Exploration of indoor environment perception and design model based on virtual reality technology

Mingli Zhang

doi:10.1515/nleng-2024-0080

Article Open Access

Exploration of indoor environment perception and design model based on virtual reality technology

Mingli Zhang

Published/Copyright: April 29, 2025

Published by

Become an author with De Gruyter Brill

Submit Manuscript Author Information

From the journal Nonlinear Engineering Volume 14 Issue 1

Abstract

With the advancement of technology, virtual reality (VR) offers designers innovative tools and perspectives in modern interior design, enhancing the intuitiveness and efficiency of indoor environment perception and design. This study adopts distance measurement modeling technology to accurately draw the spatial structure of an indoor environment using signal strength indicators and optimize the design solution. An interior design model based on virtual reality and particle swarm optimization backpropagation algorithm is proposed. Experimental results indicated that the proposed algorithm outperformed three other methods, including support vector machines (SVM), achieving the highest detection rate with a sliding window size of 50. Although increasing the length of the sliding window reduced accuracy, it still maintained its highest value at a length of 50. When the distance was less than 5 m, the detection rate exceeded 90%. Although the accuracy decreased with increasing distance, the designed method maintained a relatively high detection rate. Under the joint adaptive strategy, the algorithm achieved a 90% positioning error of 0.82 m, significantly outperforming the other methods and greatly enhancing indoor environmental perception accuracy and efficiency. These findings provide valuable insights for optimizing related technologies.

Graphical abstract

This article combines the received signal strength indication ranging technology with the particle swarm backpropagation algorithm to propose an indoor environment perception and design model based on virtual reality. The highest detection rate of 95.5% was achieved in the static state, while a detection rate of 90% was maintained in the dynamic state, demonstrating strong robustness. At different distances, the positioning error of the PSO-BP-Adaboost algorithm remains within 0.82 m, indicating its effectiveness and accuracy in real-time indoor positioning. (a) Error changes before applying filtering algorithms. (b) Error changes after applying filtering algorithms.

Keywords: virtual reality; indoor environmental perception; receiving signal strength indication; particle swarm backpropagation; positioning accuracy

1 Introduction

With the advances in technology, virtual reality (VR) has transformed from a conceptualized sci-fi dream to an advanced technology practice that penetrates various industry fields [1]. Among them, the exploration and practice in indoor environmental perception and design have demonstrated infinite possibilities of VR [2]. In the past, interior designers relied on graphic design drawings and models for design. This approach poses difficulties in conveying and understanding design concepts [3]. However, with the application of VR, designers can design and display in a three-dimensional virtual environment, making design ideas and concepts more intuitive and understandable, which greatly improves design efficiency and accuracy [4]. VR technology has changed design methods and the way people perceive indoor environments. Through VR helmets, one can enter a virtual space that is completely consistent with the real environment, fully perceive the indoor environment, and have no blind spots. This immersive experience enables designers to better understand and grasp the characteristics of the indoor environment, realizing more precise design [5]. In this context, the exploration of indoor environment perception and design models based on VR is particularly important. The application prospects of VR are broad. However, in practical operation, how to effectively utilize VR for indoor environment perception, how to achieve the organic combination of VR and indoor design, and how to handle the problems that VR may encounter in indoor design are all key issues that need to be deeply explored [6]. Therefore, the aim is to explore the indoor environment perception and design model based on VR technology and improve the accuracy and efficiency of indoor positioning by combining the received signal strength indication (RSSI) ranging technology and particle swarm optimization-back propagation (PSO-BP) algorithm. Specifically, the research aims to help designers and users better understand and perceive interior spaces while optimizing design schemes by building an intuitive three-dimensional virtual environment. Through in-depth analysis of the performance of various algorithms in dynamic environments, as well as a detailed discussion of the location tracking process, it is expected to provide new technical support and a theoretical basis for the field of interior design, promoting the development and popularization of VR technology in practical applications.

Based on this, this article aims to explore an indoor environment perception and design model based on VR technology. Through theoretical research and practical exploration, it is expected to find an effective model that organically combines VR technology with indoor design to optimize indoor environment perception and design. At the same time, it is hoped to explore the application of VR in indoor design, providing new impetus and perspectives for interior design.

The research will be conducted in four sections. The first section overviews the indoor environment perception and design models based on VR. The second section explores the indoor environment perception and design model based on VR technology. The third section verifies the method. The fourth section summarizes the research and puts forward the future research direction.

