Real-Time Closed-Loop Color Control

Abstract

Smart homes and Internet of Things are emerging concepts in modern society, with intelligent lighting being an important part of it. Besides providing visual satisfaction through its color-rendering properties, lighting also has other effects on human well-being. In order to exploit the full potential of a smartly lit home, lighting systems need to be equipped with accurate controllers that can control the spectrum and color characteristics of light in addition to conventional on-off and dimming control.

However, current commercial smart lighting products with such capabilities need to employ expensive sensors which are still lacking in terms of closed-loop feedback which is imperative for accurate color control of light-emitting diode (LED)-based luminaires. This paper presents a novel approach that uses the camera available on modern smartphones to perform closed-loop color control for lighting systems in smart homes.

The algorithm is able to perform multi-channel mixing for any color and also white light at a desired correlated color temperature with high color-rendering index. This approach proves to be very economical and convenient as no external sensors are required and can be performed using any Android smartphone on a compatible LED-based luminaire.

Introduction

Light emitting diodes (LEDs) are steadily gaining ground in lighting applications all around the world. It was reported that in the United States alone, the installations of LED products in all lighting applications have more than quadrupled from the year 2014 till 2016. The U.S. Department of Energy also forecasts that the penetration of LED-based luminaires will increase dramatically to about 86% in general lighting applications by the year 2035.

Many consumers are moving towards LEDs because of its lower power consumption compared to traditional light sources such as halogens and fluorescents. Also, LED-based luminaires offer far greater advantages than just energy savings; they come in various spectral compositions and are also easily controllable, leading to lighting systems which are tunable.

Spectrally tunable lights are expected to be the future of lighting, as studies have shown that light is a significant stimulus in influencing the human biological clock, where it has been found that the spectral composition of light strongly impacts human physiology and psychology. The attraction of a tunable lighting system is that it can close the gap between artificial lights and natural light, offering huge benefits to human well being.

Light Spectrum Control Methodology

The luminaire prototype used to test the proposed control algorithm consists of 10 channels, of which 7 are primary colors with different peak wavelengths, whilst the remaining 3 channels are phosphor-converted white LEDs. The intensity of each LED channel is controlled using pulse-width modulation (PWM), which is supplied by an Arduino microcontroller on-board the luminaires to the LED driver wirelessly using ZigBee.

An Android application was developed to execute the lighting control algorithm. The user firstly selects the target lighting color using a color picker; the algorithm converts that color to a set of u'v' coordinates which is referred to as the target set point. The information of the room lighting condition is captured by the smartphone camera, which is converted to the u'v' color coordinates of the light in the room.

The Euclidian distance between the target and measured color coordinates is computed to produce the error. The PI controller receives this error, and considers the color coordinates of the LEDs, and generates the PWM control signal to each LED channel in the luminaire wirelessly using ZigBee.

LED Channel Specifications

Channel	CIE 1931 xy x	CIE 1931 xy y	1976 CIELUV u'	1976 CIELUV v'
Red (637 nm)	0.7020	0.2975	0.5436	0.5183
Amber (625 nm)	0.6817	0.3178	0.5003	0.5247
Yellow (596 nm)	0.5899	0.4093	0.3505	0.5472
Lime (538 nm)	0.4087	0.5601	0.1836	0.5662
Green (523 nm)	0.1804	0.7281	0.0634	0.5760

Design of Multi-Channel Color Control Algorithm

The novel multi-channel color control algorithm presented in this paper is a form of the gradient descent algorithm that converges to the target chromaticity. The computations are performed in the 1976 CIELUV color space which has a uniform chromaticity scale. For this algorithm to work, the (u', v') coordinates of each LED channel has to be obtained the first time the user runs the system.

The algorithm cycles through all LED channels in the luminaire individually while computing the color coordinates using the camera reading as input. The main objective of the algorithm is to develop a fastest and shortest travel path in the CIELUV color space for the color emitted by the LED luminaire to converge to the target color using the closed-loop control design.

