A Plant Taxonomy for Designing Control Experiments

Control experiments can have a significant impact on control theory by forcing researchers to confront real-world issues that affect design tradeoffs and performance specifications. Sensor and actuator constraints, modeling and identification issues, and hardware imperfections (such as noise, drift, bias, and nonlinearity) must all be addressed for successful controller implementation. Control experimentation, however, is not an established discipline, and there are many fundamental issues and questions that are worthy of deep and careful consideration. In particular, the definition of a control experiment is open, as are guidelines for verification and reproducibility. I will not attempt to address these important questions here. Rather, my objectives are limited to assessing plant features that are appropriate for investigating system-theoretic problems in feedback technology. The term system-theoretic refers to issues such as phase variation, nonlinearity, uncertainty, accessibility, and cross coupling, which transcend a specific application or hardware realization. This discussion of control experimentation venues is based on a plant taxonomy; that is, a systematic classification of plant properties and the challenges they present to control experimentation and, indirectly, to control engineering practice. Plant properties and the limits they impose on achievable performance under linear control are analyzed in [1]-[5] and the numerous references given therein. A related analysis in the context of active noise control is given in [6]. Many of the issues and tradeoffs discussed in [1]-[6], as well as additional issues of relevance to control experiments and hardware implementation, are addressed here. This article is partially motivated by [7], which lists 16 candidate plants for undergraduate control experiments. Their selection criteria are: interesting, visual, instructive, and reasonably challenging. I am especially interested in the features of these and other plants that render them worthwhile for experimentation for either education or research. Additional motivation is provided by [8], where I discussed the lessons I learned from working with students and colleagues in designing, building, and operating various control experiments. Details concerning the experiments described in [8] are given in [9]-[17]. My objective is to provide perspective on some of the issues that arise in designing control experiments for both education and research. I hope that this discussion will serve as useful guidance for educators designing control experiments to illuminate control issues, as well as for researchers developing experiments to complement theoretical and computational research in control.