Energy-Efficient Dynamic Street Lighting Optimization: Balancing Pedestrian Safety and Energy Conservation

Abstract

Residential street lighting plays a crucial role in enhancing the reassurance for pedestrians returning home late at night. However, street lighting is sometimes recommended and required to be kept at lower levels at night, due to problems such as light pollution, energy consumption, and negative economics. To solve these problems, this study designed a new Dynamic tracking lighting control mode capable of greater interactivity. Our study aimed to determine whether this new interactive lighting model can balance pedestrian safety with energy savings, compared with other lighting approaches used in low-light environments. In this experiment, 30 participants explored four lighting conditions in a simulated nighttime street environment through virtual reality (VR) and completed their assessment of each lighting mode. The statistical analysis of the results using the Friedman ANOVA test revealed that the Dynamic tracking lighting mode had advantages in improving the pedestrians’ reassurance compared with the other three lighting modes. Moreover, an additional recognition test experiment recorded the distance between each other whenever a participant recognized a stranger agent. The experimental results showed that this Dynamic tracking lighting mode can improve pedestrians’ ability to recognize others in low-light environments. These findings provide new strategies and ideas for urban energy conservation and environmental protection.

1. Introduction

For urban areas, light pollution has increased globally, with 80% of the total population now living under light-polluted skies [1]. Although artificial light at night (ALAN) can be beneficial for societies, it also has negative economic and health impacts. For example, artificial lighting accounts for around 20% of global electricity consumption [2]. Furthermore, light pollution affects vital processes such as the production of melatonin, which can influence the health, fitness, and survival of residents [3,4,5].

In cities, light emissions mainly come from streetlights, advertising, aesthetics, etc. [6,7]. In small communities, light emissions are often dominated by public street lighting [8]. At night, clouds can now amplify skyglow to illuminances similar to that of bright twilight [9,10]. Thus, some research has presented policy recommendations that limit ALAN to guarantee major energy reductions in street lighting systems [11]. Many countries have also enacted policies to limit ALAN, such as dimming and selectively switching off lights [12]. Therefore, the traditional ways of controlling street lighting are based on time and/or daylight sensor dimming or switching off all streetlights in the street. However, traditional control approaches reduce pedestrians’ feelings of safety (FOS) at night, due to people preferring having light in their immediate surroundings [13,14]. This safety issue would hinder the promotion of energy-saving policies at night and may cause some technical problems in street lighting. These are new challenges for street lighting design.

Therefore, a new lighting design or control method is needed to ensure the safety of pedestrians in low-light environments. Advances in LED and IoT technology have led to the development of intelligent adaptive streetlighting systems capable of detecting the presence of pedestrians or cars and adjusting the light accordingly [15,16,17,18]. For example, Juntunen et al. [19] installed an adaptive intelligent lighting system that detects approaching road users with passive infrared (PIR) sensors and increases brightness in the surroundings, saving energy and making users comfortable. However, these dynamic smart lighting systems, depending on their surrounding parameters, were primitive and difficult to maintain in remote areas, especially when streetlights are dimmed to save energy and reduce light pollution. Following these systems, some studies proposed directional light sources as a new way of adapting lighting systems to radiate light in specific directions, thereby improving the efficiency of the streetlamps [20,21]. Despite this growing interest, there is limited empirical research exploring the safety benefits of intelligent adaptive lighting for pedestrians [22].

However, these studies on directional light sources were original and prototypical, focusing on technical and hardware design. Therefore, we refined the lighting design and parameters for directional lighting to improve the interactivity of street lighting, considering different scenarios of sidewalks that use it. In this paper, we called this directional lighting design “Dynamic tracking lighting”, which makes the streetlight always follow and illuminate every pedestrian on the sidewalk area. The tracking light only comes from the lamppost near the detected individual, while other lampposts remain dim. It should be noted that “Dynamic tracking lighting” was set to only function at nighttime whenever there are a few people walking alone on sidewalks. Prior research has suggested that brighter lights help law-abiding pedestrians to identify the intentions of people approaching them by lighting their facial expressions and body language [23], thereby reducing the ability of offenders to conceal themselves in the dark [24,25]. To sum up, our purpose was to demonstrate this dynamic tracking street lighting design in virtual reality (VR) and examine its impact on pedestrians’ reassurance. Additionally, in previous research, reassurance was used to describe the confidence of a pedestrian walking alone after dark and encompassed the terms perceived safety and fear of crime (FOC) [26,27]. Building upon these previous findings, we proposed and tested the following hypotheses:

H1: 

Compared to a street with all lights dimmed, “Dynamic tracking lighting” would enhance pedestrians’ reassurance, while maintaining the entire street at a dim level.

H2: 

Compared to a street with all lights lit brightly, “Dynamic tracking lighting” would approach/achieve the same level of pedestrians’ reassurance, while maintaining the entire street at a dim level.

In addition, the ordinary adaptive lighting system suggested in previous studies [19,22] could not maintain continuous illumination due to limitations in static irradiation angles, compared with the Dynamic tracking lighting system and its high interactivity. Therefore, we also compared Dynamic tracking lighting with present sensor-based adaptive lighting and tested the following hypothesis:

H3: 

Compared to the present sensor-based adaptive lighting control method, “Dynamic tracking lighting” would enhance pedestrians’ reassurance.

