AI-based modelling of individual user preferences for dynamic control of future-oriented LED lighting systems

本研究项目致力于开发一种基于人工智能的方法，根据连续记录的传感器输入值，推导出动态自适应的、面向用户的灯具控制曲线，以实现LED室内照明的HCL兼容系统解决方案。术语“控制曲线”概括了对于每个灯具可以或多或少独立地改变的所有参数(例如，照度、直接/间接光的比例、色温、间接蓝光的比例等)。为了确定最优的基于人工智能的控制曲线，受试者在一天中的不同时间进行测试，以根据主观心理和外部影响因素确定他们的个人照明偏好，并使用机器学习方法将这些偏好与相应的传感器数据相关联。其目的是开发一种照明系统，该系统基于输入的传感器数据和潜在的用户偏好模型，能够预测当前情况的最优控制曲线，并在一天中动态调整房间的照明。为了实现这一目标，将进行第一次研究，通过记录用户在几个工作周内不同时间点相对于房间当前照明情况的偏好来确定单独的控制曲线。然后，通过基于人工智能的评估，将这些用户偏好与同时记录的传感器数据相关联。根据该数据，根据环境条件和特定于对象的影响因素，以这种方式确定的控制曲线被分组为单独的特征簇。在进一步的迭代步骤中，将进行第二次研究，然后将这些聚集的控制曲线呈现给参与者，以便通过提供各种调整选项来进一步适应他们的个人偏好。这些额外的输入值与连续记录的传感器数据一起，用于进一步优化控制曲线。在研究项目的第三部分，将对这些用户偏好的最优控制曲线进行生理学评估，以解决警报效果问题，并最终推导出相应的用户偏好模型，作为未来智能照明系统的基础。

This research project deals with the development of an AI-based approach for the derivation of dynamically adaptive, user-oriented luminaire control curves on the basis of continuously recorded sensor input values for the realization of HCL-compatible system solutions for LED interior lighting. The term "control curve" summarizes all those parameters that can be varied more or less independently for each luminaire (these are e.g. illuminance, proportion of direct/indirect light, color temperature, proportion of indirect blue light, etc.). In order to determine optimal AI-based control curves, subjects are tested at different times of the day to determine their individual lighting preferences depending on both subjective-psychological and external influencing factors and to correlate these with the corresponding sensor data using machine-learning methods. The aim is to develop a lighting system which, based on the incoming sensor data and an underlying user preference model, is capable of predicting the optimal control curve for the current situation and dynamically adjusting the illumination of the room over the course of the day. To reach this goal, a first study will be conducted where individual control curves are determined by recording user preferences in relation to the current lighting situation in the room at different time points over several working weeks. These user preferences are then related to the simultaneously recorded sensor data by means of AI-based evaluation. From this data, the control curves determined in this way are grouped into individual characteristic clusters depending on the environmental conditions and the subject-specific influencing factors. In a further iteration step, a second study will be conducted where these clustered control curves are then presented to the participants for further adaptation to their individual preferences by providing various adjustment options. These additional input values, together with the continuously recorded sensor data, serve to further optimize the control curves. In the third part of the research project, a physiological evaluation of these user-preferred optimal control curves will be carried out in order to address the question of alerting effects and to finally derive a corresponding user preference model as a basis for future, intelligent lighting systems.