In recent years, digital technology has been widely used in various fields, and the concept of environmental art design has naturally changed, moving closer to digital technology. Genetic algorithm is a way to find approximate optimal solutions by simulating natural evolutionary processes and can solve various problems in the environmental art design process. Based on genetic algorithm, this article proposes a digital environmental art design based on DNA genetic theory. Unlike traditional design methods that generally only generate one design scheme, genetic algorithms can generate large-scale schemes and calculate the optimal scheme among many schemes. Through three interrelated experiments ranging from simple to complex, the results demonstrate the feasibility of the model proposed in this article. The construction elements and important links of environmental art design integrated with digital technology provide a good reference for the research on digital environmental art design based on DNA genetic theory.

1. Introduction

In the process of natural evolution in the biological world, the genes of living organisms determine the characteristics of living things, and through inheritance and variation, they continue to evolve in the process of natural selection [1]. The genetic algorithm (GA) is an algorithm that simulates the genetic laws of natural evolution [2].

The theoretical prototype of the algorithm originally originated when Alan Turing proposed the mechanism of learning the laws of natural evolution in 1950 [3]. Bledsoe later proposed in 1961 the concept of “natural selection” in biology to apply to scientific systems analysis. John Holland is considered the “father of genetic algorithms,” and in 1975 he enriched the systematic concepts and methods of genetic algorithms, which were widely promoted. Genetic algorithms have been used as an effective problem-solving technique that can be applied to derive optimal solutions for weighing parties under multiple conflicting conditions, thus proposing a new way of solving problems [4]. In recent years, genetic algorithms have been widely used in mechanical engineering, economic policy, and other fields, mainly to solve the optimization problem in multi-object situations. However, the algorithm is still in the experimental stage in the field of architectural design and urban planning and has not been widely used [5]. Today, a variety of innovative technologies based on CAD models have emerged in the field of design, which fills the gap in the integration of biology, computer technology, and architectural design. One of the most prominent tools that can be applied to architectural design is Octopus 3D, a genetic algorithm tool based on the grasshopper visual programming platform [6]. In order to better apply the genetic algorithm to the field of architectural design, a deep understanding of the logical idea of the algorithm is required [7]. Therefore, this study uses Octopus as an experimental tool to try the possibility of practical application of genetic algorithms in architecture and design.

Our daily lives are based on mobile phones and computers, which are indispensable digital technologies[8]. Specifically speaking, digital technology is the technical foundation of multimedia technology, software and hardware technology, AI intelligent technology, and even the entire information society, which includes multiple subitems such as digital, text, image, speech, programming, etc., and cutting-edge technologies such as virtual reality (VR), augmented reality (AR) and mixed reality (MR) are also included in digital technology, and the application is very common. In the field of science and technology economy in the current society, people have noticed that virtual reality and augmented reality have been quietly popularized. Alipay’s AR payment, the chain home APP’s real-life viewing, Apple’s ARkit, and Microsoft’s HoloLens VR device are also quietly making their way into the homes of ordinary people and the shopping space [ [9, 10]]. Now in some shopping malls, we can see more and more VR game centers and AR experience blocks stationed, but when we experience it, we will find that their relatively scattered layout, relatively fragmented system, and relatively solidified interaction will make the experiencer tired after the first experience. The stationing of this kind of equipment is mandatory and mechanical, and its essence only makes consumers feel its novelty and does not have an organic connection with the overall environment of the shopping center and the goods sold in the internal stores, which is a rigid set that does not have sufficient interactive experience characteristics. In this article, we propose a digital environment art design method based on DNA genetic theory. First, a large-scale design scheme is generated, and then the optimal design scheme is obtained by considering multiple design schemes. The feasibility of the proposed model is demonstrated through related experiments. The optimal design scheme in multiple design schemes is also calculated. This topic also provides a good reference for the art design research of digital environment based on DNA genetic theory.

2. State of the Art

2.1. Background of the Establishment of Genetic Theory

In 1802, the French naturalist, J.B Lamarck published a paper on the observation of living natural bodies, in which he expounded from a zoological point of view the evolutionary process from simple to complex. In 1859, the British biologist, C. Harles Darwin published his biological masterpiece The Origin of Species, which gave the laws of biological evolution [11].