As a result, we chose to focus on examining a total of four street lighting modes (all bright, all dim, dynamic tracking, and ordinary adaptive), using two scenarios (with and without stranger agents present) in the same street environment in virtual reality (VR), to test the above-mentioned hypotheses. Furthermore, we designed an additional extended objective experiment to test the ability of pedestrians to recognize others under different lighting conditions, recording the distance whenever they noticed a stranger agent. Specifically, the scope of our research was to test whether interactive lighting design (dynamic tracking) could be used to solve the problem of energy conservation at night, by verifying enhancements of this model to pedestrians’ reassurance and their recognition abilities in low-light environments.

2. Literature Review

2.1. Pedestrians’ Reassurance and Streetlighting

In the past, there has been considerable research on lighting and fear of crime (FOC). However, asking about fear of crime has been criticized on the basis that it may actually create such fears [28]. Thus, asking about the reassurance is a better methodology, e.g., to what extent lighting increases confidence to walk outside after dark [27].

This paper concerns streetlighting in residential areas. In this study, reassurance means the confidence a pedestrian might gain from road lighting (amongst other factors) to walk along a road, in particular if walking alone after dark [27]. A road in which reassurance from lighting provides confidence to walk is one that offers higher perceived safety and lower fear of crime [26,27].

Previous research investigated the effects of certain properties of lighting on reassurance such as illuminance, uniformity, color, and type of light source [13,29,30,31,32,33,34]. The research of Boyce et al. [30] has shown that higher illuminances enhance reassurance, while the effect of light spectrum is negligible. Although specific interests varied, these results have shown that higher illuminance reduces a measured level of fear of crime in general, especially illuminating approaching pedestrians [13].

Besides illuminance, illuminance uniformity has also been considered an important parameter for reassurance on residential streets [35]. Narendran et al. [36] examined the effect of illuminance uniformity on the perceived safety in a parking lot and found that a lower illuminance could be used to achieve the same level of perceived safety when the illuminance uniformity was greater. It should be noted that it was achieved by lighting up dark areas where criminals could lurk [23].

To solve these problems above, we designed Dynamic tracking lighting, improved from the research of adaptive lighting [19,22]. The Dynamic tracking lighting mode we suggest was designed to always track and brighten the area when detecting anyone existing or passing by using the nearest lamppost, while the other lampposts were kept dim. As for whether the Dynamic tracking lighting system would reduce illuminance uniformity, the research from Jedon et al. [37] noted that the attention and alertness of pedestrians would be enhanced under non-uniform lighting conditions made by dynamic lighting. From these studies above, the Dynamic tracking lighting system should have the advantages of reassurance in illuminance and illuminance uniformity compared with all lights dimmed, and even might have a few advantages compared with the normal bright street environment. Thus, we examined that whether the Dynamic tracking lighting system could enhance pedestrians’ reassurance somehow, compared with normal streetlighting modes (all lights dimmed and all kept bright).

2.2. Pedestrians’ Reassurance and Presence of Others

One of the most common justifications for the use of light at night is that streetlighting benefits pedestrian safety. Some previous studies have shown that improved lighting can reduce fear of crime [38], even if the effect on the incidence of crime is not significant [39,40]. While the benefits of using light at night to reduce crime are uncertain, lighting can still affect crime through indirect mechanisms. One of the ways is enabling people on the street after dark to recognize the intentions of others and to see well all around them, as well as allowing surveillance by the community in general and by the authorities [23]. Some studies have examined the impact of other people approaching pedestrians, such as evaluating other people as a critical task for pedestrians [41], and that pedestrians tend to look at people who are approaching [42,43].

Generally, the presence of other legitimate people can provide a sense of security and belonging [34,44]. Conversely, the presence of others is not a cue that carries a single meaning and can also instigate fear [45]. Lupton [46] suggested that the “unpredictable stranger” is a primary source of fear in urban public areas, as this is closely tied to feelings of uncertainty and a loss of control. Thus, Kang et al. [22] studied the impact of different lighting conditions on individuals’ perceived safety when they were confronted with pedestrians exhibiting uncivil behavior, using virtual agents in virtual reality (VR). As far as we know, it has not been thoroughly investigated whether the presence of other strangers impacts the reassurance of pedestrians walking during the day or at night.

To avoid this effect and distraction, we employed VR to simulate two scenarios (with or without stranger agents present) in all four different streetlight environments. At the same time, we conducted an additional experiment to compare the ability of pedestrians to recognize people under different lighting conditions.

2.3. Virtual Reality (VR) in Studies of Lighting Environments

Virtual reality (VR) is a computer-generated artificial environment simulating physical reality, which allows individuals to interact with it through some devices [47]. Generally, traditional indoor experiments using photographs and simulated environments have difficulties to completely replicate the environment of the real world [48]. One of VR’s key advantages is its immersive nature [49], which allows researchers to simulate real environments while controlling selected variables, saving time and costs [50]. Another significant benefit of VR is its ability to easily replicate and simulate social interaction scenarios in the real world using virtual human agents [51]. For example, Slater et al. [52] simulated violent incidents in VR to examine the conditions under which bystanders would intervene in a violent attack.

According to these advantages, VR has gained widespread adoption in lighting environment research, addressing various topics such as the relationship between nighttime illuminance of streets and fear of crime (FOC) [53], the influence of environmental interventions in residential neighborhoods on FOC [54], and the impact of adaptive street lighting on safety perception [22].

