Outcomes

  1. The design of a new kind of learning approach where artificial systems do not replace humans but interact, adjust and cooperate with them in order to get a shared representation to reach a same objective. Household are very interesting for the design of such an approach because inhabited systems are complex in many regards: physics in complex and is highly dependent on occupant behavior, human practices depend on desires and intentions that cannot be explicated on a regular basis. It means that knowledge must be shared but not transferred: quantitative and historical data are known by the artificial systems and preferences, desire and intentions are known by the inhabitants. Artificial systems have some perception (sensors) and action (actuators) capability whereas human actors have others: cooperation is the only way in energy management at home. It will be one of the significant innovation of the project.
  2. The usage of a pure data model approach to generate energy management advice and explanations to improve the sobriety and the flexibility of inhabitants is a second significant innovation of the project. Indeed, each home is composed of an envelope, appliances and inhabitants: each instance of this triptych is both unique and evolutive. Therefore, all the existing approaches that assume the existence of a knowledge model, generally stated as equations, are very costly because (a) it has to be develop for each home setting (b) it has to be regularly updated. Involving human actors in the search of better solutions is more engaging than trying to guess what is not realistic to explicate… and potentially doing mistakes that will alter the belief in the relevance of the artificial system reasoning.
  3. The home dashboards are currently unidirectional: they basically provides information about the current status of a home setting in its environment. The project yield another kind of dashboards, called symmetrical engaging dashboards. We believe it is more engaging because it gives a more significant role to the inhabitants. According to our first tests where interactions are limited to most interesting contexts i.e. the new ones, not only it’s not going to be boring for the inhabitants but it’s going to led to a kind of personification of the home… We believe that cooperating with artificial systems will look natural… The human actors could experiment new configurations in certain contexts, possibly suggested by the symmetrical dashboard, and the artificial system will record and analyze the resulting impacts. In this way, better usage of a home by its occupants could be discovered. Is not it the real intelligence? The cooperative intelligence?
  4. Social landlords are particularly interested in any innovative methods/tools enabling inhabitants to get involved in their energy uses and bills management. Consequently, particular interest will be paid to present main results of RF3 experiments to social landlords. In this regard, main results of RF3 experiments will be presented to the “Kocliko Lab” which is currently being set up (gathering the French social housing federation, Kocliko’s social landlords clients and Kocliko). In addition, the most promising methodologies could be integrated into Kocliko’s commercial offer.