Technology from advanced-physics research institute CERN will form part of National Instruments’ long-term strategy to improve the ability of distributed systems to support real-time control.
The company is working with the White Rabbit protocol developed at CERN for its instrumentation and which extends the IEEE 1588 protocol originally developed for telecom systems into the nanosecond range. NI has already incorporated the notion of timed control into its LabView graphical programming tool using a combination of standard microprocessors and FPGAs as execution targets.
Research by computer scientists such as Professor Edward Lee of UC Berkeley has indicated that the lack of a notion of time in most programming languages adds unnecessary complexity to real-time control systems and that clear synchronization is needed to enable applications such as automated driving where vehicles on the road co-operate with each other. A group of US universities recently received $4m from the National Science Foundation for their RoseLine timing-aware research for cyber-physical systems.
Mike Santori, vice president of product marketing at NI, said during the company’s annual user conference in Austin, Texas: “Timing and synchronization are foundation technologies for control and test.”
Jim Truchard, president, CEO and cofounder of NI claimed in his keynote speech at NI Week: “We have leading edge technologies in this area because we have FPGAs to do high-speed computation and synchronize it with the rest of the simulation.”
Truchard pointed to the use of the FPGA-based hardware programmed using LabView that could support a 1.2µs response time in a closed-loop controller for hardware-in-the-loop simulation developed by Subaru for a powertrain electronic control unit (ECU).
“We are continuing on this quest. One is in distributed time. We know we can do better. White Rabbit offers a tremendous opportunity to get into the nanosecond range and we are working intensely to bring this capability to distributed systems. As we look forward to sensor fusion where we are combining multiple distributed sensors to compute a higher-level measurement we need this kind of synchronized time support,” said Truchard.
The work at NI will also focus on programming concepts that make it easier to specify time-synchronized activities earlier in the design process.
NI cofounder Jeff Kodoskey said in his keynote at NI Week: “Most applications are concerned with performance rather than accurate time. Cyber-physical systems depend on accurate time. But, currently there isn’t any way to express this upfront in the design process. Although it is possible to build precisely timed cyber-physical systems today it would be simpler if we had a precise representation of time.”
Distributed real-time control
David Fuller, vice president of application and embedded software at NI, said cyber-physical systems built from distributed compute nodes “have to have a common notion of time. You have to have some standardized mechanisms so everything is on the same watch. The power of those things to do control comes down to the timescales that you can achieve in IoT. The less refined and accurate timing is, the less you can do.”
Fuller used the example of smart thermostats such as the Nest to illustrate what is possible with a loose notion of time. “To do more advanced things you need a more sophisticated notion of time,” he said. “We need to allow for globally defined standards of time. And needs to be adopted in the core communications technologies. IoT systems will have to have first class notion of time for electronics and software. To control anything we you have to have a notion of time.”
At NI Week, Professor Gerhard Fettweis of TU Dresden used the example of globally asynchronous, locally synchronous (GALS) concepts in chip design as an analogy for what is needed. “The synchronization all depends on the scale of time. If it is very local I need much tighter control than if I am further away.”
In other work, University of Illinois professors Lui Sha and Jose Meseguer have developed an analogous approach to GALS for distributed systems, calling their technique physically asynchronous, locally synchronous (PALS) design, which demands that clocks on each controller remain in sync with others in the system. For a dual redundant flight guidance system prototype, the results showed that PALS reduced verification time from 35 hours to less than 30 seconds compared with traditional, purely event-driven design.
Columbia University Professor Nick Maxemchuk has led a team developing protocols that use timing synchronization for automated vehicles. “In our communications, all messages are scheduled, so that all cars know when they have missed a message, and can begin recovery. Not sending a scheduled message is a reliable way to inform the other cars that you cannot take part in the maneuver.”
Truchard said NI is working on closely coupled high-rate processing to cope with the demands of communications and similar applications. “We are also working on high speed multi-rate computation for applications such as software-defined radio. We will offer 100MHz computation rates. By providing this capability I believe we will once again be able to raise the bar on time,” he said.