With the growing popularity of VR as a research tool in lighting studies, many researchers also have examined the validity of VR by comparing it with the real world. Chen et al. [48] discussed the feasibility of using virtual reality in representing lighting environments and confirmed that VR image playback methods were able to produce feelings that were very close to the true feelings of humans experiencing actual physical scenes. Similarly, Abd-Alhamid et al. [55] compared a simulated virtual office using high dynamic range (HDR) images and found no significant differences in participants’ visual performance of lighting perception between VR and real environments. While these studies suggest VR is a valid method for studying the perception of lighting environments, they also showed several limitations. For example, Rockcastle et al. [56] found that VR image playback methods might not provide an adequate medium for reproducing accurate lighting perception in lighting scenes that are dim or high in contrast with potential glare. In summary, these findings suggest that while the validity of using VR might not totally be confirmed, VR still is worthy as a useful tool for lighting environment studies because it is immersive and interactive.

Thus, we chose VR to build the streetlighting environment in our experiment. For our purpose, installing many lampposts in real spaces is time consuming, costly, and does not allow us to control all possible variables (e.g., time and crowds). Compared with the past methods, VR simulation has been greatly improved. Although the VR approach is still limited compared with reality, it is acceptable for our qualitative comparison experiment.

3. Method

3.1. Overview

The scope of this study was sidewalk lighting in residential neighborhoods at night when pedestrians were less frequent and light emissions needed to be reduced. This study employed a within-subject design, where participants were exposed to all four lighting conditions (all bright, all dim, dynamic tracking, and ordinary adaptive) randomized in a simulated virtual reality (VR) street environment. Furthermore, each condition had two scenarios: with or without stranger agents present. This approach enabled participants to compare four lighting conditions and offer feedback on their preferred lighting conditions.

3.2. VR Simulation and Lighting Conditions

3.2.1. VR Simulation Apparatus

The experiment took place in a laboratory within a science building on the university campus. The VR laboratory room measured 3.8 m in length and 3.15 m in width. Equipment included a personal computer with an NVIDIA RTX 4070 FE graphics card (NVIDIA, Santa Clara, CA, USA). The VR head-mounted display (HMD) was the Oculus Quest 2 with a dual-eye display resolution of 3664 × 1920 pixels; it supports a maximum refresh rate of 90 Hz, offering a field of view (FOV) of 110°. Participants were able to navigate the 3D virtual space freely, using the joystick on the Oculus Touch Controller (Oculus, Irvine, CA, USA).

The Unity software, version 2020.3, was employed for VR project development, while Blender, version 3.6, was used for modeling and texturing. After creating a simulated street model in Unity, we rendered realistic dynamic street lighting effects in VR via Unity’s High-Definition Render Pipeline (HDRP).

According to previous research [22,54], participants would have a near-real experience of street lighting conditions in VR through these tools. Figure 1 illustrates the development process in this study.

3.2.2. Street Environment of VR Simulation

The nighttime economy in China is demonstrating an upward trend, and an increasing number of people are going out at night [57]. In order to solve the contradiction between the reassurance of late-night pedestrians with low light density in residential areas and reducing light pollution and energy consumption, we examined the possibility of improving street lighting design by applying pedestrian tracking technology.

The simulated street environment of this VR experiment was an actual replica of a road in a residential area of China. It was modified according to the Chinese Standard for Lighting Design of Urban Road [58] to adapt to VR development requirements: the dimensions and parameters of the simulated street are based on the real streets, and the surrounding building models are based on paid models with the Chinese architectural style from the Unity Asset Store. In order to provide the participants with a near-real experience in the VR environment, we added real white noise recorded at night as background sounds (e.g., sounds of trees in the wind and cars in the distance).

As shown in Figure 2, the length of the simulated street was 60 m, and the width of the street was 12 m. There were eight luminaire poles arranged on both sides of the street, with four on each side. For the stranger agent side of the street, these four luminaire poles were marked as A1–A4. For the other side of the street, the participant’s side, these four luminaire poles were marked as B1–B4. The height of each luminaire pole in the simulated street environment was 5 m, and the average distance between the poles was 12 m in the Unity HDRP setup. In addition, the luminaire tilt was 0 degrees, and the luminaire overhang was 0.5 m. The illuminated area of each street lamppost in VR was about 8 m in diameter (4 m in radius), and the Soft Shadows Spotlight of each lamppost was set with 8 Range, 111 Inner spot angle, and 141 Outer spot angle in Unity.

Additionally, we adjusted the environmental parameters in Unity to match the real night, to make the night environment more realistic and valid in VR. We placed the VR camera on the ground of an open space in the virtual world, and the viewpoint was pointed upward toward the sky, while we used a Minolta CL-500A spectrophotometer [56] to measure the illuminance of the VR headset lens. In this process, we repeated adjustment of the sky parameter settings in the Unity interface until the illuminance value was the same as the sky illuminance value we measured via spectrophotometer at midnight in a real space with no artificial lights (0.3 lx). The sky’s main parameters are listed in

Table 1. The settings of sky parameters in the Unity project.

Settings Property		Value
Skybox
	Tint Color	0
	Exposure	0.01
Environment Lighting
	Intensity Multiplier	2.5
Environment Reflections
	Intensity Multiplier	1
Fog
	Color	0
	Mode	Exponential Squared
	Density	0.03

.

