AI and the future of "smart" lighting

1336 words / 6-minute read

The global dark-sky movement has achieved some notable victories. We have largely won the outdoor lighting hardware battle. The push for fully shielded luminaires has been successful. We have shown the efficacy of shielding as our best lever on reducing skyglow. But during the same period we saw a revolution in lighting technology nearly on par with the invention of artificial lighting itself. The arrival of light-emitting diode (LED) technology on the commercial market was a game-changer. LED has wonderful properties, controlling when lighting is on and where its light goes. It should have been a boon to light-pollution reduction efforts. Yet mostly that's not how it actually worked out.

Hardware has its limits. Shielding dictates where the light goes, but it doesn't control how much light is generated or when it is used. To make further meaningful reductions in skyglow, we have to look beyond the physical fixture and look at behavioral usage. Adaptive (or "smart") controls were intended to be that change. Carefully controlling the timing, intensity and color of lighting could be as much of a game-changer as LED itself. Yet few end-users implemented adaptive controls, leaving energy savings on the table.
​
AI is everywhere right now. For better or worse, AI has some place in technology of the future. It is now paired with adaptive controls to optimize light use. This month we examine whether this phenomenon may help increase the uptake of adaptive controls.

A victim of its own success

LED lighting seemed to hold great promise when its commercial rollout began in the 2000s. In a time of increasing worry about carbon emissions and climate change, LED arrived at just the right moment. It was up to ten times more energy efficient than earlier technologies like incandescent filament lamps. Government agencies like the U.S. Department of Energy (DoE)  subsidized the transition to LED. After a few years, the cost of LED products plummeted.

But promoting energy-efficient lighting had an unintended consequence: it caused a "rebound effect". That is, as a technology becomes more efficient, it becomes cheaper to use, which actually drives up total consumption. [1] This was further fueled by public policy shifts mandating more efficient lighting. Decision makers rushed to install LED, replacing certain lighting systems such as street lighting.

The result fueled an increasing tendency toward over-illumination. The end users no longer felt the financial pinch of leaving lights on all night at full power. Cities directed cost savings into installing new lighting where none existed before. Earth's cities got brighter at night during the 2010s, accelerating the loss of the stars. [2] In the end, LED may have had none of the expected environmental benefits that its proponents claimed. [3]

The bottleneck: why adaptive controls have failed to gain traction

By 2020, many lighting manufacturers offered hardware controls for new LED lighting. These controls change the state of lighting according to user needs. They range in nature from simple mechanical timers to complex, programmable schedulers. Their benefits seem good in principle. They help optimize efficiency of what amounts to a kind of industrial process. From a dark-sky perspective, this seemed ideal. After maxing out benefits of design changes (e.g., shielding), dimming and turning lights off became the new frontier.

The trouble is that most end-users didn't choose these controls. They were usually offered as add-on options and not integrated into the products. The main reason users opted against them was their added cost. Their low perceived value and high perceived friction further slowed adoption. Buyers tended to view the controls as unnecessary given the big jump in energy efficiency LED represented.
​
Adaptive controls have something of a PR problem. The DoE says the main barriers to adopting controls are unfamiliarity with the technology, interoperability issues, and the perceived "fussiness" of programming them. [4] What they want is a tool that lets them turn the lights on and walk away. It has to be understandable to people who know little about lighting tech. They also want something more flexible than a binary on/off switch. 

How AI may remove the friction

The way around this obstacle is to take the burden of management off end users. In that sense, Artificial Intelligence (AI) may be the ultimate "set it and forget it" solution. If human hesitation and the annoyance of programming are the bottlenecks, AI is the frictionless solution. It bridges the gap between the need for adaptive lighting and the reluctance to manage it. Manual programming gives way to systems that learn traffic patterns, adjust to weather and predict usage. In this way, AI makes reducing light pollution easy.

