For the first
time, a simulation technology has demonstrated the simultaneous simulation of
84 electronic control unit (ECU) models with very high performance and scalability,
while maintaining the timing accuracy of each ECU.
The ECUs each contained an ARM1176 processor with bus, memory, peripheral devices,
LCD and network interconnect controllers, and were connected by system network
fabrics (in this case, CAN-type buses) into 7 subsystems. Half the ECUs in each
subsystem controlled a Simulink-type ordinary differential equation plant model
(42 Simulink-type models in total), and the other half executed multi-media like
applications on top of a Linux OS. With 126 ECU and plant models executing simultaneously
under the EST simulation technology, the measured aggregate performance of the
processor models was 3.2 billion instructions per second. This is approximately
30,000 times faster than current competing timing accuracy preserving, simulation
technologies.
Translated to the engineering of automobiles: the simulation was the equivalent
of running the real-time control systems of 3 cars communicating together to
effect accurate joint control while running high-end infotainment systems in
each car. A physical car of this complexity, would execute about 1 billion instructions
per second in the same circumstance. This is faster than real-time.
EST’s technologies and methodologies enable the efficient and effective (and
optimal) design of complex control architectures, operating under severe real-time
constraints, in order to control cars, airplanes, etc. Interestingly, the same
architecture is instrumental in structuring the supply chain that designs and
manufactures validated physical super-systems.
|
