Genetic Algorithms: Optimizing Lighting and Architectural Design
Genetic algorithms were invented by John Holland in the 1960s, and since then, they have been used as stochastic methods for solving optimization and search problems, operating on a population of possible solutions. According to Darwin’s theory of evolution, the repetitive application of the aforementioned procedures alters an initial species into various other species; however, only the stronger prevail. A genetic algorithm, which is a subset of evolutionary computation, can be defined as a “population-based metaheuristic optimization algorithm.” These nature-inspired algorithms evolve populations of experimental solutions through numerous generations by using the basic principles of evolutionary biology such as reproduction, mutation, recombination, and selection.
Genetic algorithms perform the same operations on the population of possible targets, with only those that better fit the solution surviving. Like other computational systems inspired by natural systems, genetic algorithms have been used in two ways: as techniques to solve technological problems and as simplified scientific models that can answer questions about nature. These algorithms are widely used because they yield accurate results and have fast processing times in many applications such as image processing, optimization of lighting design, home automation, electronic circuit designing, wireless sensors, training artificial neural networks, automated designs of room, multi-criteria optimization problems, automation of street lights, traffic control systems and pedestrian crossing, security systems, clustering, computer automated designs, filtering and signal processing, power electronic designs, optimization of electrical distribution network, scheduling applications and many more... These nature-inspired algorithms can help engineers find solutions beyond imagination.
In the trend of semiconductor illumination technology development and application, lighting plays a unique role in many aspects of sustainable development. For indoor lighting design, the first thing to consider is the amount of luminaires. The main requirement for the amount of luminaires is the “right illuminance” (illuminance is an indirect indicator of the brightness of an object). An excessive number of luminaires will lead to energy waste and increase lighting power density and costs. Lighting power density is an indicator meant to measure the energy efficiency of a lighting system. Second, indoor lighting design could improve lighting quality, and that directly affects the efficiency of work, physical health, psychological conditions and even the atmosphere and various effects in an indoor room. However, the diversity of indoor environments and the mutual exclusion of lighting design parameters, including lighting power density, average illuminance, overall uniformity, and maximum unified glare rating, pose difficulties for indoor luminaire layout. The main emphasis should be on designing an energy-efficient luminaire layout method for a general lighting scheme, which maximizes or minimizes these design parameters within recommended limits.
Lighting design is extremely important due to the high cost of implementation and maintenance of the illumination installations. Proper illumination dimensioning is indispensable to achieve adequate lighting and reduced electrical power consumption. CAD tools are very important to the engineer design process to enable alternative simulations before the final solution, improving quality and maturity in engineering projects. In the electrical installation area, these tools have been proven suitable in several design activities. However, recent studies show that the computational facilities are still dependent on the creativity and experience of the engineer or lighting designer, which hinders a better performance from the design to the market.
Lighting design is a field of engineering that misses automatic approaches to help lighting designers. Genetic algorithms are a widely used class of algorithm for search and optimization. Many results from applying genetic algorithms for computer-generated lighting design show that the solution designed by genetic algorithms, when compared to the edified one, provides a superior uniformity of illumination and no less than a 10% reduction in power consumption. A process to be optimized is characterized by an objective function that provides the behavior of the process, the constraints that define the search space and on which the project variables tend to assume the best value after the optimization. Many of these processes can be modeled as problems of maximizing or minimizing a function whose variables must obey certain constraints. Optimizing a process is advantageous since it allows working with a vast contingent of variables and constraints that are often difficult to visualize or tabulate, thus reducing the time spent with the process and obtaining new solutions with lower expenses.
Genetic algorithms, also known as evolutionary algorithms, as an aspect of the computational thinking, are capable of generating innovative designs and objects with complex geometries in architecture that adapt to the demands of the social and environmental environment. Work with genetic algorithms began in the 1960s at the hands of the English architect John Frazer, although his ideas found greater diffusion years later through his work, "An Evolutionary Architecture" (1995). Frazer establishes a relationship between evolutionary processes and architecture. His investigation of the fundamental processes behind the generation of forms leads him to consider architecture as an artificial life form, and to propose a genetic representation that can then be subject to processes of development and evolution in response to the user and the environment. The goal of an evolutionary architecture, Frazer suggests, is to achieve in a built environment the symbiotic behavior and metabolic balance that are characteristic of the natural environment. But, with this, he does not intend to establish a superficial analogy, but rather the creation of a work methodology that allows the development of structures following the design principles underlying nature.
