Image Processing Reference
In-Depth Information
•
Target part
: Represented by a highly efficient and portable-embedded software package
offering a signal-based application programmer interface (API), handling of multi-
ple protocols, integrated gateway functionality, and post-compile-time reconfiguration
capability, together with a PC-based generation tool.
he Volcano approach is particularly beneficial as the complexity of vehicles is increasing very rapidly
and as projects will have to cope with new functions and requirements throughout their lifetime. he
computing industry has discovered over the last years that certain techniques are needed in order
to manage complex software systems. Two of these techniques are abstraction (where unnecessary
information is hidden) and composability (if software components proven to be correct are com-
bined, then the resulting system will be correct as well). Volcano is making heavy use of both of these
techniques.
The automotive industry is implementing an increasing number of functions in software. Intro-
duction of protocols like MOST for multimedia and FlexRay for active chassis systems result in
highly complex electrical architectures. Finally, all these complex subnetworks are linked through
gateways. he behavior of the entire car network has a crucial influence upon the car's performance
and reliability. To manage software development involving many suppliers, hundreds of thousands of
lines of code and thousands of signals require a structured systems engineering approach. Inherent
in the concept of systems engineering is a clear partitioning of the architecture, requirements, and
responsibilities.
A modern vehicle includes a number of microprocessor-based components called ECUs, provided
by a variety of suppliers.
Control area network (CAN) provides an industry-standard solution for connecting ECUs
together using a single broadcast bus. A shared broadcast bus makes it much easier to add desired
functionality: ECUs can be added easily, and they can communicate data easily and cheaply (adding
a function may be “just software”). But increased functionality leads to more software and greater
complexity. Testing a module for conformance to timing requirements is the most difficult of the
problems. With a shared broadcast bus, the timing performance of the bus might not be known until
all the modules are delivered and the bus usage of each is known. Testing for timing conformance can
only then begin (which is often too far into the development of a vehicle to find and correct major
timing errors). he supplier of a module can only do limited testing for timing conformance—they
do not have a complete picture of the final load placed on the bus. his is particularly important when
dealing with the CAN bus—arrivals of frames from the bus may cause interrupts on a module wish-
ing to receive the frames, and so the load on the microprocessor in the ECU is partially dependent
on the bus load.
It is often thought that CAN is somehow unpredictable and the latencies for lower priority frames
in the network are unbounded. his is untrue, and in fact CAN is a highly predictable communica-
tions protocol. Furthermore, CAN is well suited to handle large amounts of traffic with differing time
constraints.
However, with CAN there are a few particular problems:
•
Distribution of identifiers
•
CAN uses identifiers for two purposes: distinguishing different messages on the bus, and
assigning relative priorities to those messages—the latter being often neglected
•
Limited bandwidth, due to low maximum signaling speed of Mbps, is further reduced
by significant protocol overhead
Volcano was designed to provide abstraction, composability, and identifier distribution reflecting
true urgencies, and at the same time providing the most efficient utilization of the protocol.
