Engineers to be supplanted by domain experts?

SAN JOSE, CA. What’s the future of embedded systems design—or better yet, of the hardware and software engineers whose domain embedded system design has been? That topic was touched upon by panelists at an Embedded Systems Conference (ESC) presentation titled “Addressing Embedded Challenges with ARM Technology,” which coincided with National Instruments’ announcement of an extension to LabView that supports ARM’s ARM 7, ARM 9, and Cortex-M3 microcontroller families.

Panel participants included Reinhard Keil of ARM, Geoff Lees of NXP Semiconductors, Jean Anne Booth of Luminary Micro, Stuart McLaren of STMicroelectronics, and Tim Dehne of NI. ESC editorial director Moderator Richard Nass moderated.

ARM’s Keil kicked off the event by describing the evolution of embedded systems design. He said that hardware costs are falling and that the challenge today lies in software development. Citing Willert Software Tools as his source, he described the evolution of software design from the time when a project might involve 64 kbytes of code. Such a project, he said, was typically handled by one programmer—a strange guy who claimed to work long hours but delivered buggy code late. As programs became larger, he said, the US programmer was ultimately displaced by teams from India, who, because of communications difficulties and management issues, delivered buggy code late.

Lees of NXP Semiconductors said a way to deal with the limited pool of talented hardware designers was to move toward higher-level constructs, which would help alleviate the “1000-page-manual problem.” Hardware engineers are hard to find, he said, because college graduates today typically have C and C++ experience and may be familiar with DSP concepts—but they have little in the way of a hardware background.

Booth of Luminary Micro suggested that embedded design—a discipline different from either hardware or software design—could alleviate bottlenecks in getting embedded systems to market. Embedded design tools such as the LabView extension for ARM, she said, could pave the way for domain experts to design embedded systems without significant input from traditional hardware and software designers.

Dehne of NI concurred with ARM’s Keil that software complexity represents today’s biggest challenge in embedded systems design, and he said one goal of NI with its new LabView extension is to extend the use of ARM technology to domain experts, whether they are in Germany, India, or the US.

ESC’s Nass then posed the question, what is a domain expert? Dehne provided the example of a medical device designer who is developing a product for laser eye surgery—that designer wants to focus on control algorithms for positioning his device’s mirrors and on the doctor-user interface, but he or she may have limited knowledge of typical embedded-systems hardware and software design.

That prompted a question from the audience regarding how companies such as ARM would need to adjust their marketing messages to appeal to domain experts like doctors who understand the condition they are trying to treat but who are unfamiliar with multicore architectures and I/O configurations. The consensus seemed to be that companies might cite specific applications examples and describe what IP combination best suited that application.

Lees said vendors do need to address the issue, but he added that the idea of a doctor being the domain expert who would engage in embedded systems design is an extreme example, and that the domain expert targeted by embedded-system-tool vendors would more likely be someone like a hydraulic engineer.

Paradoxically, NI at the event awarded a LabView-powered Lego Mindstorms set, which is aimed in part at getting today’s 9- and 10-year-olds to become the hardware and software engineers of the future. They might just graduate from college to find their skill sets have been supplanted by design tools that target the domain experts.

See the related post "What do we want from robots?" on what artists might hope for from embedded systems. Perhaps equipped with the proper design tools they can implement what they seem to be seeking.

See also the related article "FIRST partners with NI to provide students with CompactRIO," which describes an initiative that will enable high school students to create advanced robots and program them in either LabView or ANSI C.