4654 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 39, NO.

14, JULY 15, 2021

An Accurate Ranging Algorithm Based on Received

Signal Strength in Visible Light Communication
Changeez Amini, Paeiz Azmi , Senior Member, IEEE, and Seyed Sadra Kashef

Abstract—In this paper, the line of sight (LOS) propagation is biased that is demonstrated in [8]. In [9], the impact of the
model in visible light communication for a receiver with tilting deviation of the radiation pattern compared with the Lambertian
plane is precisely derived. In addition, a relatively simple and rea- radiator on the distance estimation is quantified. In [10] the
sonably accurate model of LOS propagation is proposed to employ
received signal strength algorithm. Then, the maximum likelihood authors combines angle of arrival (AOA) and RSS information to
estimator is applied in range estimation where a new shot noise enhance positioning accuracy in an asynchronous VLP system.
(RSN) source as a function of received signal power, is considered. In most papers, the receiver plane is set horizontally. But in
Moreover, the Cramer-Rao lower bound is derived in closed form practice, a receiver may not lay horizontally and has a tilt angle
by considering RSN and tilting plane. Next, the influence of the which requires three-dimensional positioning approach to be
Lambertian emission order on the channel model is investigated,
theoretically and the optimum value of the Lambertian emission adopted. As a fact, increasing the line of sight (LOS) power by
order is presented. Finally, performance is evaluated numerically. tilting receiver plane provides an opportunity for indoor high
precision three-dimensional positioning. Different works are
Index Terms—Visible light communication, received signal
strength, range estimation, tilting, dependent shot noise.
done on VLP, but they remain silent on the subject of using
channel parameters related to tilting technique (such as irradi-
ance and incidence angles) to improve the VLP system [11]. The
I. INTRODUCTION VLC received signal depends on the geometry of the LED source
and photo-detector (receiver), including the irradiance angle [12]
RADUALLY, visible light communication (VLC) has
G become a research hotspot in indoor environments be-
cause of its potential large bandwidth [1]. It can be deployed
as well as the variance of noise detected by the receiver. There
exists an approach to obtain the angle differences of arrival
(ADOA) with arbitrary tilting in 3-D coordinate systems [13].
for safe, secure, and high-speed communication where radio
The required VLP data includes the 3-D positions of the LEDs
frequency is hazardous or even forbidden (such as submarines,
and their identifying information, such as the signal frequency or
air planes, underground mines, intrinsically safe environments
light color. If the positions of the LEDs are known to the receiver,
like petrochemical industries and so on). As a result of immunity
error analysis on AOA of the light signal can be estimated
to electromagnetic interference, the visible light spectrum serves
approximately [14]. 3-D localization with multiple LEDs using
as a promising candidate for the industrial Internet of Things [2].
a weighted angle-of-arrival (WAOA)-based locater is studied
Besides communication, the light-emitting diode (LED) based
in [15], which is then used as the starting point of the recursive
localization can lead to high positioning accuracy, especially
estimator based on RSS.
for indoor visible light positioning (VLP). Since LED lighting
In recent years, VLP has been developing through complex
systems are installed everywhere, a seamless positioning service
processes with multiple transmitters based on the angle of arrival
can be provided by visible light within much larger coverage than
technique which has resulted in approximate formulas. Mean-
current techniques.
while, [16], [17] supposed that tilting does not change the posi-
By using the photodiode (PD), extracting features from the
tion of the receiver which can be inaccurate. In [18], the influence
received signal strength (RSS) is a typical VLP method. [3],
of transmitter and receiver orientation on the channel gain for
[4]. By using RSS measurements, the Cramer-Rao lower bound
RSS ranging-based VLP is investigated. A Monte-Carlo simula-
(CRLB) of the positioning error is determined in [5], [6].
tion study of the impact of LED tilt on the accuracy of RSS-based
CRLB as a fundamental limit on RSS based range estimation
VLP for different localization approaches is presented in [19].
is investigated in [7]. RSS-based distance estimation in VLP
The simultaneous position and orientation estimation problem
using RSS is studied in [20], for a visible light system containing
Manuscript received December 1, 2020; revised March 2, 2021 and April 8, multiple LEDs and multiple PDs. In general, four sources of
2021; accepted May 5, 2021. Date of publication May 10, 2021; date of current
version July 16, 2021. (Corresponding author: Paeiz Azmi.) noise in VLC are considered which are thermal noise, shot
Changeez Amini and Paeiz Azmi are with the Department of Electrical noise of background radiation, shot noise of dark current and
and Computer Engineering, Tarbiat Modares University, Tehran, Iran (e-mail: shot noise of received signal (RSN). These noise sources are
[Link]@[Link]; pazmi@[Link]).
Seyed Sadra Kashef is with the Department of Electrical and Computer introduced in section II in detail. Unlike other noises, the RSN
Engineering, Urmia University, Urmia, Iran (e-mail: [Link]@[Link]). depends on the received signal power. However, the effect of
Color versions of one or more figures in this article are available at https: RSN on VLP is generally considered exactly the same as the first
//[Link]/10.1109/JLT.2021.3078544.
Digital Object Identifier 10.1109/JLT.2021.3078544 three noise sources. It is demonstrated in [21], the dependence