3.2.3. Lighting Simulation

In addition to tracking pedestrians’ movements, the main differences in illuminance for different lighting modes in the simulated street environment were designed. According to design of the lighting modes (all bright, all dim, dynamic tracking, and ordinary adaptive), there were two levels (bright and dim) of the illuminance. In this study, we used mean horizontal illuminance to measure different lighting levels of luminaire poles in streetlighting modes.

Based on the guidance in BS EN 13201-3:2015 [59], we recorded illuminance values at ten evenly spaced locations between the two lamp posts in the VR street environment. These measurements were taken along the center of the lanes on both sides of the road, yielding 20 measurement points.

Although VR is better suited for exploring the subjective perception of lighting, namely perceived brightness [22], it is worth noting that the lux values in this study do not provide an exact representation of illuminance. Due to the characteristics of the VR HMD’s display, it was hard to measure the illuminance of artificial lights in the VR environment. Thus, we chose to measure the equivalent lens’ illuminance via the Minolta CL-500A spectrophotometer [56] while the VR game camera approached the ground and was aimed down towards the road surface on the vertical axis in the VR environment. In this way, we measured the 20 points and obtained the equivalent average illuminance.

Previous studies suggested that for illuminances below about 10 lx, small increases in illuminance produce a large increase in perceived safety [30], and 10 lx mean horizontal illuminance was suggested as good lighting [27,35]. Therefore, given that a minimum illuminance of 1 lx is suggested for the detection of trip hazards [60], we chose 10 lx as the bright lighting level and 1 lx as the dim lighting level, to make sure there was enough difference between the lighting modes.

Finally, we repeated the illuminance measurement in VR until the mean illuminance was as close to 10 lx and 1 lx as possible, and then we had the Unity light source intensity values of the luminaire poles set at about 130 and 10, respectively.

3.2.4. Interaction in Virtual Reality

In this experiment, each lighting mode included two test scenarios, that is, with or without a stranger. We expected that the no-stranger scenario would make participants focus on their experience of perceived safety, while the with-stranger scenario would make participants focus on their experience of fear of crime (FOC). In this way, participants combined the two experiences to assess the reassurance of different street lighting modes.

The simulated street was a two-way lane. People are used to walking on the right in China. Thus, we chose to have the participant and the stranger agent walk on one side of the road on their right. Their movement routes are shown in Figure 2.

The action mode of the simulated stranger agent was to repeat his walking route at a speed of 1 m/s, complete with sounds of footsteps. When the agent reached the end of the route, it was teleported to the beginning of the route, and then continued to walk. This ensured that no matter how the participant played in the with-stranger scenario, there was always a stranger on the road.

Participants were able to navigate the 3D virtual street environment freely. There were no restrictions on the participants’ activities; they were simply told to walk from their starting position to the end of the road at will. When using the stick on the controller to walk instead of body movement, the upper limit of the movement speed of the participant was set at 1.5 m/s, slightly faster than the agent, due to playing design. This meant that the participant could control their moving speed freely from 0 to 1.5 m/s using the joystick on the Oculus Touch Controller. Moreover, 1.5 m/s is the walking speed of most people [61].

3.2.5. Details of the Four Lighting Modes

The streetlamps in the four lighting conditions of the VR experiment had two brightness levels, and the value was set according to the method mentioned in Section 3.2.3.

(1)

Mode 1: All bright

In the first mode where all lights are bright, every lamppost was set to static, just like normal lampposts in the real world, and only kept the street bright. The mean illuminance was set to 10 lx (equivalent average 10 lx/130 lamppost light intensity in Unity).

(2)

Mode 2: All dim

In the second mode where all lights are dimmed, every lamppost was set to static, just like normal lampposts in the real world, and only kept the street dim. The mean illuminance was set to 1 lx (equivalent average 1 lx/10 lamppost light intensity in Unity).

(3)

Mode 3: Dynamic tracking

In the third mode, “Dynamic tracking lighting”, the streetlight always followed the participant and kept the sidewalk area bright (using the LookAt function in Unity), with the tracking light only coming from the lamppost near the individual detected, while the other lampposts were kept dim. The brightness of the lamppost that detected people was increased to the same level as used in mode 1 with all lights bright (10 lx/130 intensity in Unity). Meanwhile, the brightness of the other lampposts were set to the same level as mode 2 with all lights dim (1 lx/10 intensity in Unity).

The proximity threshold for activating the lighting was set at 6 m, which aligns with the width of the road or half of the distance between two lampposts. This meant that when a pedestrian leaves the detection range of one lamppost, he or she can be detected by the next lamppost immediately, which ensures the pedestrians are always traced by a brighter light. Meanwhile, the mean illuminance of street was kept dim.

(4)

Mode 4: Ordinary adaptive

In the fourth mode, “ordinary adaptive”, compared with mode 3 (dynamic tracking), only the tracking of the pedestrian by the street lamppost was cancelled, that is, the irradiation angles of lights were kept static. Instead, the lamppost only increased in brightness when a human was detected. This meant that pedestrians would not be illuminated by street lights for part of the time, because for a single lamppost, the lighting range is smaller than the inductive detection range.

The conditions for the four modes are shown in Figure 3.

