Qualitative-evolutionary design of greenhouse environment control agents

A novel technique that fuses qualitative reasoning and evolutionary computa tion in the design of control agents for the control of the environment of a greenhouse is proposed in this paper. Linguistic rules relate the attribu tes of the greenhouse performance in response to perturbations in the desir ed set points to the suitability of the agent. De-fuzzification of the fuzz y suitability membership function yields a quantitative measure of the fitn ess of the agent to satisfy the system performance specifications. The fitn ess measure is used in a genetic algorithm that performs a stochastic searc h for the global optimum parameters of the agent. (C) 2000 Elsevier Science B.V. All rights reserved.