0733-8724 © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See [Link] for more information.
AMINI et al.: ACCURATE RANGING ALGORITHM BASED ON RECEIVED SIGNAL STRENGTH IN VISIBLE LIGHT COMMUNICATION 4655

of the RSN variance on the geometry of the system noticeably x = Rp H(0) Pt + w = P + Pdiff + w, where P = Rp ηLOS Pt
affects the total variance
√ of the noise and so CRLB. Typically, and Pdiff = Rp ηdiff Pt . ω refers to additive white Gaussian noise
RSN can increase CRLB in RSS considerably. Consequently, (AWGN) and Pt refers to the transmitted optical power. The
CRLB and estimators are needed to drive reasonably accurate. responsivity Rp is the conversion factor from the optical to
Whereas it is shown in other fields such as channel capacity [22] electrical domain. In this paper, we assume that the responsivity
that the effect of input-dependent Gaussian noise (RSN) can not of the photodiode is a constant taken at the central optical
be ignored. To the best of our knowledge, all of the VLP works frequency of visible light. In the case that the receiver faces
focus only on the VLC channel DC gain and remain silent on up and the transmitter points down, cosφ = cosϕ = hd and the
the subject of RSN effect on positioning accuracy. initial LOS power can be written as [12]
In this paper, we take into account the impact of irradiance (m + 1)S hm+1
angle on the channel gain of the VLC system and ranging P0 = R p P t , (2)
2π dm+3
algorithms. We will show that the optimum tilting angle provides
where vertical distance is represented by h.
the maximum received optical power. Optimal tilting is inves-
Generally, four noise sources are considered in a visible light
tigated in this paper and suboptimal tilting angle is introduced
system as follows [27]:
in a closed form for the first time. The latter angle increases
the LOS gain and results in a simple but effective estimator. 1) Johnson Noise σTH : The Johnson (thermal) noise is σT H =
8πκTe 2 16π 2 κTe Γ 2 2
Moreover, by considering the relation between the RSN and G ηSI2 B + gm η S I3 B 3 , where the bandwidth
signals transmitted by LED, the maximum likelihood estimation of the electrical filter that follows the photodetector is repre-
(MLE) and CRLB are carried out. The results of this article can sented by B, k is the Boltzmann’s constant, Te denotes the
be employed to calculate the security rate of the physical layer absolute temperature, G represents the open-loop voltage gain,
in the presence of eavesdroppers. η is the fixed capacitance of photodetector per unit area, Γ
The rest of paper is organized as follows: Section II describes a is the Field-effect transistor (FET) channel noise factor, gm is
VLC propagation channel model and related basic assumptions. the FET transconductance, and the noise-bandwidth factors are
General LOS channel model and suboptimal LOS channel power I2 = 0.562 and I3 = 0.0868.
are derived in Section III. We employ MLE in section IV. 2) Shot Noise from Background Radiation σBG : The current
In order to evaluate the accuracy of the proposed estimation noise frombackground radiation is given as
and ranging algorithm, the CRLB is computed analytically in σBG = 2qRp pBS S λ̆B, where q is the electron charge, pBS
Section V. Simulation results are presented in Section VI and
finally, conclusions are drawn in Section VII. is the background spectral irradiance, λ̆ is the bandwidth of the
optical filter.
3) Shot Noise from Dark Current σDC : The noise from the
II. SYSTEM MODEL
dark current
√ of a photodiode can be expressed as
This paper adopts the commonly used model for the channel σDC = 2qIDC B, where IDC refers to the photodiode dark
LOS gain of VLC [23] current.
 (m+1)S 4) RSN σSS : The shot noise
 which depends to received signal
2πd2 cos φ Ts (ϕ) g(ϕ) cosϕ, 0 ≤ ϕ ≤ ϕcon ,
m
ηLOS = power is given by σSS = 2q(P + Pdiff )B.
0, ϕ > ϕcon
Eventually, the variance of noise can be summarized by
(1)
σ 2 = σT2 H + σBG
2 2
+ σDC 2
+ σSS = σT2 + σSS
2
. (3)
where ϕ and φ are the angle of incidence and the angle of
irradiance according to the vertical normal axis against trans- It should be mentioned that the first three noise sources do not
mitter surface, respectively. Furthermore, m is the Lambertian depend on received optical power. By contrast, RSN is directly
emission order and d is the distance between the LED and the PD affected by received optical power and range of receiver from
surface area S. The value of m can be expressed as m = − ln(θ
ln2
1)
, the transmitter. In other words, RSN depends on the estimated
2 parameter d and this dependence affects the total noise vari-
where θ 21 is defined as the view angle when irradiance is half
ance considerably [21]. As a result, common MLE and CRLB
of the value at 0◦ . When the value of m increases, it means the
formulas in literature are not quite accurate.
value of θ 21 decreases. The signal is more concentrated and the
signal-to-noise ratio (SNR) is higher. Therefore, the accuracy is
higher. ϕcon expresses the field of view (FOV) of the receiver. III. RECEIVER PLANE TILTING TECHNIQUE
Ts (ϕ) and g(ϕ) are the filter transmission coefficient and the This section describes a receiver plane tilting technique to
concentrator gain, respectively. It is assumed for simplicity that improve the VLC performance. Since the light from the line of
Ts (ϕ) = g(ϕ) = 1 [24]. sight is the dominant source of the optical power, the higher will
Although visible light is less vulnerable to multipath issue be obtained if the channel LOS gain increases. Obviously, cosϕ
than radio wave, it is necessary to consider the impact of reflec- reaches its maximum when the receiver plane faces the source.
tions [25]. Relating to multipath transmission, the channel gain In the case where the receiver is not located on the desk right
of VLC is given by H(0) = ηLOS + ηdiff , where ηdiff denotes below the source on the ceiling, especially when the receiver
the channel diffuse gain and can be treated as a constant [26]. is positioned in one of the corners of the room, the maximum
Therefore the received optical power can be summarized by channel DC gain may dramatically be increased as the incidence
4656 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 39, NO. 14, JULY 15, 2021