Additionally, in order to avoid tracking interference in the with-stranger scenario because of the Unity settings, we set lampposts A1–A4 to track the agent and lampposts B1–B4 to track the participant. As for how this lighting mode can avoid interference in real-world applications, we will discuss it further in the Discussion. In this part, we only focused on achieving this dynamic tracking effect in the VR environment conveniently, in order to study its impact on pedestrians’ reassurance.

Among these modes, the latter two require pedestrian detection technology, which we simulated using VR programming code. The flowchart of the program code controlling pedestrian detection and tracking modes is depicted in Figure 4. This set of control codes was applied to each of the eight streetlights in Unity. Moreover, the participants and stranger agents were assigned the same priority level within the system, triggering identical dynamic lighting effects.

3.3. Participants

A previous study from South Korea mentioned that adults in their 20s have a higher fear of crime [53] and a lower resistance to VR equipment [62]. Thus, we recruited experiment participants from among college students through the university’s posters and social media channels. To determine the necessary sample size for testing the within-subject design of this study, a priori power analysis [63] indicated that a minimum of 27 participants would be required to achieve a power of 0.95 for a large effect (0.4), with a significance level (α) of 0.001 for the four repeated measurements of experimental lighting modes.

According to previous similar research [53,54], gender could have an impact on the results of VR experiments. To eliminate the impact of different genders on the results of the experiment, we maintained the gender ratio of the participants at 50:50. Every participant confirmed that he or she had no eye diseases, no motion sickness, and no psychological disorders in the past, according to their self-report.

A total of 34 participants completed all materials. Finally, 30 participants were recruited, a number deemed sufficient to test the study hypothesis. Four participants were excluded from analyses because of severe motion sickness (n = 1), color weakness (n = 1), and to control the gender ratio (n = 2).

3.4. Procedure and Survey Questionnaire

Before the beginning of the experiment, each participant was invited to our VR laboratory located at the university. At the start of the study, participants were taught how to put on the HMD of the Oculus Quest 2. Subsequently, to get used to the game controller and HMD, participants first practiced in a test environment (an empty space sample).

After demonstrating an understanding of the game controller and HMD, the participant was briefed on the background of the experimental tasks, that he or she would walking home alone on the street, late at night, for some reason such as having worked late, and that there were few other people on the road. Meanwhile, the participant was informed that they would explore the street with three different lighting modes and was told the characteristics of each lighting condition before beginning each exploration.

Then, the participant would navigate, at random, one of four street lighting modes with the no-stranger scenario, and was told to walk through from the start point to the end of the road at will. Similarly, the participant would walk with the same street lighting mode, applied once again with the with-stranger scenario. After completing VR explorations with one of the lighting conditions, the participants were asked to complete a survey questionnaire, assessing their reassurance of the lighting scenarios they just experienced. A comparison between the four lighting modes with the with-stranger scenario is illustrated in Figure 5.

After a short break of several minutes, the participants would explore the remaining three lighting modes in random order and complete the survey questionnaire. At the end of the entire experience, we allowed participants to revise their assessments if they wished to do so. The experimental procedure is shown in Figure 6.

After finishing the four questionnaires, an additional objective experimental test was designed to examine the differences in participants’ ability to recognize the stranger agent approaching from a distance in darkness under the four lighting modes. We reloaded the with-stranger scenario that was the same as the assessment experiment. The participant kept standing at the starting point and pressed a preset button on the controller whenever he or she thought they could see the presence of the agent. At the same time, the system calculated and recorded the distance between the participant and the agent. When the participant pressed a button, the scene was switched to the next lighting condition after a three-second black screen transition, while the positions of both the participant and the agent were reset. The preceding steps would be repeated four times in random order.

The questionnaire assessing reassurance was based on previous studies [35,37,64]. We chose the same three questions [64] addressing safety perception (Q1), comfort (Q2), and the sense of potential criminal risk (Q3) to assess the reassurance of pedestrians. After finishing the first exploration of lighting conditions, the questionnaire was explained to the participants. Based on previous studies [35,37], we modified the definition of the terms for our scope and better consistency. Specifically, it was clarified that the term ‘safety’ pertained to the cognitive aspect of safety, such as confidence of being alone, whereas ‘comfort’ related to the sensation of comfort or anxiety while walking, and ‘risk’ was associated with the alertness to potential threat of crime. The survey questions are shown in

Table 2. The survey questions to assess the reassurance of pedestrians.

No.	Questions	Please Provide a Rating According to Your Assessment
1	How safe do you feel when you are walking alone on this road	No confidence				Feel safe
		1	2	3	4	5
2	How comfortable do you feel when you are walking alone on this road	Anxious				Comfortable
		1	2	3	4	5
3	How risky do you feel when you are walking alone on this road	Very risky				Not at all risky
		1	2	3	4	5

. Responses were captured on a 5-point scale, with higher ratings indicating better (better reassurance) outcomes.

3.5. Measures

After completing all four repeating explorations of lighting conditions, each participant was then asked to answer the 13 items of a post-questionnaire designed to test the validity of the VR experiment (e.g., ‘I had a sense of being in the scenes displayed’), rating each item on a 5-point Likert scale (1 = strongly disagree, 5 = strongly agree). For this post-questionnaire, we used the scale used by Son et al. [53].