From 1857 to 1865, the Austrian priest G. Mendel (1822–1884), from a poor peasant family, conducted an 8-year experiment on pea hybridization in the monastery garden. He crossed peas of different colors, tracked the distribution of these traits in the offspring, and performed a careful statistical analysis. Mendel showed through large-scale experiments that germ cells are carriers of genetic traits. Mendel thought since germ cells are carriers of genetic traits, which part of germ cells is the carrier of genetic traits? Since the detailed structure of the cell had not yet been understood, he speculated that the trait was carried by a scattered unit (genetic factor) inside the germ cell and transmitted over generations. He provided a great deal of experimental evidence for the existence of such a genetic factor, but little was discussed about the factor itself. This genetic factor was still an “invisible, untouchable” hypothesis in Mendel.

In 1866, Mendel published his paper “Experiments in Plant Hybridization” in a local magazine in Austria, expounding his theory of biological genetics, but it did not attract the attention of his contemporaries. It was not until 1900 that the paper was rediscovered by three scientists, H. de Vries (1848–1935) in the Netherlands, C. Correns (1864–1933) in Germany, and S.E. von Tscher-mak (1871–1962) in Austria, and Mendel’s genetic theory became a starting point for biological development in the 20th century. Genetic algorithms can find approximate optimal solutions by simulating the natural evolution process, which can solve various problems such as contradictory design objectives in the design process of environmental art and provide a new solution for environmental art design.

2.2. Background of Digital Environmental Art

In the development process of ancient Chinese environmental art, environmental art design in the pre-Qin, Sui and Tang dynasties and Ming and Qing dynasties is the most representative. In the pre-Qin period, with the emergence of private ownership, unprecedented changes occurred in the social economy, and environmental art design also reflected its unique characteristics, and in the same period, there were also forms of settlement environmental art, urban environmental art, and architectural environmental art. The cultural characteristics of the Sui and Tang dynasties injected vitality into environmental art, forming a distinct era characteristic of simplicity, authenticity, majesty, and boldness, which also had a profound impact on East Asia, including Japan and the Korean Peninsula. In terms of urban environmental art, the unprecedented unity and strength of scale, and the unprecedented tolerance and intake of style are concentrated in the urban environmental art of the imperial capital Chang’an. The Ming and Qing dynasties were the third climax of environmental art after the pre-Qin and Sui and Tang dynasties, and it was also a summary of ancient Chinese environmental art. In terms of urban environmental art, the most representative work is the capital city of Beijing. It is worth mentioning that during the Ming and Qing dynasties, the art of the mausoleum environment sprang up and achieved world-renowned achievements. The Ming Tombs adopted the arrangement method of grouping tombs into groups, breaking through the tradition that the mausoleums of each dynasty were built separately in the past, set up their own Shinto, and were not related to each other and adopted the method of sharing Shinto with the Chengzu Long Tomb as the center and other mausoleum rings forming an arc, which not only reduced the labor, but also made the momentum of the mausoleum area more grand, which is a major creation worthy of attention [12].

Among the development processes of foreign environmental art and design, the most representative periods are the ancient Egyptian period and the Italian Renaissance. The ancient Egyptian environmental art known from the ruins is mainly reflected in their mausoleums and temple complexes built of stone materials. In the dry and hot climate conditions, the ancient Egyptians were good at using trees and water bodies to create a cool and humid environment, reflecting the early civilization’s pursuit of a garden environment. During the Italian Renaissance, as well as the wide application of Euclidean geometry and three-dimensional perspective in architecture and environmental design, Alberti, Blamonte, Pallamento, Michelangelo, and others carried out a large number of single buildings, city squares, and ideal urban designs based on round, square, and golden rectangles, resulting in geometric integrity and a clear, strong sense of concentration of shape and spatial environment composition, reflecting the rational spirit of human places, and had a wide impact in Europe. From the mid-15th century, the Italian Renaissance culture gradually spread to other countries, forming a ShangGu tendency in European architectural and environmental art in the following centuries. In the process of environmental art design, many contradictory design objectives are often encountered, and the traditional design methods usually can only generate one design scheme, not with diversity as shown in Figure 1.

2.3. Background of Digital Technology

Digital technology is also known as digital control technology. In the digital age, human interaction is media on the Internet. People study, live, and work more using the Internet, home appliances will be organized into a home network by computers to manage, and people can be anywhere and at any time with any device to get the information they need.

