Comparing Architectural Design Styles

ion should also suppress implementation decisions. Several designs unnecessarily reveal implementation decisions: Booth and Yin and Tanik included a global system clock and the fact that it has a millisecond resolution. Kirby included extensive information about range, resolution, and accuracy of values. Ward and Keskar included details of the defmition of events at the beginning of the design. On the other hand, Ward and Keskar did not include timing information, even though they claimed to be producing a real-time design. include some provision that leads the designer to consider an even more serious problem: the possibility that the cruise control’s model of current speed is radically different from the actual speed. This can happen, for example, if the sensor is on a nondrive wheel and the drive wheels start spinning. It can also happen if the sensor fails. The Kirby p roblem formulation calls for a calibration capability, but this addresses only the problem of gradual drift, not that of sudden inaccuracy. The only designs that explicitly direct the designer’s attention to the accuracy of the current speed model are Higgins’ and my feedback-loop models. This is not surprising, since these are the only two designs that explicitly treat the problem as one of feedback control. Safety. Can the system fully control the vehicle, and can it ensure that the vehicle will not enter an unsafe state as a consequence of the control? Cruise-control systems exercise largely auto-nomous control over moving machines. De-signers of such systems should consider explicitly if the machines will be safe in operation. Two particular questions arise here: Does the system have full control over the speed of the vehicle? Can the system’s model lntegrotion with the vehicle. How well do inputs and outputs match the actual controls? How do the manual and automatic modes interact? What are the characteristics of real-time response? How rapidly and smoothly does the vehicle respond to control inputs? How accurately does the system maintain speed? Ultimately, the designed system must not only satisfy the problem As a result, if the car picks up excess speed (coasting down a hill, for example), the system cannot slow it down. The only designs that recognize this problem are Kirby’s NRL/SCR design, Atlee and Gannon’s state-based formulation, I !