2 Related works

VR is an interactive virtual environment that integrates vision, hearing, and touch, generated using computer technology. Designers and users can interact in this environment, providing real-time feedback and adjusting designs through various devices and software. How to improve the simulation effect of VR and bring more realistic feelings and experiences to users has become a research hotspot today. Ya studied the application of VR in art design. The basic theory of virtual prototyping was summarized, expanding the core technology of virtual prototyping. By analyzing the current situation of design technology in China, the application of VR in art design was explored. Compared with traditional art design, the application of VR in art design was more extensive and feasible [7]. Zhang et al. explored the application of point cloud computing on object surfaces based on VR. By reconstructing the geometric shape of object surfaces in VR technology, the problems in point cloud data algorithms for object surfaces were solved. It indicated that the proposed method had higher independence and flexibility [8]. Cummings et al. explored the impact of psychological characteristics, including immersion tendency, absorption, sensory seeking, cognitive needs, and fear of new things, on VR adoption rate and usage time. The research found that psychological factors were often more capable of determining VR adoption than demographic factors. On the contrary, those who believed that their emotions were more susceptible to face-to-face communication had lower VR usage [9]. Yangfei proposed a ship interior environment design system based on 3D VR. The system hardware included a visual module and a data processing module, among which the visual module was composed of multiple stereo vision sensors. The data processing module consisted of a PC and a GPU. The system software was configured as an indoor environment design module. This module mainly designed the indoor environment of ships through direct 3D software and Open GL software. By combining hardware and software, indoor environment design for ships was achieved. The indoor scene simulation of the system took the shortest time and achieved an improvement in design performance [10]. Shimeng and YIchao summarized the development and direction of VR applied in light environment simulation. The implementation of VR in the light environment simulation was introduced. The visualization of a monochromatic light environment was analyzed. The results indicated that the virtual light environment could realistically restore the real light environment in terms of light color and brightness. However, for certain color schemes, there was a significant deviation in monochromatic photoreduction. Real and virtual light environments exhibited high consistency in pleasure, arousal, and visual comfort [11].

Relying on the characteristics of VR, the virtual simulation experimental teaching based on VR can showcase highly simulated 3D virtual experimental environments and experimental objects. This is not only conducive to cultivating the spatial perception, thinking, and imagination abilities of students majoring in environmental design, but also enhances their skills and understanding in environmental perception and design. Jian and JiE elaborated on the concept and necessity of project design. The process and teaching methods of virtual simulation experiment teaching were explored. The difficulties and countermeasures of virtual simulation experiment teaching under existing technological conditions were analyzed [12]. Liu proposed a user interaction experience art design method based on situational awareness and machine learning. The user knowledge model was constructed, and artistic recommendations were made through tensor decomposition. The results indicated that this method could still obtain excellent recommendation results in sparse data. It could solve problems in folk art appreciation classes [13]. Saucedo et al. revised the history and current status of freshwater pearl farming in Asia and Latin America, particularly the pearl farming potential in southeastern Mexico. The first batch of pearls in Latin America was successfully produced, committed to optimizing production technology and promoting local community development [14].

In summary, with the continuous progress of VR, the application of VR is no longer limited to the military, medical, or film production fields, but gradually appears in the public and enters the lives of ordinary people. Users can experience unprecedented immersion and create unprecedented scenes and experiences. Exploring indoor environment perception and design models based on VR technology not only has high application value, but also has great development space in the market.

3 Exploration of VR and design patterns in indoor perception

In modern interior design, VR technology provides designers with a new perspective and tools, making the indoor environment perception and design more intuitive and efficient. The ranging modeling technology of RSSI helps to accurately draw the spatial structure of indoor environments and optimize design schemes. An interior design model based on VR and the PSO-BP algorithm is constructed.

3.1 Distance measurement modeling of integrating RSSI into VR technology in indoor perception

In indoor environment perception, traditional methods have problems such as inaccurate spatial understanding and incomplete information transmission. However, based on VR technology, designers have a more intuitive and comprehensive understanding of indoor environments through immersive experiences. Especially when VR technology is combined with RSSI ranging modeling, it can more accurately grasp the spatial structure of the indoor environment, thereby optimizing the design scheme. The deployment and communication diagram of the RSSI ranging application is shown in Figure 1.

Figure 1

RSSI ranging flowchart.

In Figure 1, when the node that needs to be located starts working, it broadcasts a request signal to the surrounding anchor nodes. The anchor node that receives the request records the RSSI value between it and the target node in the data packet. It is sent back to the target node as a response signal. The target node then packages and transmits all received information to the coordinator. The coordinator further passes it on to the server side. The RSSI signal has instability and significant fluctuations. After receiving multiple data packets containing RSSI values, the server needs to perform preprocessing, such as using mean filtering or Gaussian filtering methods, to reduce errors caused by environmental factors. The preprocessed RSSI values are converted into distance values through indoor wireless signal propagation models to obtain the actual distance between the target node and each anchor node. Through these three steps, precise indoor positioning can be achieved. The flowchart of indoor positioning technology incorporating RSSI ranging is shown in Figure 2.