3.6. Analysis

The Shapiro–Wilk test was conducted to examine the distribution of the raw data and validate its normality. The results indicated that the data did not follow a normal distribution. Therefore, we employed non-parametric tests to analyze the data. Specifically, using the lighting mode as the independent variable and the assessment of the reassurance (safety, comfort, and risky) as the dependent variable, we conducted Friedman’s ANOVA tests to examine the reassurance assessment data under the four lighting modes, based on the within-subject design of this study.

4. Results

4.1. Association Between Lighting Modes and Reassurance

The results of the questionnaire from 30 participants with three assessment scores for the four lighting modes are shown in Figure 7 and

Table 3. The assessments of the four lighting modes in nine questions (Mean and SD).

Lighting Mode	Q1 Safety		Q2 Comfort		Q3 Risky
	Mean	SD	Mean	SD	Mean	SD
All bright	4.30	0.79	4.37	0.81	4.27	0.83
All dim	1.37	0.67	1.33	0.61	1.40	0.67
Dynamic tracking	4.53	0.63	4.17	0.87	4.57	0.73
Ordinary adaptive	3.77	0.86	3.60	0.77	3.83	0.87

, which include the average scores and SDs obtained from the experiment. The results indicated that the data did not follow a normal distribution. Therefore, we employed the Friedman ANOVA test to analyze the data in

Table 4. Results of the Friedman ANOVA test assessing the four lighting modes.

Variable	Lighting Mode								Friedman ANOVA
	All Bright		All Dim		Dynamic Tracking		Ordinary Adaptive		Test Result
	Mean Rank	Sum Rank	Mean Rank	Sum Rank	Mean Rank	Sum Rank	Mean Rank	Sum Rank	Chi-Square	DF	p
Q1 Safety	3.15	94.5	1.03	31	3.4	102	2.42	72.5	61.03	3	<0.001
Q2 Comfort	3.43	103	1	30	3.15	94.5	2.42	72.5	63.91	3	<0.001
Q3 Risky	3.05	91.5	1	30	3.42	102.5	2.53	76	61.09	3	<0.001

. The Friedman ANOVA test revealed statistically significant differences in ratings among the four lighting conditions in the reassurance of participants, based on the average responses to multiple questions. Specifically, the reassurance scores of the All-bright lighting mode and the Dynamic tracking lighting mode were higher than the All-dim mode. However, the differences between the scores of the All-bright mode and the Dynamic tracking mode were not significant. Therefore, we used Dunn’s test to compare paired results of the four lighting modes in

Table 5. Comparative paired post hoc analysis of the results of four lighting modes using Dunn’s test.

Lighting Mode	Q1 Safety			Q2 Comfort			Q3 Risky
	Sum Rank Diff	Z	Sig	Sum Rank Diff	Z	Sig	Sum Rank Diff	Z	Sig
All bright–All dim	63.5	6.35	1	73	7.3	1	61.5	6.15	1
All bright–Dynamic	−7.5	0.75	0	8.5	0.85	0	−11	1.1	0
All dim–Dynamic	−71	7.1	1	−64.5	6.45	1	−72.5	7.25	1
All bright–Adaptive	22	2.2	0	30.5	3.05	1	15.5	1.55	0
All dim–Adaptive	−41.5	4.15	1	−42.5	4.25	1	−46	4.6	1
Dynamic–Adaptive	29.5	2.95	1	22	2.2	0	26.5	2.65	1

Sig = 1 indicates that the difference of the means is significant at the 0.05 level.

.

The results provided support for our hypotheses H1 and H2, namely, the “Dynamic tracking lighting” would provide better pedestrians’ reassurance compared with all lights dimmed, and would approach the same pedestrians’ reassurance level compared with all lights lit brightly.

Additionally, while the difference was not significant between All-bright and Dynamic tracking modes, we noticed that Dynamic tracking modes had a small advantage in Q1 (safety) and Q3 (risky) and had a disadvantage in Q2 (comfort) compared with All-bright mode. This was consistent with previous studies [37] where although dynamic lighting might reduce the visual comfort, attention and alertness of pedestrians would somehow be enhanced under the non-uniform lighting conditions made by dynamic lighting. This needs to be confirmed further through experiments and studies.

At the same time, consistent with previous studies [22], the difference in safety between Ordinary adaptive lighting and All-bright lighting was not significant (Q1 and Q3), while the Ordinary adaptive lighting brought less sense of comfort (Q2) and less reassurance compared to All-bright lighting. The results in Table 3 and Table 4 show that the two intelligent control modes (modes 3 and 4) can greatly improve the reassurance compared with the All-dim lighting mode. However, the difference between the Ordinary adaptive lighting and the Dynamic tracking lighting was significant, and the latter gained a higher reassurance score than the Ordinary adaptive lighting. This supported our hypothesis H3 in which lighting design with stronger interactivity could enhance pedestrians’ reassurance.

4.2. Recognition Ability Test

This experiment tested participants’ recognition abilities, through recording the recognition distances between the participant and the agent. The participant remained standing and was called not to move, and to just push the record button on the controller whenever they noticed a walking stranger agent. The results of the test are shown in

Table 6. The distances between each other when the participant recognized the stranger agent. (higher is better).

Lighting Mode	Mean (m)	SD
All bright	40.59	1.20
All dim	34.59	1.08
Dynamic tracking	43.44	1.04
Ordinary adaptive	39.95	1.14

.

