Is fuzzy logic a practical choice in resource-constrained embedded control systems implemented using general-purpose micro controllers?

The emergence of "fuzzy logic control" (FLC) has been viewed by many researchers as a way of avoiding the need for complex mathematical analyses when developing control systems. Instead, FLC was originally seen as a way of using simple linguistic rules to implement an effective controller. Unfortunately, for many real applications, it has proved difficult to identify an appropriate set of such rules. The "second generation" of FLC techniques has therefore tended to focus on self-learning inference methodologies in which the required rules (and fuzzy sets) have been determined automatically. Despite some attractive features, many self-learning approaches present significant challenges for the developers of resource-constrained embedded control systems. One such challenge relates to what has been called the "rule explosion" problem: this describes the fact that FLC inference methodologies tend to create very large rule sets for multivariable control systems. In this paper, we compare both the performance and resource requirements of a "conventional" (LQR) and fuzzy control system for an inverted pendulum testbed. In each case, the controller is implemented on a low-cost microcontroller with limited CPU and memory resources. The results obtained reveal that the resource requirements of the FLC design greatly outweigh those of the "conventional" controller in this study (even when the fuzzy rule set is "pruned"). We suggest that where embedded systems have severe resource constraints and "off the shelf" microcontrollers are used - fuzzy control is unlikely to be a practical option for non-trivial control systems.