Systems Engineering and the META Program

I pe­ri­od­i­cally look for in­for­ma­tion on sys­tems en­g­ineer­ing. This time I came across a pow­er­point pre­sen­ta­tion from the MIT Open Course­ware course Fun­da­men­tals of Sys­tems Eng­ineer­ing. Pro­fes­sor de Weck, who taught the course, had done some re­search on state-of-the-art meth­ods de­vel­oped as part of DARPA’s META Pro­gram.

A few years ago DARPA wrapped up the pro­gram, de­signed to speed up de­liv­ery of cy­ber-elec­tro-me­chan­i­cal sys­tems (war ma­chines) by 5x. Since the par­ent pro­gram Adap­tive Ve­hi­cle Make seems to have con­cluded with­out pro­duc­ing a ve­hi­cle, I in­fer the META Pro­gram lost its fund­ing at the same time.

The work it pro­duced ap­pears to be ad­ja­cent to our in­ter­ests along sev­eral di­men­sions though, so I thought I would bring it to the com­mu­nity’s at­ten­tion. The pitch for the pro­gram, taken from the ab­stract of de Weck’s pa­per:

The method claims to achieve this speedup by a com­bi­na­tion of three main mechanisms:
1. The de­liber­ate use of lay­ers of ab­strac­tion. High-level func­tional re­quire­ments are used to ex­plore ar­chi­tec­tures im­me­di­ately rather than wait­ing for down­stream level 2,3,4 … re­quire­ments to be defined.
2. The de­vel­op­ment and use of an ex­ten­sive and trusted com­po­nent (C2M2L) model library. Rather than de­sign­ing all com­po­nents from scratch, the META pro­cess al­lows im­port­ing com­po­nent mod­els di­rectly from a library in or­der to quickly com­pose func­tional de­signs.
3. The abil­ity to find emer­gent be­hav­iors and prob­lems ahead of time dur­ing vir­tual Ver­ifi­ca­tion and Val­i­da­tion (V&V) and gen­er­at­ing de­signs that are cor­rect-by-con­struc­tion al­lows a more stream­lined de­sign pro­cess and avoids costly de­sign iter­a­tions that of­ten lead to ex­pen­sive de­sign changes.

Which is to say they very care­fully ar­chi­tect the sys­tem, use known-to-be-good com­po­nents, and em­ploy for­mal ver­ifi­ca­tion to catch prob­lems early. In the pa­per a simu­la­tion of the META work­flow suc­cess­fully achieved a 4.4x de­vel­op­ment speedup com­pared to the same pro­ject’s ac­tual de­vel­op­ment us­ing tra­di­tional meth­ods.

There are a bunch of in­di­vi­d­ual di­rec­tions ex­plored which are of in­ter­est. Some that struck me were:

It looks like the repos­i­tory for the pro­gram’s out­puts is here. A mis­sile en­g­ineer try­ing to ap­ply the method to health­care is here. Re­duc­ing health­care costs is rocket sci­ence!

In a nut­shell it con­tains a whole bunch of things we have long dis­cussed all in a sin­gle pack­age, and now there are a few peo­ple out and about try­ing to get parts into prac­tice.