With the continuous impact of Internet business, experiential shopping malls continue to compromise with the market in order to achieve a good survival situation in the absence of a secondary innovation-driven environment [13]. Affected by the sudden outbreak of the new crown pneumonia epidemic in 2020, the retail industry is facing the potential risk of reshuffling, which has also made many retailers and designers think deeply about the trend of the next physical business, hoping to open up the future of commercial new retail through innovative measures [14].

As early as 2017, Chen Yunfeng, secretary-general of the China Real Estate Managers Alliance, said in a commercial real estate event that the golden ratio of shopping, entertainment, and catering in shopping malls has been broken, and the previous model has been replaced by new models such as “1:1:1” for shopping, leisure, and catering. The proportion of the entire catering industry has soared to about 40% in some developed cities in 2019. In addition, in the shopping space, children and education formats have also sprung up to 15% of the proportion, and the two have gradually swallowed up the main shopping formats of shopping malls, and shopping malls have no longer mainly assumed the function of buying and selling and gradually lost the original intention of their interactive experience, but have generally become “big canteens” and “small classrooms” (Figure 2).

Figure 2 illustrates the interrelationship of these digital technologies; the left side of the relationship differentiation diagram is the real world and the right side is the virtual world [15]. The real world is the real three-dimensional space in which we live. Augmented reality (AR), as the name suggests, is an augmented representation of the real world, using related equipment to scan and match the specific physical objects of the real world and through image recognition and other technologies, presenting part of the content of the virtual world in the real world, so as to give the physical object more information and achieve an enhanced role. On the one hand, virtual reality (VR) is a collection of computer graphics, human-machine interface, simulation technology, sensing technology, and network transmission technology, and this collection can create a multisource fusion, three-dimensional interactive dynamic simulation vision through the corresponding equipment. On the other hand, mixed reality (MR) technology strives to create a trend effect close to the virtual world through the fusion and sublimation of multiple technologies [16].

AR and VR technology devices developed and mass-produced today play a pivotal role in today’s technology community. At the level of creating an interactive experience in space, the difference between the two, in addition to existing in hardware and technology, is ultimately to see which of the two technologies can have a complete virtualization effect on the enhancement of reality [17]. In terms of immersive virtual experiences, VR will win because it immerses the user fully, but it will fail if it aims to make virtual things more like real life. Although the ultimate goal of MR is to completely integrate the above two and interact with virtual and real time in “real time,” this ultimate form of development is still in the stage of theoretical research and laboratory development, and it is extremely difficult to achieve and popularize in the short term [18]. Therefore, under the corresponding technical conditions, AR technology that seamlessly connects the real world with virtual information, thereby enhancing the human-computer interaction of real-world scenes, is far more comprehensive and applicable than VR in many senses, and it is more appropriate to become the next generation of universal human-machine terminals [19].

3. Methodology

3.1. Characteristics of Environmental Art Design in the Digital Age

When human beings gradually move from the postindustrial era to the digital age, the concept of environmental art design has naturally changed, which is mainly manifested in the following two aspects: First, the transformation from the contradiction of the environment to the coordination of the environment. In the digital environment, the transformation of environmental art design concepts to environmental coordination has become inevitable. The second is to change from “form following function” to “form following emotion.” The Internet and virtual communities do not make the relationship between people closer, but rather strengthen the solitude and personalization of the individual’s way of life, so that interior design also carries the heavy responsibility of human spiritual and spiritual comfort [20].

3.1.1. Visualization of Design Ideas

Design ideation is a critical step for designers after obtaining a large amount of first-hand data. In the digital environment, due to the development of computer hardware and software, design sketches have been computerized. In the past, these drawings could only be drawn with bare hands. Therefore, the entire design performance is not only cumbersome and inconvenient to modify, but also has an excessively long cycle and incomplete expression. In the digital environment, the design performance has undergone tremendous changes. First, due to the development of computer technology, the hand-drawn in the past can be completely drawn by the two-dimensional software AutoCAD. With the three-dimensional software 3D StudioMAX, you can first build a three-dimensional model in the computer and use the functions of the camera in the software to make a multi-angle, all-round observation, and finally render the rendering. Therefore, the design performance presents the advantages of simplicity, speed, easy modification, intuition, and complete. Second, with the maturity of network 3D technology (network virtual reality), designers can use Cult3D, Pulse3D, Ser, 3DML, and other network 3D software to carry out 3D modeling, so that owners and related personnel can multi-angle the design scheme through the network. Comprehensive observation and evaluation could be set.

