2002, “Coordination specification for distributed optimal system design

Coordination plays a key role in solving decomposed optimal system design problems. Several coordination strategies have been proposed in the multidisciplinary optimization (MDO) literature. They are usually pre-sented as a sequence of statements: the parallel na-ture of the multidisciplinary subproblems is often ei-ther not addressed or only briefly mentioned. How-ever, a more formal description of the concurrency in the coordination is essential, in particular for large and non-hierarchic coordination architectures. This paper proposes to use concepts from communicating sequential processes (CSP) developed in concurrency theory. CSP allows the description of the MDO co-ordination as a number of parallel processes that op-erate independently and communicate with each other synchronously over pre-defined channels. For this pur-pose, we introduce elements of the language χ, a CSP-based language that contains data types such as reals, arrays, lists and tuples. The accompanying software tool set that enables the execution of a χ specifica-tion has been extended with a Python interface so that function calls to external software can be carried out. Through this interface, χ has been coupled with Mat-lab to run coordination specifications of distributed op-timal system design problems on single or on multiple parallel computers. An optimal design example is used to illustrate this. It can be concluded that the use of a CSP-based language such as χ for coordinating the so-lution of MDO problems is quite promising.