Figure 2

Flow chart of indoor positioning technology integrated with RSSI ranging.

In Figure 2, the target node sends a request signal to multiple anchor nodes in an indoor environment. After receiving the signal, the anchor node records the RSSI value between them and the target node and returns it to the target node. The target node then aggregates all the received RSSI data and sends it to the server. The server then performs data preprocessing on the RSSI values, such as mean filtering or Gaussian filtering, to reduce errors caused by environmental interference. After that, the preprocessed RSSI values are converted to the actual distance by an indoor wireless signal propagation model. Finally, the precise coordinates of the target nodes are calculated using a geometric positioning algorithm (such as the trilateral positioning method or maximum likelihood estimation method). The phased positioning and tracking are realized to complete the indoor positioning process. In the tracking stage, a filtering algorithm is applied to correct the node position to be located. In this conversion process, the indoor wireless signal propagation model is crucial. During the indoor propagation of signals, they are easily affected by environmental factors such as obstacles, personnel movement, wall reflections, etc. These factors lead to significant differences in signal propagation between actual channels and ideal channels [15]. In the free space propagation model, the received power between these two is shown in the following equation [16]:

(1) P r ∞ 1 d 2 .

In Eq. (1), P r is the signal power received by the target node. d stands for the distance value between anchor nodes. The received power of the signal is shown in the following equation:

(2) P r ( d ) = P t G t G r λ 2 ( 4 π ) 2 q 2 L .

In Eq. (2), P t stands for the transmission power, G t is the antenna gain of node t, G r is the antenna gain of node r, λ stands for the wavelength of the radio wave, q stands for the distance value between the transmitter and the receiver, and q stands for a system loss parameter that is independent of the propagation environment. The path loss in free space is shown in the following equation:

(3) P L ( d ) = 10 log P t P r = − 10 log G t G r λ 2 ( 4 π d ) 2 .

In actual positioning environments, due to the movement of personnel, numerous obstacles, and complex indoor environments, signals can be affected. Simple free space channel models are not suitable for real indoor positioning environments. The experience of the logarithmic path loss model (LPLM) is shown in the following equation:

(4) P r ( d ) = P r ( d 0 ) − 10 n log d d 0 + X σ .

In Eq. (4), P r ( d ) stands for the strength of the received signal, d 0 is the reference distance, n stands for the path loss factor, and X σ stands for related to the environment and represents the magnitude of environmental noise. Eq. (4) improves the accuracy of distance estimation by correlating the received signal strength with the distance relationship. Therefore, in subsequent positioning algorithms, the location information of the target node can be more accurately determined. In indoor environments, floors and walls have an impact on signals, resulting in the modified K-factor model, as shown in the following equation :

(5) P r ( d ) = P r ( d 0 ) + 10 n log ( d ) + ∑ i = 1 M N w i L w i + ∑ j = 1 Q N j j L j j .

In Eq. (5), N w i stands for the total certain walls, M stands for the total wall types in the signal transmission environment, L w i is the loss index corresponding to a certain wall type, N j j is the total number of floors that pass through a certain type, Q is the total number of floor types during signal transmission, and L _jj is the loss index corresponding to the floor type. To make the signal closer to the true value, the mean filtering algorithm or Gaussian filtering algorithm is usually used. The mean filtering algorithm is to average the RSSI values collected at the same distance, as shown in the following equation:

(6) RSSI = 1 L ∑ k = 1 L RSSI k .

In Eq. (6), Q is the total value. To reduce the impact of such fluctuations, filtering techniques are usually used to process the RSSI values, filtering out singular values in the RSSI values and making the collected RSSI values more stable. The constraint principle is expressed as the relationship between A and B. If there is an interconnection relationship between them, then the sum of the confidence levels of A-type position node n and all B-type position nodes is less than 1. Otherwise, the number of location nodes n and type B will exceed 1. When there is an input sequence L , the sequence is shown in the following equation:

(7) L = { l 1 , l 2 , l 3 , … l M } Y = { y 1 , y 2 , y 3 , … y N } Y ˆ = f ( x 1 , x 2 , x 3 , … x T ) .

In Eq. (7), Y ˆ = { y ˆ 1 , y ˆ 2 , y ˆ 3 , … y ˆ N } refers to the predicted output result. It is assumed that the time series are consistent, and there is a certain correlation between the data in the time series. Their relationship is interdependent. The pore convolution method is adopted. The number of compression parameters is used to increase the area of the receiving field. The algorithm for processing a frame is displayed in the following equation:

(8) F ( t ) = ( X × f d ) ( t ) = ∑ i = 0 k − 1 f ( i ) ⋅ x t − d ⋅ i .