Modern, AI-driven "smart" lighting systems use machine learning to process real-time data on their own. They can predict traffic flow, detect pedestrian movement, and even adjust light output according to weather conditions. The systems program themselves by sensing use patterns and applying the information to predict future conditions. As the circumstances change, they adapt. In that sense, "adaptive" controls come full circle and achieve their greatest potential.

If adaptive lighting is autonomous and easy to manage, AI could finally incentivize the widespread adoption of smart controls. This in turn could begin to slow the tendency to "over-deploy" LED lighting around the world. It also addresses the prevalence of full-night lighting in situations where it's unnecessary. [5]

Almost all efforts to reduce or extinguish unnecessary light encounter some resistance. AI-based controls may ease some of this concern. Through the use of dimmers, light remains available throughout the night on the basis of need. This is a more sophisticated approach than the use of motion-sensing switches alone. These tend to trigger at the wrong times and leave lighting on far longer than needed. AI systems can learn more about their environments. They adjust light levels, areas of illumination and other lighting parameters in subtle ways. We know that much of the utility of lighting for security purposes has to do with subjective "feelings of safety". AI could help improve users' perception of outdoor spaces as 'safe'.
​
Some early experiments are proving hopeful. The European "SMARTLIGHT" project is  developing predictive, AI-powered adaptive street lighting. It promises 60–80% energy savings and big reductions in light pollution without manual human oversight. In Greece, Tvilight (a smart lighting controls company) installed an AI-powered adaptive lighting system across a major highway. It predicts traffic, expected weather, and road events to automatically brighten or dim over 9,000 streetlights. It does this without compromising driver safety. 

A smarter path to darker skies

Industry-standard, fully shielded luminaires were a big win for the dark-sky movement. But hardware alone can't slow humanity's insatiable appetite for artificial light at night. The next great leap in light pollution reduction will come from intelligent control, not just physical fixture design.

The greatest influence AI may yield on the future of lighting is not mere energy savings. Rather, it's about removing the human aversion to engaging with complicated lighting controls. Systems can now predict needs and adjust supply of a public good. End users get their "set it and forget it" solution, and the nighttime environment gets a break. In this sense the solution may be a true win-win scenario.

It may furthermore realize the true cost savings proponents pushed two decades ago. If the cost of AI controls falls, they could finally bring the reduced carbon emission once expected. Decreasing skyglow would then be a fortunate side-effect. For as many problems that technology has solved, it tends to create others. The arrival of AI-controlled lighting systems may mean that the pendulum now swings in the other direction.
​
Planning experts dream of cities of the future. They imagine light following human presence with ease instead of blanketing empty streets. Light glides through the urban fabric, creating inviting outdoor spaces. Instead of broadcasting light with reckless abandon, systems guide it with surgical precision. And as urban wildlife returns, there are a few more stars overhead. It brings promise rather than peril, and everyone can sleep a little better at night. 

Sources

1. Schulte-Römer, N., et al. (2019). The LED Paradox: How Light Pollution Challenges Experts to Reconsider Sustainable Lighting. Sustainability, 11(21), 6160. https://doi.org/10.3390/su11216160

2. Kyba, C. C. M., et al. (2023). Citizen scientists report global rapid reductions in the visibility of stars from 2011 to 2022. Science, 379(6629), 265–268. https://doi.org/10.1126/science.abq7781

3. Kyba, C. C. M., et al. (2017). Artificially lit surface of Earth at night increasing in radiance and extent. Science Advances, 3(11). https://doi.org/10.1126/sciadv.1701528

4. Poplawski, M. (2014, September). Emerging technology primer: Networked outdoor lighting control systems. U.S. Department of Energy, Office of Energy Efficiency & Renewable Energy. https://www.energy.gov/sites/prod/files/2015/09/f26/ssl_outdoor-lighting-control-tech-primer_0.pdf

​5. Asif, M., et al. (2022). Adaptive Control of Streetlights Using Deep Learning for the Optimization of Energy Consumption during Late Hours. Energies, 15(17), 6337. https://doi.org/10.3390/en15176337