The main research question the LearningHome project tackles is « how to involve inhabitants of residential buildings in the everyday energy management in order to support them to improve their sobriety but also their flexibility while getting the expected comfort (thermal, air quality, limited number of actions) and limiting environmental and economic impacts ». Some solutions have been proposed for guiding inhabitants thanks to advice, analysis and explanation but it is admitted that it is not realistic to design a detailed physical model of every home behavior for the sake of time, expertise and therefore cost: it is one of the scientific lock the project will address. The research question will be solved in replacing behavioral knowledge model by historical data coming from an as-limited-as-possible number of sensors, but also possibly, from interaction with inhabitants. Different approaches for involving inhabitants are going to be compared: existing ones, like Kocliko dashboards and analysis reports, and LearningHome ones, like detailed analysis reports, gamification with competitions and contests in collective buildings and finally interactive and cooperative approaches, with engaging dashboards to encourage and maintain changes of inhabitant behaviors. Innovative case based reasoning together with interactive and cooperative learning engaging dashboards are going to be developed and tested. Depending on the approaches, the assessment is going to be done either with few experimental homes with a high density of sensors (with a study of what is possible when reducing the number of sensors), or with many homes (typically 100 to 200) in collective buildings with a limited number of sensors.
The LearningHome project starts with the key findings of the INVOLVED project but focusing on the following limitation: each site is unique because of its architecture, equipment and sensors but also because of its occupants. The site models, including sensors locations and human activities/preferences used in INVOLVED are mostly not available. Therefore, an alternative approach has to be designed considering that site knowledge results from a symmetrical confrontation between occupants’ knowledge and the knowledge of an artificial system, called Interactive Home Energy Management Aid System (IHEMAS). Each party is informing, explaining, asking, suggesting and learning from the other in a symmetrical way. In {tooltip}1{end-texte}A. Awada, P. Reignier, S. Ploix, M. Jacomino, A. El Safadi, Identification collaborative d’activités dans une zone habitée. Conference IBPSA France, 12-13 November 2020, Reims, France.{end-tooltip}, we showed a first application of a symmetrical co-definition of activities where occupants depicted daily what they did with their own meaningful labels and the IHEMAS analyzes whether these labels can match its own perception based on sensor data and, if not, it suggests a more global activity label or labeling errors depending on the context. In LearningHome, we want to go further in (1) actively discovering non Pareto-optimal occupant behaviors regarding costs/comforts compromises by driving occupants in exploring the space of possible actions and generating contextual dynamic causal explanations {tooltip}2{end-texte}A. A. Alyafi, V. B. Nguyen, Y. Laurillau, P. Reignier, S. Ploix, G. Calvary, J. Coutaz, M. Pal, and J. P. Guilbaud. From usable to incentive-building energy management systems. Modeling and Using Context, 2(Issue 1), 2018.{end-tooltip} (2) validating the efficiency of the approach by comparing it to different ways of pricing and/or billing, of involving by gaming approaches, of providing inhabitants with regular home energy/activity reports.
1.. A. Awada, P. Reignier, S. Ploix, M. Jacomino, A. El Safadi, Identification collaborative d’activités dans une zone habitée. Conference IBPSA France, 12-13 November 2020, Reims, France.↩
2.. A. A. Alyafi, V. B. Nguyen, Y. Laurillau, P. Reignier, S. Ploix, G. Calvary, J. Coutaz, M. Pal, and J. P. Guilbaud. From usable to incentive-building energy management systems. Modeling and Using Context, 2(Issue 1), 2018. ↩