In Eq. (8), d refers to the expansion coefficient, k refers to the convolutional kernels, t − d ⋅ i refers to the size of the time slice. To further improve the accuracy of 3D spatial relationships, the project plans to obtain connection points from images and videos. 2D connection points are used to reconstruct 3D spatial relationships. By collecting the original pose, sequence images are obtained. Then, 2D drawings are modeled. The vibration is removed by the time-domain convolution method. Combined with the operation of the angle vector method, a 3D spatial display model is obtained.

3.2 Construction of interior design model based on VR and PSO-BP

Traditional interior design models mainly rely on floor plans and physical models. This approach is inefficient and difficult to quickly iterate and optimize the design scheme [17,18,19]. However, if VR technology is combined with the PSO-BP algorithm, a new interior design model can be constructed. In this model, designers can visually find the effect of the design scheme in a virtual environment. Meanwhile, through the PSO-BP algorithm, the design scheme can be automatically optimized. This improves design efficiency while also improving accuracy and quality. The flow chart of indoor positioning based on the PSO-BP algorithm is shown in Figure 3.

Figure 3

Flow chart of indoor positioning using the PSO-BP algorithm.

In Figure 3, the following six steps are included. The sample data is collected. The PSO algorithm trains the optimal weight and bias values. They are applied to train the BP network. The real-time collected RSSI values are input into the network to obtain prediction results and locate and track them [20,21]. For the acquisition of the initial training set, a training set composed of personnel in a stationary state is shown in the following equation:

(9) train _ data _ s = [ ( RSSI 1 ′ , d 1 ′ ) , … , ( RSSI i ′ , d i ′ ) , … , ( RSSI L ′ , d L ′ ) ] T .

In Eq. (9), AP 1 and AP 2 are the two anchor nodes deployed indoors and d i ′ is the distance value. The training set for personnel movement status is shown in the following equation:

(10) train _ data _ m = [ ( RSSI 1 ′ ′ , d 1 ′ ′ ) , … , ( RSSI i ′ ′ , d i ′ ′ ) , … , ( RSSI L ′ ′ , d L ′ ′ ) ] T .

In Eq. (10), the method for obtaining the training set in the personnel movement is the same as that in the personnel stationary. The data acquisition in the personnel movement only requires personnel to move between two anchor nodes. The deployment diagram of indoor anchor nodes is shown in Figure 4.

Figure 4

Indoor anchor node deployment diagram.

In Figure 4, three deployment routes are selected. One of the anchor nodes is moved along three different trajectories. Personnel are trained in stationary or mobile training to obtain stationary or mobile training datasets. If we want to obtain a more accurate training network, more paths can be added to collect more RSSI values. The RSSI value is saved corresponding to the corresponding distance value, forming the training sample set of the training network [22,23]. The training sets under different personnel states are collected. The training set is input into the PSO-BP algorithm of the joint adaptive enhancement strategy. The mapping relationship between RSSI values and distance is obtained, which is the barium distance model under two different personnel states. The indoor personnel activity perception and detection algorithm achieves automatic switching of ranging and positioning models based on different personnel states. When a person is detected to be in a stationary state, a mapping network trained on a human stationary training set is used for ranging and localization. When a person is detected to be in a mobile state, a mapping network trained using the training set of the person’s mobile state is used for ranging and positioning. The trajectory of the detection point during motion can be approximated as a curve. The first-order derivative relationship between the position and time of the detection point coordinates is shown in the following equation:

(11) x t ′ = d x d t = m ′ ( t ) y t ′ = d y d t = n ′ ( t ) .

In Eq. (11), x t ′ is the motion rate of the detection point on the x-axis and y t ′ is the motion rate of the detection point on the y-axis. A discrete data is collected at fixed equal time intervals. The time interval is very small, which can be replaced by difference. The motion direction of the detection point is shown in the following equation:

(12) φ ( t ) = tan − 1 y t − y t − 1 x t − x t − 1 + π, x t − x t − 1 < 0 tan − 1 y t − y t − 1 x t − x t − 1 + 2 π, x t − x t − 1 < 0 , y t − y t − 1 < 0

In Eq. (12), φ ( t ) is the motion direction of the detection point. The angle change and threshold change in the detection direction are shown in the following equation:

(13) Δ φ t = φ t − φ t − 1 Δ φ t < τ , Data retention Δ φ t > τ , Data filter .

In Eq. (13), τ is the threshold. The threshold is taken as 15°. The angle change in the motion direction of the detection point is used as a preliminary data filter. When the change rate is less than 15°, it is retained. When it is greater than 15°, it is filtered out. To further suppress high-frequency noise, the simple moving average filtering method can also be used for data denoising, as shown in the following equation:

(14) S = ( x i + x i − 1 + … + x i − n + 1 ) n , ( n = 1 , 2 , 3 , … ) .