3.1.2. Improve Design Accuracy

The wide application of computer technology in the design discipline has brought about changes in design methods and concepts, become a new trend in design and creation, and also brought revolutionary changes to the development of environmental art design. The superiority of computer-aided design is manifested in the following four aspects: one is to improve the accuracy of the design, reduce errors, and complete complex designs that cannot be completed by manual means alone. The second is to improve productivity. It can make people get rid of tedious and repetitive drawing work, save the time of consulting information, and greatly accelerate the design speed. It shortens the design cycle, so that new advanced design ideas can be quickly turned into works. The third is to facilitate the realization of optimal design. Due to the fast design speed, designers have the potential to select the best design from many design solutions to achieve better economic benefits. Finally, the fourth is to reduce design costs. It can save more manpower and material resources and change in time according to the requirements and intentions to meet the special needs of all aspects.

3.1.3. Intelligent, Reduce Costs, and Save Costs

Computer-aided design will evolve mainly towards the following aspects. The first is to be intelligent, expand functions, and make the drawing system have richer colors and levels. The second is to develop high-efficiency special hard equipment, such as drawing software hardening, CAD/CAM database machine, etc., expand the scope of application, in addition to environmental art design, and further develop a broader field of application, such as packaging design, clothing design, and even the design of household equipment. The third is standardization, so as to facilitate the sharing of results and mutual exchanges, and reduce costs. There have been many achievements in the standardization of programming languages, numerical control languages, and in-computer representation of graphics, but most of them are application packages for specific environments, and there is still a lack of unified standards. However, like all high-tech, the use of computer technology in the design discipline also has many negative effects.

There are great differences between hand-painted environmental art design methods and digital environmental art design methods. In the contest between the two, for now, digital methods have the upper hand. But is it possible to directly say that digital environmental art design methods are superior to hand-drawn environmental art design methods? It is obviously unscientific to draw conclusions in this way. The two have their own advantages and disadvantages, although hand-drawn is more cumbersome, but it can be closer to the designer’s idea and more flexible to deal with some details. Digital environmental art design depends on certain technical conditions, and when designers are inspired but cannot find the help of modern tools such as computers, they cannot immediately play the advantages of digital design methods. Only by combining the two and using different techniques according to the actual situation can we achieve the highest state of environmental art.

3.2. Theories Related to Genetic Inheritance Algorithms

The genetic algorithm (GA) was first proposed in 1975 by Professor Holland of Michigan University in his monograph “Applicability of Natural and Artificial Systems.” Genetic algorithm, also known as an evolutionary algorithm, is a heuristic search algorithm inspired by Darwin’s theory of evolution and drawn on the process of biological evolution. Drawing on the theory of biological evolution, genetic algorithms simulate the problems to be solved into a biological evolutionary process; generate the next generation of solutions through replication, crossover, mutation, and other operations; and gradually eliminate the solutions with low adaptability function values and increase the solutions with high adaptability function values. In this way, after the evolution of N generations, it is very likely that individuals with high adaptability function values will evolve and be applied to various fields.

3.3. Design of GA-CNN Algorithm

The test results of the traditional and basic CNN methods in the experimental environment are shown in Table 1.

From the experiment, for the traditional classification algorithm, SVM performs high. In the data case, the accuracy of Naive Bayes algorithm is low, followed by logistic regression and linear SVM algorithm.

In the CNN network structure, there are many parameters and structures that can be discussed. In the algorithmic exploration of GA-CNN, each layer of network structure is treated as a chromosome. The system architecture of the GA-CNN algorithm is shown in Figure 3; its overall process is shown in Algorithm 1.

