The goal of Advanced Life Support Systems (ALSS) is to provide self-sufficiency in life support for productive research and exploration in space. Important in reaching this goal is the production of crop plants in one or more controlled environments for the purpose of providing life essential food, air, and water to a human crew. To do this reliably and efficiently, it is necessary to achieve control of the rates of various plant physiology processes.
To develop an efficient control system that will be able to manage, control, and optimize plant-based life support functions, system identification and modeling of plant growth behavior must first be accomplished. We have developed a plant growth (physiology) model using artificial neural networks. Neural networks are suitable for both steady-state and dynamic modeling and identification tasks, since they can be trained to approximate arbitrary nonlinear input-output mappings from a collection of input and output examples. In addition, they can be expanded to incorporate a large number of inputs and outputs as required, which makes it simple to model multivariable systems.
In this paper, we describe our motivation and approach to developing these models and the neural network architecture. Initial use of the artificial neural network for modeling the single plant process of transpiration is presented. The approach is to develop and validate neural network submodels describing the individual plant-based functions (assimilation, biomass allocation and accumulation, and resource demands) and to integrate them for full control of plant-based life support functions. With the use of neural networks, these complex, nonlinear, dynamic, multimodal, multivariable plant growth models will be able to better interpolate between various environmental conditions and parameters and be able to simulate responses and performance of various plants.