C O V E R F E A T U R E The Discipline of Embedded Systems Design Thomas A Henzinger EPFL Joseph Sifakis Verimag The wall between computer science and electrical engineering has kept the potential of embedded systems at bay It is time to build a new scientific foundation with embedded systems design as the cornerstone which will ensure a systematic and even handed integration of the two fields C omputer science is maturing Researchers have solved many of the discipline s original defining problems and many of those that remain require a breakthrough that is impossible to foresee Many current research challenges the Semantic Web nanotechnologies computational biology and sensor networks for example are pushing existing technology to the limits and into new applications Many of the brightest students no longer aim to become computer scientists but choose to enter directly into the life sciences or nanoengineering 1 At the same time computer technology has become ubiquitous in daily life and embedded software is controlling communication transportation and medical systems From smart buildings to automated highways the opportunities seem unlimited yet the costs are often prohibitive and dependability is generally poor The automotive industry is a good example As each car receives an ever increasing number of electronic control units software complexity escalates to the point that current development processes and tools can no longer ensure sufficiently reliable systems at affordable cost Paradoxically the shortcomings of current design validation and maintenance processes make software the most costly and least reliable part of embedded applications As a result industries cannot capitalize on the huge potential that emerging hardware and communication technologies offer 32 Computer We see the main culprit as the lack of rigorous techniques for embedded systems design At one extreme computer science research has largely ignored embedded systems using abstractions that actually remove physical constraints from consideration At the other embedded systems design goes beyond the traditional expertise of electrical engineers because computation and software are integral parts of embedded systems Fortunately with crises comes opportunity in this case the chance to reinvigorate computer science research by focusing on embedded systems design The embedded systems design problem certainly raises technology questions but more important it requires building a new scientific foundation that will systematically and even handedly integrate computation and physicality from the bottom up 2 Support for this foundation will require enriching computer science paradigms to encompass models and methods traditionally found in electrical engineering 3 4 In parallel educators will need to renew the computer science curriculum In industry trained electrical engineers routinely design software architectures and trained computer scientists routinely deal with physical constraints Yet embedded systems design is peripheral to both computer science and electrical engineering curricula Much of the cultural wall between the two fields can be traced to differences between the discrete mathematics of computer science and the continuous Published by the IEEE Computer Society 0018 9162 07 25 00 2007 IEEE mathematics of traditional engineering The industry desperately needs engineers who feel equally at home in both worlds The embedded systems design discipline has the potential to produce such integrated talent But defining its scientific foundation will take a concerted coordinated effort on the part of research academia industry and policy makers Band limited white noise Scope Input generator Discrete controller xd xo x xo Jitter In1 Out1 Pendulum x xo xo THE DESIGN PROBLEM An embedded system is an engineering artifact involving computation that is subject to Figure 1 An analytical model The block diagram models an inverted penphysical constraints The physical constraints dulum controlled by a discrete controller described in Matlab s Simulink arise through the two ways that computa From feedback signals and a periodic input the controller generates an tional processes interact with the physical ideal control signal which jitter and noise functions transform world reaction to a physical environment and execution on a physical platform Common reaction con design theories and practices for hardware and software straints specify deadlines throughput and jitter and orig are tailored toward the individual properties of these inate from behavioral requirements Common execution two domains often using abstractions that are diametconstraints bound available processor speeds power rically opposed and hardware failure rates and originate from impleHardware systems designers for example compose mentation choices Control theory deals with reaction a system from interconnected inherently parallel buildconstraints computer engineering deals with execution ing blocks which can represent transistors logic gates constraints The key to embedded systems design is gain functional components such as adders or architectural ing control of the interplay between computation and components such as processors Although the abstracboth kinds of constraints to meet a given set of require tion level changes the building blocks are always determents on a given implementation platform ministic or probabilistic and their composition is determined by how data flows among them A building General versus embedded systems design block s formal semantics consist of a transfer function Systems design derives an abstract system representa typically specified by equations Thus the basic operation from requirements a model from which a sys tion for constructing hardware models is the compositem can be generated automatically Software design tion of transfer functions This type of equation based for example derives a program from which a compiler model is an analytical model such as the example in can generate code hardware design derives a hardware Figure 1 Examples of analytical models include netlists description from which a computer aided design tool dataflow diagrams and other notations for describing can synthesize a circuit In both domains the design system structure process usually mixes bottom up activities such as the Software systems designers in contrast use sequential reuse and adaptation of component libraries and top building blocks such as