Step 1 Process the data in a standardized manner and divide into training sets, evaluation sets, and test sets.
Step 2 Initialize the population of the CNN framework structure and preset the maximum number of iterations G and the current population algebra  = 1.
Step 3 Learn to train each frame structure in the CNN population.
Step 4 Evaluate the trained CNN model with the evaluation set to obtain the corresponding adaptability of the CNN framework structure population.
Step 5 Use the roulette method to generate mating targets.
Step 6 Crossover the mating target and perform training to assess fitness.
Step 7 Use the variation operation to mutate the crossover results and perform training to assess fitness.
Step 8 Determine whether the newly generated results are better than the mating target, update the CNN structure population, and update the corresponding adaptability.
Step 9 If  < G and the convergence condition is not met,  =  + 1, go to step 5, otherwise go to step 10.
Step 10 Output the elite individual model as the final classification model.

The above pairs of GA-CNN algorithm and the results after the traditional CNN test are shown in Table 2.

In conclusion, it can be seen that the GA-CNN algorithm, which has evolved, effectively self-adjusted, adjusted its own structure and model parameters, and improved the model accuracy, rising from 52.68% to 77.08%. The evolution reaches convergence at 85 times, obtaining an approximate optimal solution.

3.4. Effect Analysis between Genotype and Environment

A mode H model contains four parameters: original length l, degree O (H), define length and pattern dimension D (H), l is the length of the string, O (H) is the number of fixed bits in the pattern, (H) is the distance between the first and last fixed bits, for example, O ( 111 ) = 3, ( 111 ) = 4 − 2 = 2, D (H) represents the number of strings in the pattern, is

Assuming k-generation inheritance, there are m specific pattern H in population p (k), written m (H, k)

Since the average fitness of the population is

Suppose that the fitness of the H is

If starting from k-0, c is constant

The above analysis shows that a qualitative analysis of global convergence based on the pattern theorem holds that the genetic algorithm is globally convergent. Based on the actual settings of many researchers, the genetic algorithm parameter set in this article is shown in Table 3.

4. Result Analysis and Discussion

In this study, the experiment consisted of three interrelated sequences, the subject of which was a three-dimensional model built in rhino and grasshopper modeling software, and the operating software of the gene algorithm was Octopus. The complexity of these three sequences increases from low to high, and the study subjects develop from the three-dimensional form of a single body to a set of urban textures. In the three sequences of experiments, the concept of “Body Plan” is introduced, that is the subjects are specifically divided into parts of the body that can be controlled by genes. In addition, the parameters of the control variables that function as “genes” are also set, and these variables and parameters together form the “Gene Pool” that controls the development of test subjects. In addition, the experiment will also set a number of conflicting “Fitness Criteria,” which is equivalent to playing the role of natural selection in the process of biological evolution, which can be used to judge the specific performance level of the test subjects, and finally screen out the optimal scheme of environmental art design.

4.1. Experimental Process

Each sample consists of a random string of 6 genes that controls its performance in genetic operations. In addition, the “standard of excellence” for this sequence test is set to two conflicting criteria, the contradictory design objectives of the traditional design method are simulated to investigate the feasibility of this method: (1) the maximum projected area is obtained with the smallest volume, and (2) the structure has the highest total height. After setting the number of samples per generation to 10 and performing 4 generations of cross-genetic operations without variation, a group of 10 samples was generated, which was counted and arranged for comparison with subsequent experimental samples. Three individuals were randomly selected from the fourth to fifth generations of genetic operations and a variant gene B was added, that is, the three individuals had a gene string length of 7. Two more individuals were randomly selected and added to the genetic operations from the fifth to sixth generations, H and C, which are 8 in length.

Comparing three groups of individual samples of the fourth, fifth, and sixth generations, the test found that the difference in the generated sample groups became significantly larger after the addition of mutated genes. It is worth noting that from the fourth to the sixth generation, the average “standard of excellence” values of the sample groups showed a clear upward trend (Figures 4 and 5). This is mainly because the variability of the sample groups is affected by the mutated genes.

Each body part has a corresponding genetic control over the variable range of that part in the genetic algorithm. The gene pool of this group is composed of a total of 8 genomes, which control the variable range of height, depth, and retreat of the inner and outer group buildings (interrelated with the various parts of the building, the building area and location of the tower, retail store, and courtyard will also be affected by these genes), and the other part of the gene controls the height of the tower house and the depth of the retail store facing the courtyard.