In Eq. (14), n is the length of the data sequence. The value directly affects the detailed information of the detection point.

4 Analysis of indoor perception and design model based on VR

This study fully utilizes advanced VR. It is applied in interior design and perception. This study mainly analyzes how VR changes indoor perception and how space can be optimized through design models. This model can more intuitively display the design scheme, helping designers and customers better understand and perceive indoor space.

4.1 Performance analysis of VR in indoor perception

The model parameters include group size, maximum iterations, inertia weight, learning factor, etc. In order to implement and validate this model, the open-source database MySQL is used for data storage and processing. The parameter table is shown in Table 1.

Table 1

Configuration details for VR indoor perception performance analysis

Category	Configuration
Model parameters	Population size, maximum iterations, inertia weight, learning factor
Database	MySQL
Processor	Intel Core i7
Memory	16 GB RAM
Graphics card	NVIDIA GeForce GTX 1080
VR device	Oculus Rift
Storage	1TB SSD
Operating System	Windows 10
VR software	Unity 3D

Table 1 lists the configuration details of VR indoor perception performance analysis in this study and clearly clarifies the impact of software and hardware environments on the experimental results. The combination of the Intel Core i7 processor and 16 GB RAM ensures efficient performance in data processing and computation, enabling complex algorithms to run in real time and process large amounts of ambient data. The NVIDIA GeForce GTX 1080 graphics card provides the necessary graphics processing power for VR scenes, ensuring a smooth and realistic visual experience. 1 TB SSD storage ensures the speed of data read and write, which can effectively support interaction and real-time feedback in dynamic scenarios. In addition, the Windows 10 operating system and Unity 3D development environment provide system stability and development flexibility. This series of configurations provides an efficient and stable experimental platform for research, which helps to achieve more accurate indoor perception and positioning results. In addition, to achieve an interactive experience in the virtual environment, the Oculus Rift VR helmet is also adopted. Anchor nodes are located at different distances in different personnel states. The fluctuation of RSSI values varies. The RSSI values in different situations are collected and aggregated. The fluctuation of RSSI values in different situations is calculated. Table 2 shows the changes in the mean RSSI values.

Table 2

The average RSSI value under different personnel statuses and distances

Personnel status	4 m	5 m	6 m	6.5 m	7 m	7.5 m	8 m
Unmanned	−41.9475	−43.3490	−46.0385	−48.6993	−50.5725	−52.4263	−54.2695
Personnel stationary	−50,6793	−50.5175	−49,9958	−48.9862	−49.5892	−50.1923	−50.7953
Personnel movement	−51.7803	−51.0592	−50.1814	−49.5693	−49.8592	−50.1492	−50.4392
Personnel running	−52.8903	−52.1692	−51.2914	−50.6793	−51.1292	−51.5792	−52.0292
Personnel Jumping	−53.9803	−53.2592	−52.3814	−51.7693	−52.3992	−53.0292	−53.6592
Personnel squatting down	−54.0703	−53.3492	−52.4714	−51.8593	−52.6692	−53.4792	−54.2892

Table 2 displays the signal strength values at different distances (4–8 m) for different personnel status. Personnel status includes unmanned, stationary, moving, running, jumping, and squatting. As the distance increases, the signal strength shows a gradually decreasing trend. The detection performance under different sliding window lengths is shown in Figure 5.

Figure 5

Comparison of detection performance under different sliding window lengths. (a) Detection performance under different sliding window lengths. (b) Detection performance of JAYA under different sliding window lengths. (c) Detection performance of Naive Bayes under different sliding window lengths. (d) Detection performance of SVM under different sliding window lengths.

Figure 5 shows the influence of different sliding window lengths on the performance of personnel motion status detection. With the increase of sliding window length, the overall detection rate showed a decreasing trend. This is because long sliding windows may cause response delays in detection algorithms when dealing with dynamic changes, thereby affecting the accuracy of real-time detection. When the sliding window length was 50, the detection rate reached the highest, exceeding 90% accuracy, showing that this parameter setting had the best performance in dynamic detection. However, when the sliding window length was too long, the detection accuracy was significantly reduced. Therefore, in practical applications, it is necessary to choose the appropriate sliding window length according to the environment and demand to balance the detection sensitivity and accuracy. The detection performance at different distances is shown in Figure 6.

Figure 6

Comparison of detection performance at different distances.

