Andrew Birnie, Systems Engineering Manager for Automotive Microcontrollers
and Processors. This week saw the publishing of the shocking United Nations
report on the transformation of Earth’s natural resources and
landscapes by humans. It is a timely reminder for us at NXP why we are
pushing so hard to move forward on more sustainable measures for
transportation. Even without the added impetus of social
responsibility, rapid innovation in automotive means that developing
solutions capable of responding to the requirements of new sustainable
vehicles is necessary.
One way of being able to deal more efficiently with these innovations is
through evolving electrical architecture to one of domain control. This means
that each functional area can be split into its own collaborating sub-network,
with a domain controller acting as an access point from the vehicle. Domain
control is a philosophy, not an ECU. Each controller in the hierarchy combines
to implement the functionality of the domain, and each controller can make
appropriate decisions for the purpose of the domain. Previously there
was one ECU per function, with low data exchange between them. But domain
architecture provides a more holistic efficient system level approach, with
reduced ECU count and reduced overall wiring.
In electrification, the propulsion domain a high-performance node that
delivers computational and mathematics performance. Predictive model control
techniques can be used to improve instantaneous efficiency of ICE or electric
motor using Kalman filtering or state space modeling. Torque decision making
can be improved by using an array of vehicle data combined with crowd sourced
data of other road users. That data can be utilized in mathematically intense
control strategies, using machine learning techniques, to realize a
significant increase in vehicle efficiency and therefore range. Studies have shown
that as you grant more predictive control to the domain controller, efficiency
savings of up to 30% are possible.
To deliver on this functionality and performance, many MCU suppliers have been
instantiating 6, 8, 10 cores and more. However, this approach can cause
software architecture challenges due to the difficulty of parallelizing the
system.
At NXP, we have struck a new path by splitting the MCU into its constituent
functions and implementing an optimum process technology for each. The
separate die with high-frequency interfaces are then combined into a
system-in-package. From the outside, the 3-die are invisible—they look,
behave and operate exactly like a traditional MCU, but with the full
performance of the base CMOS process.
Moving beyond the advantage of this SIP concept for energy optimization within
the propulsion controller, ECU consolidation cannot happen efficiently without
virtualization techniques like a hypervisor. Here, tasks can be separated
within an MCU into virtual machines (VMs), giving a method to assign and
reserve resources to each task. This helps guarantee freedom of
interference, which is essential within these consolidated ECUs where the
majority of the tasks are safety critical.
For electrification, where safe and energy efficient management of the torque
sources are needed, new microcontroller approaches are needed to deliver the
real-time performance, safety, security and virtualization in the MCU needed
to make it possible. This type of MCU requires innovative approaches.