Excerpts from a white paper.
Today's automobiles contain many complex electronic systems, each of which may incorporate a large number of electronic control units (ECUs) performing a single function, communicating through layers of networks. Even as complexity increases, design cycles are under pressure to shorten, so that manufacturers can deliver the latest in safety, fuel efficiency and convenience to consumers in a highly competitive industry. At the same time, quality and reliability remain of paramount concern. The challenges of managing these factorscomplex technology, time-to-market and qualityare dictating a new approach to automotive electronics, yet one that is as old as the automotive industry itself.
Since its inception, the automotive industry has used models for the physical and mechanical aspects of vehicle design and development. Engineers build scale models of cars to predict how the real vehicles will look, feel and behave long before production begins. This tried-and-true method is equally appropriate for the design and development of automotive electronics. Modeling the silicon systems, ECUs and entire networks of ECUs in automobiles allows engineers to evaluate design alternatives, weigh tradeoffs and predict results long before the electronics are built. Modeling also simplifies troubleshooting and can reveal dangerous interactions between systems early, to prevent potentially disastrous results that may result in product recalls.
This paper discusses the changes in automotive electronics that dictate the evolution to model-based design and virtual prototyping. Virtual prototyping is contrasted with bench testing to illustrate the specific differences. Other software-based modeling techniques are described to clarify the requirements of an effective model. Finally, virtual prototyping is explored in depth and an example of an automotive safety system designed using a model-based process is provided.
The Challenge of Embedded Systems
The explosive growth of electronics in the automotive industry, especially the growth of embedded system software, changes the dynamics of automotive design and presents significant challenges. Now that there are so many ECUsup to 70 in a vehicle, connected by up to five busesproject managers and engineers are faced with new issues arising from how those ECUs behave internally and how they interact with each other. Warranty and quality issues hinge on being able to diagnose difficult hardware/ software problems within and between ECUs. Bus communication between ECUs is complex and critical.
These new challenges escalate when coupled with the demands of a highly competitive industry. Automotive OEMs and Tier 1 suppliers who can bring new functionality to market in the areas of safety, fuel efficiency and convenience will reap the benefits of market share. But this new functionality must meet stringent quality standards, at competitive costs. The cost/quality dynamic is very difficult given the increased complexity of ECUs, the interactions between them and the software proliferation within them. A BMW executive reported at the Jan 2004 automotive electronics conference in San Diego, California, that 50-70% of the development cost of an ECU is related to software and already 40% of a vehicle's cost is determined by electronics and software.
The confluence of these factors is driving the change in automotive electronics design methodology to model-based design, using virtual prototypes. The traditional hardware-based bench development methodology no longer suffices. Car manufacturers realize the difficulties of changing to an embedded systems approach, and are showing a new level of attention to embedded software development success in other markets. For example, General Motors reportedly has two major, high-level strategic programs: fuel cells and embedded software.
Growth of ECUs and Networks
ECUs used to be relatively simple, hardware-oriented systems. Today, they are multi-purpose, multi-chip computer systems where more functionality often is delivered in software than hardware. The most complex ECUs operate the power train. Simpler ones operate functions such as seat and mirror adjustments, but even these ECUs need to be networked so that the seat and mirrors can be adjusted for different drivers. The massive network requires a wiring harness that can weigh over 50 Kg; its many connectors contribute negatively to automotive reliability and hence to increased warranty costs.
The trend of increasing automotive electronic content is the direct result of many new features that will greatly increase both safety and comfort but that will require more sophisticated ECUs with a large embedded software component. The safety features include steer-by-wire, brake-by-wire and drive-by-wire (collectively known as "X-by-wire"), automatic lane-following, drowsy driver detection, intelligent cruise control and airbag systems that can adjust deployment based upon passenger weight and the specific nature of an accident. Improving fuel efficiency is an important goal: hybrid (internal combustion and battery) and fuel-cell electric drive place high demands on software. Telematics and in-car entertainment will further increase the electronic content of cars, requiring the combination of such technologies as wireless connectivity, global positioning, digital radio, and Internet access, all with hands-free voice activation whenever possible.
Various factors work against each other to create a complex optimization problem. To reduce wiring weight, ECUs should be located close to whatever they control: for example, the engine control ECU should be mounted in the engine compartment and brake controls should be close to the wheels. This proliferates ECUs, with each performing a few specific functions. On the other hand, to reduce connector count, the number of ECUs should be as low as possible. But to keep reliability up, inter-ECU wiring must be kept to a minimum and redundancy for the most critical functions must be built into the system. The network connecting the ECUs needs to guarantee service for functions such as X-by-wire.
Resolving this series of contradictory requirements requires a more systemic approach to the architecture of the electronics system. Many potential trade-offs must be evaluated before the overall hardware/software system is defined and detailed development can begin. Evaluation, in turn, requires the capability to do quantitative analysis of potential architectures of both networks of ECUs and of the internals of the ECUs themselves.