In Figure 6, as the distance increases, the detection rates of the four detection algorithms for detecting personnel movement status all decreased. The proposed algorithm had the highest detection rate compared to the other three algorithms such as support vector machines (SVM). When the distance value was less than 5 m, the detection rate was greater than 90,070. With the increase of the distance value, the detection accuracy decreased, but the detection rate was still relatively high. As the distance value increases, the detection rate remains relatively high. The results show that the proposed method has higher robustness and adaptability when dealing with an indoor complex signal environment and can effectively improve the positioning accuracy of personnel. The results emphasize the importance of selecting suitable algorithms and improving RSSI-ranging technology in indoor positioning, providing strong support for practical applications.

4.2 Analysis of design model positioning results based on VR

This experiment is deployed in an indoor positioning environment based on the optimal horizontal distance value of anchor nodes, while considering shaded areas as obstacles. The node to be located is carried by the tester, which can be in a stationary or moving state. The LPLM, backpropagation algorithm (BP), PSO-BP algorithm, and the combination of personnel state detection and PSO-BP-Adaboost algorithm proposed in this article are used for distance calculation. Then, the maximum likelihood estimation algorithm is applied to locate the target node. Finally, the extended Kalman filtering algorithm is used to track and filter the positioning results. This process can effectively locate and track the position and movement status of the nodes to be located in the indoor environment. Table 3 provides the detailed parameters of the VR indoor perception design model positioning configuration.

Table 3

VR indoor perception design model positioning configuration details

Experiment parameters	Parameter values	Experiment parameters	Parameter values
Virtual Reality Scene Scale	100 m²	Monitor resolution	1920 × 1080
Processor performance	Intel Core i7	Network connection speed	100 Mbps
Memory size	16 GB RAM	Operating system	Windows 10
Graphics card performance	NVIDIA GeForce GTX 1080	VR device	Oculus Rift
Localization accuracy requirement	Within 1 m	Storage space	1 TB SSD
VR software	Unity 3D	Programming language	C#

From Table 3, this configuration ensures that the model runs in high-performance environments, improving the accuracy and efficiency of indoor environment perception and localization. Figure 7 shows the cumulative distribution functions of errors for four positioning techniques. The PSO-BP-Adaboost indoor wireless positioning method with a joint adaptive strategy showed the best positioning performance and generated the smallest error. Next was the indoor wireless positioning method based on PSO-BP distance measurement. The formula method performed the worst in positioning performance. The PSO-BP-Adaboost indoor wireless positioning algorithm in this article achieved a 90% positioning error of 0.82 m under the joint adaptive strategy, which was significantly superior to the other three methods, greatly improving the positioning performance. The results show that the proposed method has higher robustness and adaptability when dealing with an indoor complex signal environment and can effectively improve the positioning accuracy of personnel. The results emphasize the importance of selecting suitable algorithms and improving RSSI ranging technology in indoor positioning and provide strong support for practical applications.

Figure 7

Error cumulative distribution functions for four positioning methods.

Next, in an indoor positioning environment, the positioning nodes are carried by fixed personnel and moved indoors. Combining maximum likelihood estimation localization and extended Kalman filtering algorithm, the position is tracked. Figure 8 shows the percentage measurement error of the four positioning algorithms when personnel move. Due to personnel activities, the indoor positioning environment has undergone significant changes. This has a significant impact on the signal, thereby affecting the accuracy of the positioning results. In the table, the formula method, the BP, and the PSO-BP all showed an increase in error. The formula method had the worst positioning performance, with 67% of data positioning errors reaching 4.01 m. However, the PSO-BP-Adaboost algorithm proposed in this study had a mechanism for detecting personnel status. By adjusting the distance measurement model, high positioning accuracy was maintained. This method had a positioning error of less than 0.68 m in 67% of data. Compared to the static state of the personnel, the positioning error increased by 0.15 m. In many indoor positioning applications, a 0.15 m error is still acceptable, especially in environments such as smart homes, augmented reality, and indoor navigation that require high flexibility and real-time performance. However, compared to the other three methods, this method has significantly higher positioning accuracy and better positioning performance.

Figure 8

Measurement error percentage of four positioning algorithms.

Figure 9 shows the error comparison of using the extended Kalman filtering in the actual personnel positioning process. Figure 9(a) shows the comparison between the measurement error before filtering and the actual error. The results showed that the maximum measurement error before filtering was 1.4 m. The maximum measurement error was 0.7 m. Figure 9(b) shows the comparison between the measurement error after filtering and the actual error. The result shows that the measurement error after filtering was close to the actual error, and the maximum longitudinal error value was 0.8 m. The results show that the error value is reduced, the path is smoother, the location path is closer to the real path, and the location performance is improved.

Figure 9

Comparison of errors during the actual positioning process of personnel. Error changes (a) before and (b) after applying filtering algorithms.