The results show that All-bright lighting mode and Ordinary adaptive lighting mode had greater advantages than All-dim lighting mode in the ability of pedestrians to recognize the presence of others. This suggests that increased brightness could also help pedestrians to recognize others. Especially, in all four lighting conditions, the Dynamic tracking lighting system makes it easier for pedestrians to detect the presence of others in advance. This proves that interactive lighting design could improve pedestrians’ safety and recognition ability when walking alone at night.

4.3. Validity of the VR Experiment

To test the validity of the VR experiment, namely, the sense of physical presence in a VR space, immersion, ecological validity, and side effects, we conducted a post-questionnaire survey using the assessment indicators modified by Son et al. [53].

As shown in

Table 7. The results of the feasibility evaluation survey of the VR experiments.

Item	Frequency of Response					Mean	SD
	1	2	3	4	5
Sense of physical space						4.61
I had a sense of being in the scenes displayed.	1	0	3	5	21	4.50	0.88
I felt I was visiting the places in the displayed environment.	0	0	1	3	26	4.83	0.21
I felt that the characters and/or objects could almost touch me.	1	1	0	8	20	4.50	0.88
Engagement						4.61
I felt involved in the displayed environment.	0	0	2	6	22	4.67	0.37
I enjoyed myself.	0	0	0	3	27	4.90	0.09
My experience was intense.	0	1	1	17	11	4.27	0.48
Ecological validity						4.54
The content seemed believable to me.	0	0	0	2	28	4.93	0.06
The displayed environment seemed natural.	0	0	0	1	29	4.97	0.03
I had a strong sense that the character and objects were solid.	1	2	8	12	7	3.73	1.03
Negative effects						1.36
I felt dizzy.	25	1	2	2	0	1.37	0.79
I felt nauseous.	29	1	0	0	0	1.03	0.03
I felt I had a headache.	30	0	0	0	0	1.00	0.00
I had eyestrain.	14	5	8	2	1	2.03	1.34

, sense of physical space, engagement, and ecological validity scored high average values (≥4.5 points). Most participants responded that they had no physiological problems. Adverse reactions mainly reported included dizziness (n = 2) and eye fatigue (n = 3).

5. Discussion

5.1. The Impact of Lighting Conditions on Reassurance

Firstly, this study mainly analyzed the relationship between the two modes (all bright and all dim) of static streetlighting conditions and the dynamic lighting mode in the reassurance of pedestrians. An experience-based questionnaire survey based on simulated VR street environments was conducted on 30 young adults. The data collected were subjected to a Friedman ANOVA test analysis and a comparative paired post hoc analysis. Meanwhile, we eliminated factors such as age and gender that might have influenced the experimental outcomes and simplified the experiment, since the scope of this qualitative experiment was to compare the impact of lighting modes on reassurance rather than examine the best parameters of dynamic lighting conditions.

The analysis results revealed that the “Dynamic tracking lighting” system was advantageous in reassuring pedestrians, namely, the “Dynamic tracking lighting” system provided better pedestrians’ reassurance compared with all lights dimmed (H1), and would approach the same pedestrians’ reassurance level compared with all lights lit brightly (H2). The results indicate that brighter lights contribute to a higher level of reassurance, which was consistent with previous studies [22,23,27,30], especially since pedestrians preferred lights approaching them [13]. Furthermore, the study findings indicate that the tracking strategy can enhance the reassurance by lighting others. As in the previous studies reviewed by Boyce [23], our results reconfirmed that the perceived safety could be enhanced by lighting up those dark areas in which criminals can lurk. It also reconfirmed that dynamic lighting enhances the attention and alertness of pedestrians [37], although this needs to be examined further through experiments and studies.

Additionally, from the results between the All-bright and the Dynamic tracking modes, we noticed that Dynamic tracking modes had a small advantage in Q1 (safety) and Q3 (risky), meaning that the Dynamic tracking streetlighting could provide the same reassurance level with fewer lighting resources to conserve energy and reduce lighting pollution compared with normal street lighting at night. However, the Dynamic tracking mode had a slight disadvantage in Q2 (comfort) compared with All-bright mode, which means that the effects of different dynamic methods needs more examination. There might be a better design of dynamic adaptive lighting that increases reassurance and alertness without reducing comfort and causing anxiety.

Secondly, this study examined the impact of ordinary adaptive lighting on reassurance compared with other lighting conditions. However, from the results between the All-bright and the Ordinary adaptive modes, we found that ordinary adaptive lighting could approach the reassurance level of all-bright lights, with some disadvantages. This is consistent with previous studies [22], meaning that current ordinary adaptive street lighting still needs to be improved, which was one of the purposes of this study, as the results show that the Dynamic tracking lighting mode was better than ordinary adaptive lighting in reassuring pedestrians (H3).

Some studies suggest that lighting uniformity was also important, and people generally prefer consistent illumination across an environment rather than lighting that targets specific areas [65]. Cutting and Vishton [66] suggested that the boundary of vista space beyond the action space, which encompasses the immediate circular area, be approximately 30 m around a person’s location. In another words, pedestrians become anxious because of possible threats in the dark space far from their location. The results of this study show that when a potential threat was extra illuminated, the overall brightness at further distances could be reduced, which is an advantage of the Dynamic tracking lighting mode. However, it is worth noting that this dynamic mode only performs well in well-paved roads, since for roads that are broken or have complicated environments, lampposts that are too dim increase the risk of injury when walking at night.