Evolutionary architecture explores the possibilities of buildings to improve their performance according to the demands of their inhabitants and the natural environment. These demands are dynamic, and continually change over time, so the architecture must also adapt and change accordingly. Evolutionary algorithms offer an effective solution to this problem: when architecture needs to re-adapt to the environment, the algorithms are capable of searching numerous adaptation possibilities and finding the most suitable solution, providing methods for structures to behave efficiently, and addressing well-defined construction factors such as structural, mechanical, thermal and lighting performance.
Additionally, this type of application allows routine processes in architecture and engineering to be streamlined and automated. Imagine automatically evaluating multiple design alternatives for the implementation of a project on a given lot. For this type of exercise, the determinants of the lot, both physical and normative, are established, and design parameters are established that direct the evolutionary verification of multiple alternatives that meet the success indicators of the project. Thus, the decision on the best option has a repertoire that broadly informs reality and supports the choice.
In 2020, Foster + Partners won a competition to design the new offices for Alibaba, the Chinese e-commerce giant, in Shanghai. Its construction will be guided by a design process based on evolutionary algorithms. The algorithm combines several aspects that are crucial to the project, such as being highly sensitive to environmental cconditions and maximizing exterior views, while meeting specific area requirements for different building functions. Another challenge to be solved is to achieve better user comfort in the central public space, protecting it from strong winter winds and harsh summer sun, while creating bespoke workspace solutions for the different departments of Alibaba.
The application of evolutionary architecture falls under the realm of computational design, a revolutionary way of thinking about architecture that is associated with digital arts and engineering. Computational design incorporates algorithms, mathematics, and generative thinking to create innovative designs and objects with complex geometries. Through the connection between technology, society and spaces, design by data allows us to gather invisible data from the environment to integrate into structures, which gives us the possibility of creating a "living architecture" based on the environment, the context and the culture. In this framework, genetic algorithms represent one of the most interesting and disruptive aspects of computational thinking, not only because of its ability to generate different solutions to structural problems, but also because of its potential to generate new business opportunities focused on sustainability and the relationship with a changing social environment.
So what does the future hold for genetic algorithms, and how far can they take us? In his book "The Light of Evolution in the Darkness of the Universe," Dr. Çağrı Mert Bakırcı-Taylor, Ph.D. at Texas Tech University, claims that machines utilizing evolutionary algorithms will be the beginning of the technological singularity. This will not happen with some engineer somewhere coming up with a single equation, which may or may not be complex, and an artificial general intelligence emerging just like that. This is because our machine learning methods rely on extremely simplistic models of neurons, and learning based on this simplicity cannot achieve the complexity of human brains, just like overly simplified engineering designs are bound to perform worse than their evolutionarily designed counterparts, he argues.
If Bakırcı-Taylor is right, and if we can overcome some of the hurdles, in a couple of decades evolutionary algorithms should be the ‘next big thing.’ If he is wrong, we might find a brand new approach, since obviously the current deep learning methods are great at special artificial intelligence but are not enough to yield a general one. What that approach could be, of course, we don't know. We just look at the history of engineering and realize that the main pattern is following the lead of nature to come up with better techniques. And if our goal is to create the most complicated, most beautiful, most powerful technologies of all times, we cannot just go with learning—we have to substantiate it with evolution. Because that is how it has worked in nature for the past 4 billion years. We'll see.
With AI developing at its current dizzying rate, genetic algorithms and evolutionary computation might be what will drive our technological abilities beyond deep learning and enable AI to become fully 'creative.' It is undeniable that engineers generally favor more traditional design processes where they can be fully in control of everything with mathematics and certified methods of calculation. An approach where they let an algorithm do the job of figuring out what works might seem like uncharted territory; however, when achieving a level of creativity that rivals nature becomes a necessity, the right thing might be to take your hands off of the wheel and let evolution do what it does best.
When the researchers do take on the challenge, genetic algorithms can be quite fruitful, as we've mentioned before. The difficulties aside, many fields have been influenced by the advances in genetic algorithms, with control engineering being one of the first. ■