The performance of the proposed method is further analyzed through practical application, and the obstacle factor is added for testing. The results are shown in Table 4. From Table 4, the detection rate of the PSO-BP-Adaboost algorithm reached 95.5% in the static state and short distance. Despite significant signal attenuation at 6 m and in a moving state, the detection rate remained at 85.6%, showing stronger adaptability compared to 75.3% of the BP algorithm and 60.1% of the traditional algorithm. These results show that the proposed method has better stability and accuracy in complex indoor environments, providing a solid foundation for daily application.

Table 4

Comparative analysis results of practical application

Personnel status	Distance (m)	Daily obstacle effects (description)	PSO-BP-Adaboost detection rate (%)	BP algorithm detection rate (%)	Traditional algorithm detection rate (%)	Detection accuracy (error/m)
Static	4	Minor impact, good signal	95.5	87.2	75.2	0.52
Static	6	Medium impact, shaded	92.0	80.6	70.7	0.73
Move	4	Large impact, frequent occlusion	90.3	80.7	65.8	0.91
Move	6	Great influence, signal attenuation is significant	85.6	75.3	60.1	1.18
Run	4	Some influence, some path good	88.5	76.7	62.8	1.05
Run	6	Big block, weak signal	83.0	71.4	58.9	1.23
Jump	4	Minor impact, signal fluctuation	87.0	75.5	60.5	1.24
Jump	6	Moderate impact, increased signal volatility	81.5	70.2	55.1	1.31

The study recognizes that there are multiple challenges to effectively integrating theoretical models and algorithms into existing design workflows, including data acquisition, real-time processing, and environmental adaptability. To overcome these challenges, it is recommended to establish a standardized data acquisition process to ensure consistent RSSI data under different environmental conditions and possibly multiple sensors to improve signal accuracy. At the same time, the algorithm needs to be optimized for real-time data processing to improve the response speed and ensure that the delay is controlled within the acceptable range. In addition, it is recommended to work closely with the designer to conduct user testing and feedback to adjust and optimize the algorithm parameters according to the specific project needs. Based on these measures, the practical applicability of the model can be significantly improved, making it flexible and scalable in different industrial applications, thereby improving the overall design efficiency and effectiveness.

5 Conclusion

By combining RSSI ranging technology and PSO-BP algorithm, a new indoor environment perception and design model based on VR technology is proposed. Experimental results showed that under the join adaptive strategy, the positioning error of PSO-BP-Adaboost algorithm was 0.82 m, up to 90%, which was obviously better than the other three methods, greatly improving the positioning performance. The positioning error of PSO-BP-Adaboost was less than 0.68 m in 67% of the data. The proposed PSO-BP-Adaboost algorithm was superior to other comparison algorithms in terms of human motion state detection and location accuracy, especially in dynamic environment. The model not only improves the accuracy of indoor positioning, but also provides more intuitive support for designers to design in complex indoor environments. However, there are still some shortcomings in the research, such as insufficient treatment of dynamic obstacles in complex indoor environments, lack of testing in larger application scenarios, and lack of discussion on the direct impact of user behavior patterns and psychological states on indoor environment perception and design in VR environments. Future research can further explore how to combine more complex environmental factors and optimize algorithms to improve the adaptability and universal applicability of the system. By combining user experience research, future work will help build a more humane and intelligent VR design platform, thus promoting higher design innovation ability and efficiency.

Funding information: Authors state no funding involved.
Author contributions: All authors have accepted responsibility for the entire content of this manuscript and approved its submission.
Conflict of interest: Authors state no conflict of interest.
Data availability statement: All data generated or analysed during this study are included in this published article.

References

[1] Alonso A. Ocuweld brings welding education to virtual reality. Weld J. 2022;101(3):9–10.Search in Google Scholar

[2] Chalmers D. Virtual reality is as real as physical reality, but just different. N Scientist. 2022;253(3376 Suppl):42–5.Search in Google Scholar

[3] Narang K, Imsirovic A, Dha J, Claire FS. Virtual reality for anatomy and surgical teaching. Adv Exp Med Biol. 2022;1397:135–49.10.1007/978-3-031-17135-2_8Search in Google Scholar PubMed

[4] Xue L, Yang C. Virtual reality painting dexterous hand gesture control algorithm and simulation. J Electron Imaging. 2022;31(5):51422.1–13.10.1117/1.JEI.31.5.051422Search in Google Scholar

[5] Takami A, Watanabe K, Makino M. Immediate effect of video viewing with an illusion of walking at a faster speed using virtual reality on actual walking of stroke patients. J Phys Ther Sci. 2021;33(8):560–4.10.1589/jpts.33.560Search in Google Scholar PubMed PubMed Central

[6] Marre Q, Caroux L, Sakdavong JC. Video game interfaces and diegesis: The impact on experts and novices’ performance and experience in virtual reality. Int J Human-Computer Interact. 2021;37(11/15):1089–103.10.1080/10447318.2020.1870819Search in Google Scholar

