Benefits of fuzzy logic in pig house atmosphere control

Actually, the regulators dedicated to pig-house atmosphere follow the same principle. Their working is in fact essentially based on a temperature set-point and on a temperature range. However, a statistical study has shown that the atmospheres obtained with this kind of controllers do not satisfy the zootechnic constraints in a number of situations.Nevertheless, the specification and realization of a correct regulator seem extremely hard to carry out. This is the reason why we propose to use a learning method to tune automatically the controller. To implement this regulator, we have chosen a Fuzzy Inference System with a modifiable linguistic rule conclusion part.To tune these conclusions, the learner uses a reinforcement learning method that consists in modifying the conclusions to optimize the global sum of reinforcement signals received over time during the interaction with the environment. In the computer science community, these reinforcement signals consist of rewards and punishments, and allow to indicate to the learner the zone to reach and to avoid respectively.After a series of experimental studies, it appears that the atmospheres obtained with this new kind of controller satisfy totally the zootechnic constraints expressed by the experts. Indeed, the fuzzy controller reach to maintain temperature and hygrometry in the goal zones and follow a good policy when external conditions are unfavorable.