Furthermore, the actual performance of the Dynamic tracking lighting system in terms of reassurance depended heavily on whether the user understands that all people will be illuminated; the actual effect may be compromised by inadequate public awareness if pedestrians are not told that such lighting would help them to recognize others. Through the introduction to the experiment, the participants understood the characteristics of each of the lighting conditions. Otherwise, it could be foreseen that the participants would become confused and terrified without any prior information.

However, the present study has several limitations. First, this study still has some limitations in VR technology as other experimental studies [53,62,67,68]. Due to their imaging principles, VR devices have a significant gap in display quality compared to the real world. For instance, the narrower field of view (FOV) in VR headsets compared to the human FOV may impact participants’ perception of the virtual environment [69]. Even with calibration measures, it is still impossible to completely ensure the accuracy of VR displays in photometry. Additionally, this study’s simulation in Unity also has certain limitations, as the simulation effects of characters and lighting fixtures differ from real-world pedestrians and streetlights, although simulations in VR using models rendered by game engines could provide more realistic effects compared with other experimental methods of simulation. For an initial conceptual study and discussion, we considered VR devices to be an efficient tool and attempted to mitigate the drawbacks of VR in the experimental design as much as possible. Based on this research, in the future, we should strive to develop corresponding hardware devices in the real world and conduct more targeted quantitative studies to explore details such as lighting parameter settings.

Second, in order to secure the generalizability of the research, analysis of sufficient amounts of samples must be accompanied; however, the participants were limited to adults in their 20s in this study. Although this limitation is generally shared by other VR experimental studies, it is necessary to acquire statistically validated sample sizes and to expand the participants to include other population groups vulnerable to crime, such as older adults and children.

5.2. Role of Interactivity in Street Lighting Design

The Dynamic tracking lighting mode proposed in this paper is more interactive than the adaptive lighting proposed in previous studies [19,22]. Through real-time dynamic adjustments of the irradiation angle, light has the function of accompanying the lone walker and helping them recognize the existence of others. The experimental assessment results showed that the Dynamic tracking lighting system results in higher reassurance than ordinary adaptive lighting, which features only simple detection and adjustment (H3). The additional test also proved that such interactivity could not only improve pedestrians’ subjective reassurance, but also enhance pedestrians’ ability to recognize others such that pedestrians can notice the presence of others at greater distances. This interactive lighting design has obvious advantages for pedestrians, whether in seeking help from others or preventing potential threats.

This interactive tracking design has a certain reference value for both developed and underdeveloped areas. For developed areas, it could improve residents’ well-being and reduce light pollution. For less developed areas, this design can achieve a higher reassurance with fewer lighting resources and reduce the pressure on infrastructure, using some low-power electronic chip devices instead of large-scale lighting installation.

5.3. The Possibility of Real-World Applications

We did not consider the implementation of multi-person tracking in the simulation experiment, which is not the focus of this paper and represents a technical problem. The expectation of the Dynamic tracking lighting mode is that lights will be able to aim at everyone, and the lampposts will not interfere with each other. Although only simple cases were simulated in the experiment, we can still discuss the possible problems and solutions of tracking lighting in reality.

Firstly, matrix lighting technology [70] could achieve one light source illuminating multiple targets at the same time, and Internet of Things (IoT) technology [21] could achieve collaborative operation between street lights where multiple targets are involved that could lead to possible deficiencies of light sources.

Second, maintaining tracking and illuminating pedestrians might cause glare problems. One possible solution is through controlling the lighting angles and using pedestrian path prediction, avoiding high-lighted lampposts aiming at pedestrians from the front of their path, and trying to illuminate the paths of pedestrians only from light sources behind them. Another possible solution is using matrix lighting technology to avoid lighting the pedestrians themselves, but to only to light up the ground around them instead. Additionally, lighting angle control strategies are needed to keep the lights only on the ground of the road area, and to avoid light pollution caused by lighting into the windows of houses.

Third, the study did not address participants’ perception of surveillance or concerns related to privacy that may arise with the implementation of the Dynamic tracking street lighting. These abovementioned issues are important considerations for the design of networked intelligent lighting and warrant exploration in future research.

6. Conclusions

In this study, we sought to examine how the Dynamic tracking lighting mode influenced the three factors (safety, comfort, and risky) of pedestrians’ reassurance. The results of the virtual reality (VR) simulation experiment showed that both the All-bright and the Dynamic tracking modes significantly increased pedestrians’ reassurance compared to the All-dim lighting condition, which reconfirmed that increased illumination as well as dynamic tracking could enhance reassurance levels. However, the difference in impact on the reassurance between the All-bright and the Dynamic tracking modes was not significant, which meant that the Dynamic tracking streetlighting condition could achieve a higher reassurance similar to having all lights up, while using lower levels of streetlighting. Moreover, the Dynamic tracking lighting system achieved a higher reassurance than the ordinary adaptive lighting system. The additional recognition experimental test results showed that the pedestrians in Dynamic tracking lighting had a higher recognition ability than the other three lighting modes. However, whether this system could completely replace the normal full streetlighting condition in late-night conditions still needs to be further examined.

As such, the Dynamic tracking lighting system has great potential to improve pedestrians’ reassurance, while contributing to energy conservation and light pollution control. Researchers and practitioners alike need to explore better dynamic control strategies through further examination.