|
Automobiles are becoming computer environments on wheels, filled with embedded electronic systems that power everything from the drive train and automatic rollover sensing to climate control and infotainment systems. As embedded systems proliferate, the auto industry is facing increasing design challenges. Design engineers are under increased pressure to deliver high-quality software on time because increased software functionality is driving this proliferation of embedded automotive systems.
These articles will provide background information on software proliferation in automotive; outline the common reasons for late delivery of software and the ways in which traditional design methodologies are failing; and explain how leveraging the automotive industry's familiarity with computer-based modeling techniques is increasing acceptance of virtual system prototypes (VSPs)— fully accurate and high performance simulation models of digital hardware—to shorten the development cycle while reducing cost and improving quality.
With their ability to efficiently iterate across multiple architecture alternatives—whether at the microcontroller, the ECU, or the network level—VSPs are allowing developers to fine tune system designs using realistic software loads without the need for expensive hardware design. At the same time, the ability to develop software on such models greatly increases the quality of systems and reduces the risk of late delivery.
Emerging standards, such as those found in communication protocols (i.e. FlexRay) or in software standardization (i.e. AUTOSAR) still require understanding in advance of the exact behavior for critical real-time reactive systems. Virtual system prototypes provide the best means for accurate communication of system level behavior in terms of both software and hardware. The value of communicating executable specifications applies all across the supply chain for automotive electronics.
Automotive software proliferation
As automotive electronics design is increasing in software functionality, designers are under increasing time pressure in delivering high quality software. This functionality is driven by the proliferation of embedded automotive systems. A BMW executive reported at a recent US automotive electronics conference [Ref. 1] that 50-70% of the development cost of an electronic control unit (ECU) for an automotive application is related to software and 40% of a vehicle's cost is determined by electronics and software. Yet, according to a survey of embedded developers conducted in June 2006 by Venture Development Corporation (VDC), 37% of embedded system projects in the automotive industry for power train, safety, chassis, and infotainment applications are delivered late.
"These developers cited unrealistic schedules, changes to the specifications, and inadequate specifications as the main reasons for late projects," said Matt Volckmann, senior analyst at VDC. Applications complexity and too few developers and testers are additional problem factors. This trend does not sit well, because projections are that software alone will soon account for over 12% of the value of a car.
Why is it so difficult to deliver software on time? To begin with, automotive software complexity is growing (see figure below). Today's high-end automobiles may have up to 100 ECUs connected by up to five buses under the control of millions of lines of software interacting in real time. Add to this the advent of highly complex 32-bit micro architectures, decreased time to start-of-production (SoP) and the need for better quality automotive systems.
IMAGE: Toyota
View a full-size image
It is becoming more obvious that the old sequential design methodology, where hardware must be created before software development can begin, can no longer serve the needs of automotive OEMs, their suppliers, and the end user. Challenges faced by design teams are steep and growing; the need for radical improvements in the development and delivery process is becoming critical to deliver working systems that are right the first time.
Traditional design method inadequacies
Automotive electronics system development requires an extremely high level of quality and reliability at a very low cost. Automotive industry Tier 1 suppliers often do not have the same time-to-market concerns as other industries do, however cost-cutting measures are paramount. To meet ever-increasing demand for lower costs, it is imperative that development teams are able to complete more work in shorter time while simultaneously getting the design and implementation correct on the first try. Rushed and late deliveries often contain hard-to-find bugs and may lack features, resulting in both future development and warranty costs that can quickly erode already-small profit margins on components.
In addition to tough market constraints, the complexity of automotive electronics is rapidly rising. New features and market demands such as hybrid power train, stability control, and infotainment systems are pushing the envelope of automotive electronics by introducing tightly coupled multi-core ECUs running millions of lines of code. At the same time, product development windows are shrinking. Major OEMs are pushing to go from concept to production in as little as 18 months.
Add to this the introduction of complex new hardware standards such as FlexRay. Ultimately, these standards have the goal to improve the predictability and quality of new systems—yet it is extremely difficult to implement new systems on top of such new hardware standards without having the ability to analyze their impact in a system context.
The software content of automotive electronics is increasing at an exponential rate. More and more the intellectual property (IP) that differentiates one automotive company from another is defined by the implementation of features found in software. For many, the solution to the problem of increasing software features is covered under the AUTOSAR initiative, which is a standards organization created to provide an open source framework for vehicular software.
AUTOSAR will provide a middleware runtime environment that will abstract hardware dependencies to a common denominator. That will allow application software to be the sole differentiator. While this may be applicable for simpler control systems, such as those found in body and dashboard applications, the real-time critical systems will need to remove or circumvent some of the API levels defined by AUTOSAR. (An excellent discussion of the implications of AUTOSAR for automotive semiconductor suppliers can be found in Ref. 2.)
Traditional design methodology exacerbates the problems created by rising complexity, increasing software features, time-to-market pressures, and new standards. Poor quality and rising costs and must be addressed through improved design methods to enable a more accurate and comprehensive specification process, better architectural design, and improved quality of implementation.
When you combine the problem factors above with conventional systems development—where hardware design must be complete before software design begins—it is obvious that a sequential process simply cannot support the development of today's complex systems. Only new design methods capable of addressing all these factors promise to contain the rising cost imposed by complexity growth.
|
