisation values for federate instances, these can also be used to specify a design space. CoHLA has been extended by allowing the user to specify lists of different initialisation values for federate instances or attributes. As there are three methods to assign initialisation values, we also consider three different types of lists that can be used to define a design space. • Lists of initialisation values are lists that specify specific values for initialisation attributes of a specific federate instance. For every initialisation attribute for every ▇▇▇▇▇- ation instance, such list may be given in a DSE definition. 1 Set bottomSlider.startPosition : "0.15", " 0.05", "-0.05", "-0.14" Listing 5. Example of a list of initialisation values. • Lists of ClassInitialisations are lists that specify ClassIn- itialisations that should be applied to a specific federate instance. A DSE definition may contain one list for every federate instance in the federation. 5▇▇▇▇▇://▇▇▇▇▇.▇▇▇/ 6▇▇▇▇▇://▇▇▇.▇▇▇▇▇▇.▇▇▇/ element in the list with all other design space elements defined. Table I shows a design space for a hypothetical system. The system consists of two federates, each having two attributes. All but one attributes have three possible design parameters. When using independent exploration behaviour, there are 27 possible system configurations: (x, a, q, u), (x, a, q, v) Independent behaviour can only be used when leaving out Attribute C from the design space, which results in three possible system configurations: (x, a, u), (y, b, v) and (z, c, w).
1 DSE strokeStartPositions { 2 SweepMode Independent 3 Set bottomSlider.Position_realInitial : "0.15", "0.05", "-0.05", "-0.14" 4 Set supervisoryController.initSpeed : "1.0", " 1.5", "2.0", "2.5", "3.0" 5 } Listing 8. Design Space definition for the SliderSetup. Listing 8 displays a design space definition for the Slider- Setup. The design space specifies two lists, each containing initialisation attribute values. The initialisation attributes are used to specify a starting position for the bottom slider and an initialisation speed for the supervisory controller. As the SweepMode is set to Independent, the design space has a size of 20 (4 5) possible configurations. From this design space definition, a DSE configuration file is generated by the code generator. These configuration f...
Design Space Exploration. In the first year of the DSE task the focus has been on assessing what the industrial partners currently understand of DSE, what their aspirations are for DSE within their case studies and what DSE approaches and algorithms should the INTO-CPS project aim to support. As outlined earlier in Section 2.1.1 the DSE approaches currently undertaken by the WP1 partners are manual and rely heavily on engineer expertise to both define product parameters and to analyse results. While there will always be a need for engineer expertise within the DSE methods to both define what aspects of a product design should be varied and how to measure simulation outputs to evaluate and rank designs, the tool support within INTO-CPS aims to support the engineer in several ways: reduce the workload in defining the parameters for and running multiple simula- tions to explore the design space by automating the configuration and launch of simulations; provide a range of DSE algorithms for the engineer to select from along with guid- ance about the suitability for different problem domains; provide templates and scripts to obtain objective measures of simulated CPS per- formance from the raw simulation results; support the use of both scripted ranking functions and pareto optimality to rank the results of each design globally; reduce the total number of simulations required to have confidence in finding a globally optimum solution by using design ranking and closed loop optimisation methods. The details of both the currently implemented DSE method and the set of proposed methods may be found in D5.1a [GHJ+15]. This deliverable also contains the aspirations of the WP1 partners for DSE along with comment on how the proposed methods will meet them.
Design Space Exploration. A comprehensive overview of the state of the art in DSE is provided in Deliverable 5.1a [GHJ+15]. 23▇▇▇▇://▇▇▇.▇▇▇▇▇▇-▇▇▇.▇▇ 24▇▇▇▇://▇▇▇.▇▇▇.▇▇▇.se/labs/pelab/OpenProd/ 25▇▇▇▇://▇▇▇.▇▇▇▇▇.▇▇/ 26▇▇▇▇▇://▇▇▇.▇▇▇▇▇▇▇▇.▇▇▇/external-projects/modrio 27▇▇▇▇://▇▇▇.▇▇▇▇▇▇.▇▇ 28▇▇▇▇://▇▇▇.▇▇▇▇▇▇.▇▇
Design Space Exploration. The Design Space Exploration (DSE) extension of the INTO-CPS tool chain allows a user to automatically explore a range of parameters and candidate implementation options automatically while avoiding the problem of state space explosion. The extension is launched from the main INTO-CPS appli- cation but is actually a separate entity. The interested reader can find more details about the DSE extension in [GHJ+15]. There is three main parts to the extension: by which of the DSE search algorithms the user has elected to use given the nature of the model and problem domain; DSE Analysis has the role of processing the results of simulation to obtain the key objective values by which it should be evaluated. Such anal- ysis could be as simple as finding the maximum power consumed by a device during the simulation, or it could be more complex such as computing the deviation from a path or some temporal constraint over the states visited during the simulation (the latter functionality will be implemented in part using the RT-Tester tool, see Section 6.7); and
Design Space Exploration. In order to be able to make right system platform selections, the feasibility of candidate application-platform bindings need to be predicted w.r.
Design Space Exploration. During the process of developing a CPS, either starting from a completely blank canvas or constructing a new system from models of existing compo- nents, the architects will encounter many deign decisions that shape the final product. The activity of investigating and gathering data about the merits of the different choices available is termed Design Space Exploration. Some of the choices the designer will face could be described as being the selection of parameters for specific components of the design, such as the exact position of a sensor, the diameter of wheels or the parameters affecting a control algo- rithm. Such parameters are variable to some degree and the selection of their value will affect the values of objectives by which a design will be measured. In these cases it is desirable to explore the different values each parameter may take and also different combinations of these parameter values if there are more than one parameter, to find a set of designs that best meets its objectives. However, since the size of the design space is the product of the number of parameters and the number of values each may adopt, it is often impractical to consider performing simulations of all parameter combinations or to manually assess each design. The purpose of an automated DSE tool is to help manage the exploration of the design space, and it separates this problem into three distinct parts: the search algorithm, obtaining objective values and ranking the designs according to those objectives. The simplest of all search algorithms is the exhaustive search, and this algorithm will methodically move through each design, performing a simulation using each and every one. This is termed an open loop method, as the simulation results are not considered by the algorithm at all. Other algorithms, such as a genetic search, where an initial set of randomly generated individuals are bred to produce increasingly good results, are closed loop methods. This means that the choice of next design to be simulated is driven by the results of previous simulations. Once a simulation has been performed, there are two steps required to close the loop. The first is to analyse the raw results output by the simulation to determine the value for each of the objectives by which the simulations are to be judged. Such objective values could simply be the maximum power consumed by a component or the total distance travelled by an object, but they could also be more complex measures, s...