The fact that the color resulting from adding two colors always falls on a line connecting the colors on the chromaticity diagram is used as a basis to iteratively arrive at the final intensities of each LED channel.

Algorithm Implementation

The first step in the algorithm is to scale the size of the image obtained from the camera by reducing the width and height of the image by 10 times respectively, thus resulting in a final image that is 100 times smaller than the original. Then, the average RGB values in the image are calculated.

The RGB values are then used to compute the measured color coordinates (u', v'). The error signal is computed using the Euclidian distance formula between the target and measured coordinates.

A proportional-integral (PI) controller is used in the design of the feedback control algorithm in order to achieve zero steady-state error. It was tuned using the well-known Ziegler-Nichols method to calculate the step size for each iteration which provides the algorithm with the capability of adaptive step sizes.

Experimental Results and Discussions

The experimental system was set up in a mock living room which is a 5.8 m × 3.4 m space, equipped with six wirelessly-controlled and tunable 10-channel research prototype luminaires. The seven channels which are pure-colored LEDs span across the visible wavelength range and can be mixed to obtain white light with a wide range of color properties.

The smartphone with its secondary camera facing upwards is used to capture the lighting condition, i.e. the RGB values and illuminance of the light incident on the surface where the phone is placed. A Konica Minolta CL-500A illuminance spectrophotometer is placed in close proximity of the smartphone to validate the color-control algorithm.

Bi-Channel Mixing Results

Experiment	Average Δu'v'	CCT Range	Average Absolute CCT Error	Average CRI
Warm white & cool white	0.0103	2700K to 5600K	4.45%	77.7
Cool white & yellow	0.0089	2700K to 5600K	3.62%	59

Multi-Channel Mixing Results

The color control algorithm was tested using various scenarios including:

• Seven primary colors to produce white light
• Ten LED channels to produce white light
• Seven primary colors to produce a colored light
• Ten LED channels to produce a colored light

For multi-channel mixing, the feedback algorithm was programmed to stop when it detects that the color difference Δu'v' is less than 0.003, a more stringent value compared to bi-channel mixing. This target was met for every chosen color detected by the smartphone camera.

The average CRI was considerably high at 82.76 for seven-channel mixing with pure-color LEDs and 94 for ten-channel mixing. By optimizing the choice of LED primaries in the luminaire, the number of LED channels required to produce a larger color gamut and high CRI light can be further reduced.

In terms of timing performance, each step in the closed-loop control takes approximately 658ms, with the algorithm taking about 10 iterations to converge the output of the luminaries from a random color to the target color. This is equivalent to about 6-7s. This rate of convergence by the algorithm is reasonable as well as acceptable in real-world applications.

Conclusion

This paper has presented a novel approach for color control of a multi-channel LED lighting system in a smart home environment using the camera found on most modern Android smartphones as the main feedback sensor. The algorithm is able to optimize the output spectrum of the luminaries to produce a light with tunable CCT, accurate color and high color rendering index.

A closed-loop feedback control system helps maintain robustness towards external disturbances from other light sources such as sunlight coming in through the windows. The algorithm is able to work at reasonable accuracy with potential improvement if customized camera calibration data is used.

The proposed method of using the smartphone as both a sensor and a processing unit proves to be very economical and convenient as no additional sensors are needed to be installed. Future work can include among other features, mood lighting based on user preference and the replication of a lighting scene that a user has captured on the smartphone from a different location.

References

The complete paper contains 39 references covering topics in LED lighting, color control algorithms, smartphone applications in IoT, and smart home technologies. Key references include works on:

• LED adoption forecasts by the U.S. Department of Energy
• Studies on the effects of light spectrum on human circadian rhythms
• Previous research on color control methods for LED systems
• Smartphone applications in home automation and IoT
• Color constancy algorithms and the Gray World Assumption
• Controller tuning methods including Ziegler-Nichols