[7] Ya L. The application of virtual reality technology in art design. Vol. 827(May). Singapore: Springer; 2022. p. 739–46.10.1007/978-981-16-8052-6_93Search in Google Scholar

[8] Zhang W, Fu X, Li W. Point cloud computing algorithm on object surface based on virtual reality technology. Comput Intell. 2021;38(1):106–20.10.1111/coin.12449Search in Google Scholar

[9] Cummings JJ, Cahill TJ, Wertz E, Zhong Q. Psychological predictors of consumer-level virtual reality technology adoption and usage. Virtual Real. 2022;27(12):1357–79.10.1007/s10055-022-00736-1Search in Google Scholar PubMed PubMed Central

[10] Yangfei C. Research on the design system of ship interior environment based on 3D virtual reality technology. Ship Sci Technol. 2020;42(8):17–9.Search in Google Scholar

[11] Shimeng H, YIchao X. Construction method of indoor monochromatic light environment based on virtual reality technology. J Beijing Univ Civ Eng Archit. 2022;38(9):27–35.Search in Google Scholar

[12] Jian L, JiE Z. Design and exploration of virtual simulation experiment teaching project based on VR – taking “Landscape visual spatial analysis virtual simulation experiment teaching project” as an example. ChJ ICT Educ. 2021;22(7):42–8.Search in Google Scholar

[13] Liu L. The artistic design of user interaction experience for mobile systems based on context-awareness and machine learning. Neural Comput Appl. 2022;9:34.10.1007/s00521-021-06160-xSearch in Google Scholar

[14] Saucedo PE, Acosta‐Salmón H, McLaurin‐Moreno D, Castillo-Domínguez A, Melgar-Valdés CE, Mazón-Suástegui JM. Freshwater pearl culture in Mexico: historic context, present status and future perspectives. Rev Aquac. 2021;13(3):1379–96.10.1111/raq.12527Search in Google Scholar

[15] Hui J, Zhou Y, Oubibi M, Di W, Zhang L, Zhang S. Research on art teaching practice supported by virtual reality (VR) technology in the primary schools. Sustainability. 2022;14:1–4.10.3390/su14031246Search in Google Scholar

[16] Meng W, Ding H, Liu H. Research on the application of virtual reality technology in environmental art design. International Conference on Machine Learning, Image Processing, Network Security and Data Sciences. Vol. 1(1). Cham: Springer; 2022. p. 1–10.Search in Google Scholar

[17] Pack A. University EAP students’ perceptions of using a prototype virtual reality learning environment to learn writing structure. Int J Comput-Assist Lang Learn Teach. 2020;10(1):27–46.10.4018/IJCALLT.2020010103Search in Google Scholar

[18] Steel A, Robertson CE, Taube JS. Current promises and limitations of combined virtual reality and functional magnetic resonance imaging research in humans: A commentary on Huffman and Ekstrom. J Cogn Neurosci. 2020;33(2):1–8.10.1162/jocn_a_01635Search in Google Scholar PubMed PubMed Central

[19] Berti M, Maranzana S, Monzingo J. Fostering cultural understanding with virtual reality: a look at students’ stereotypes and beliefs. Int J Comput-Assist Lang Learn Teach. 2020;10(1):47–59.10.4018/IJCALLT.2020010104Search in Google Scholar

[20] Bridget C. Development of a virtual reality clinically oriented temporal bone anatomy module with randomised control study of three-dimensional display technology. BMJ Simul Technol Enhanc Learn. 2020;7(5):352–9.Search in Google Scholar

[21] Vaghela KR, Trockels A, Carobene M. Active vs passive haptic feedback technology in virtual reality arthroscopy simulation: which is most realistic. J Clin Orthop Trauma. 2021;16(1):249–56.10.1016/j.jcot.2021.02.014Search in Google Scholar PubMed PubMed Central

[22] Ferraz-Torres M. Passive or interactive virtual reality? The effectiveness for pain and anxiety reduction in pediatric patients. Virtual Real. 2022;14(2):1–10.10.1007/s10055-022-00633-7Search in Google Scholar

[23] Yang M. Research on vehicle automatic driving target perception technology based on improved MSRPN algorithm. J Comput Cognit Eng. 2022;1(3):147–51.10.47852/bonviewJCCE20514Search in Google Scholar

Received: 2024-08-29

Revised: 2024-12-05

Accepted: 2024-12-16

Published Online: 2025-04-29

This work is licensed under the Creative Commons Attribution 4.0 International License.

Articles in the same Issue

https://doi.org/10.1515/nleng-2024-0080

Keywords for this article

virtual reality; indoor environmental perception; receiving signal strength indication; particle swarm backpropagation; positioning accuracy

Creative Commons

BY 4.0