In addition, three conflicting “excellent standards” are set for the model: first, the smallest building volume; second, the largest courtyard sunshine; and third, the largest building sunshine. The third generation was calculated by the first operation of Octopus. During the operation, the number of samples per generation was 10, the elite screening rate was 0.5, the probability of variation was 0.1, the mutation ratio was 0.5, and the cross-reproduction rate was 0.8. The 10 samples of the third generation generated by the operation were optimized for performance across three health criteria and achieved convergence. Under the same setting, the operation continues to the sixth generation, and although the 10 samples are still optimized, the convergence of individual samples fluctuates and shows more diversity. Therefore, it was found in the experiment that maintaining the same settings and continuing to calculate did not guarantee that the samples would remain convergent and develop. At this point, the variation scale parameter in the setting is raised from 0.5 to 0.7, and the other probability parameters remain unchanged, and the operation continues to the tenth generation.

The sample results show that the 10 samples of the tenth generation have developed convergence in volume, and the convergence of these two aspects of architectural sunshine and courtyard sunshine is still weak. Comparing the three groups of samples of the third, sixth, and tenth generations, the results show that the best-performing sample group under the three excellent standards is the tenth generation, and the best-performing individual is the sixth generation G6.1.

4.2. Experimental Results and Discussion

When discussing and analyzing the experimental results, the four parameters of variation probability, variation proportion, cross-reproduction rate parameter, and sample diversity are mainly taken as the main indicators. The results of the sequence 1 experiment show that although the sample as a whole shows an optimization trend with the iterative calculation of the genetic algorithm, the optimal individual does not necessarily appear in the latest generation of sample results. The results of the sequence 2 test showed that the elite screening rate in the probability parameters was inversely proportional to the sample diversity. The results of the sequence 3 test showed that the probability of variation, the proportion of variation, and the parameters of cross-reproduction rate were proportional to the sample diversity. It is worth noting that the three best options for sequence 3 are G80.11, G80.19, and G80.16 and the initial model scheme. It can be found that the newly generated scheme has been significantly improved by the evaluation of the three “standards of excellence” initially established by the trial. Looking at these three schemes from the perspective of the floor plan, the spatial variation is more abundant than that of the initial scheme. In addition, for these three optimization schemes, in addition to evaluating the performance of the three “excellent standards” involved in the genetic algorithm process, other common urban spatial evaluation criteria have also been measured, including plot coverage, volume ratio (FAR), and building façade sunshine. The performance ranking of G80.11, G80.19, and G80.16 in terms of building coverage, plot ratio, and building façade insolation is inconsistent with the “excellent standard” in the original genetic algorithm. Therefore, when selecting the most suitable solution among several optimization schemes derived from the genetic algorithm, the performance of these schemes in other aspects can be referred to make a choice.

5. Conclusion

When traditional design methods encounter multiple conflicting design goals, they lack logical and rigorous solutions, and how to find a balance between multiple conflicting goals has become a design difficulty. In the face of this situation, the gene algorithm provides a good idea, by learning the principle of “natural selection” in evolutionary theory, iteratively producing an unlimited number of design schemes under the control conditions of set parameters, and by quantifying the numerical value of each scheme in various aspects, to measure and compare the optimal solution in the generated samples. Through the experiment of the antecedents of three sequences, this study explores the possibilities and specific methods of gene genetic algorithms based on the concepts of biological evolution and embryonic development theory (Evo-Devo). The subjects of the three experiments evolved from the simplest cube to the urban blocks with a certain degree of complexity, and the effect of different parameter changes on the results of the experimental sample group was tested. Experimental results show that the genetic algorithm can be used as a scheme optimization tool to solve design problems, produce large-scale possibility schemes, provide a good research direction for digital environment art design, and select the best scheme. However, although genetic algorithms are a very powerful tool that can optimize schemes indefinitely, how to use genetic algorithms correctly is the key to the problem. In the three sequences of experiments, it can be found that the most influential on the results of the experiment is the input, the input parameters in the experiment include: the model and its various “body parts” cut and the parameters that control the change of the model. The final results show that the final output direction of the scheme generated by genetic algorithm depends on the initial input conditions to a large extent, so genetic algorithm can be optimized when the input conditions are determined. Future genetic algorithms can be considered for optimizing the output direction.

Data Availability

The labeled data set used to support the findings of this study is available from the corresponding author upon request.

Conflicts of Interest

The author declares that there are no conflicts of interest.


This work was supported by the Jingdezhen Institute.