DARPA iFAB Program

The manufacture of complex cyber-physical systems is generally limited to two options: large scale manufacturing that is efficient at producing one known product quickly and repeatably, or prototype assembly of a single (or few) test article(s).  Factories that do one generally cannot efficiently do the other – it is currently not possible to achieve efficient large-scale manufacturing that is also flexible.  iFAB (instant Foundry, Adaptive through Bits) is an attempt to accomplish both in one manufacturing entity.  It aims to take a META-articulated system design representation and automatically configure a digitally programmable manufacturing facility tailored to fabricate it.  The IT infrastructure required to accomplish this is also able to provide detailed manufacturability information in the form of cost, schedule and identification of critical components to the designer throughout the design process.  Thus, the final verified META-articulated design sent for iFAB manufacturing already has manufacturing considerations factored in.  The physical instantiation of the iFAB manufacturing capability, called a foundry, includes a network of participating manufacturing facilities and equipment, the sequencing of the product flow and production steps, as well as work instructions and training modules.  In essence, iFAB seeks to eliminate the learning curve inherent in large-scale manufacturing even for limited build numbers.

The iFAB tool suite is centered around two functions: 1) providing manufacturability feedback to the designer at a level of detail commensurate to the design; and 2) configuring a foundry of networked manufacturing capabilities tailored to the final verified design, including supply chain considerations, assembly planning, and automatically generated computer-numerically-controlled (CNC) and human work instructions. 

Much like META, iFAB is predicated on detailed formal models representing the capabilities of various manufacturing machines and processes.  By mapping these models into the same semantic domain as the vehicle design, iFAB can automatically constrain the design trade space such that designs that are not manufacturable in a given iFAB Foundry network of manufacturing capabilities are automatically culled.  Though we term iFAB a “foundry” – principally to differentiate it from a conventional factory that, at least in the defense world, tends to be a custom facility tailored to a specific product or small set of product variants –, in actuality it is primarily an information architecture.  Only the final assembly capability needs to be co-located under a single roof in anything resembling a conventional fabrication facility.  The rest of the iFAB Foundry can be geographically distributed and extend across corporate and industrial boundaries, united only by a common model architecture and certain rules of behavior and business practices.

The year one iFAB effort was organized into the following technical areas:

  • (FCT) Foundry Configuration Tools: a set of tools that enable the initial design and rapid reconfiguration of foundry capability based on an objective system design and possible excursions from that system design
  • (MML) Manufacturing Capability and Process Model Library: a library of various fabrication processes and associated “factory components,” i.e., machines and techniques employed to produce the various constituent elements of the objective product
  • (MCF) Manufacturability Constraint Feedback to System Design: a mechanism for the feedback of manufacturability constraints into the design and design tradespace exploration process

Successive development on iFAB has focused on developing an integrated tool suite, instantiating the manufacturability constraint feedback as a virtual test bench to be executed within the CyPhy META design process, and updating the manufactuing model library with relevant data in preparation for the first DARPA Fast Adaptable Next-Generation Ground Vehicle (FANG) Challenge.

  • (iFAB-FANG1) iFAB tool suite integrating FCT, MML, MCF and added feature sets as implemented in DARPA FANG 1 Challenge.