Optimizing and Controlling Nonlinear Metrics of Multichannel LED Systems With Ambient Light Interference via Reinforcement Learning

Controlling multichannel LED systems is essential in many applications, ranging from stage lighting to multispectral imaging and light therapies. However, the increasing number of channels coupled in a compact package and housing leads to higher nonlinearity in the photo-electro-thermal response. This presents a substantial control challenge for tracking nonlinear lighting indices, which has not yet been solved in the existing literature, especially in the presence of ambient light interference. To address these challenges, this work proposes a spectral optimization and control framework based on reinforcement learning for multichannel LED lighting systems. By comprehensively considering ambient light interference throughout the modeling and training stages, the framework effectively handles the complex nonlinear optimization and control tasks. A seven-channel LED control and measurement system incorporating ambient light interference was established to evaluate its effectiveness in experiments. Through rigorous testing, the proposed framework demonstrated its ability to successfully execute complex lighting control tasks, including achieving desired color rendering index, color difference, and intensity error, in the presence of ambient disturbance.