gain of a particular position can be attained if the value of cos(ϕn )

reaches its maximum; thus βopt = φn = π4 . The electrical SNR
can be expressed in terms of received optical power and noise
2
variance as SNR  (P +P σ2
diff )
[29]. Notice that by increasing the
tilting angle, SNR changes.
Given the fact that small changes of β make minor changes
in P , it is experimentally possible to introduce the suboptimal
1 

tilting angle βsub = min(m 7− 80 π4 h , π2 ) and Psub = P 
3m
.
β=βsub
In other words, Psub is a suboptimal value for P , i.e., Psub ≤
Popt and βsub ≈ βopt . About how βsub was obtained, a detailed
discussion is provided in Section VI. It is worth noting that
without using the applied model of VLC geometry, it is not
possible to determine βsub . In the future works, we would extend
our proposed model to the 3-D case.

IV. MAXIMUM LIKELIHOOD ESTIMATION

In this Section, the estimation of parameter d is calculated
based on MLE. Note that RSS can estimate only one parameter,
so we assume that vertical distance (height) is known which is
Fig. 1. The geometry of the LED source and photo-detector after tilting the a common assumption [26]. To find the MLE for this problem,
receiver plane: (a) Two-dimensional, (b) modelling.
first, we write the probability density function (PDF) of the initial
received optical power as
angle ϕ reduces. However, ϕ is not the only parameter of the  
1 1
VLC system model and d and φ should be considered in order p(x; d) = √ exp − 2 (x − P0 − Pdiff )2 , (8)
to attain maximum LOS gain. As it is shown in figure 1, d and σ0 2π 2σ0
φ are not constant for a given tilting process. After tilting, the where σ0 is the variance of Gaussian noise detected by mobile
new value of cosϕ becomes station (MS) if P = P0 . By using the natural logarithm, we have
cosϕn = cos(φn − β) = cosφn cosβ + sinφn sinβ √ 1
ln p(x; d) = − ln σ0 2π − 2 (x − P0 − Pdiff )2 . (9)
  2σ0
hn n hn n
= cosβ + sinβ = cosβ + sinβ , (4) Differentiating the log-likelihood function, we have
dn dn dn hn
∂σ0 ∂σ0
∂ ln p(x; d)
where denotes the horizontal
√ distance between the transmitter = − ∂d + ∂d3 (x − P0 − Pdiff )2
and receiver and = d2 − h2 . β is the inclination angle which ∂d σ0 σ0
is equal to the tilting angle [28]. After tilting, φn , hn , and n are 1 ∂P0
the new values of φ, h, and  , respectively. Furthermore, dn is + (x − P0 − Pdiff ) , (10)
2 + h2 .
σ02 ∂d
the new value of d and dn = n n
regarding the relationship between the shot noise from received
hn = h − h0 = h − ζsinβ, (5) signal and the received optical power, ∂σ 0
∂d is equal to
where ζ is the length of the receiver, as shown in Figure 1(b), ∂σ0 ∂P0 qB ∂P0
− 12
= − = − ζ(1 − cosβ). (6) = qB σT2 + 2q(P0 + Pdiff )B = .
n 0 ∂d ∂d σ0 ∂d
By these modifications, Equation (2) can be revised as (11)
(m + 1)S Alternatively, by using Equation (11), we can rewrite Equation
P = R p Pt
2π (10) as
(h − ζsinβ)m+1 ∂ ln p(x; d) qB ∂P0 qB ∂P0
×  m+3 = − 2 + 4 (x − P0 − Pdiff )2
2 ∂d σ0 ∂d σ0 ∂d
(h − ζsinβ)2 + ( − ζ(1 − cosβ))2 1 ∂P0
+ (x − P0 − Pdiff ) , (12)
  σ02 ∂d
− ζ(1 − cosβ)
× cosβ + sinβ . (7) and setting it to zero produces
h − ζsinβ
Let βopt be the optimal tilting angle that maximizes Equation (7), − σ02 + x2 − 2x(P0 + Pdiff ) + (P0 + Pdiff )2
which depends only on the vertical distance h and horizontal σ02 σ2
distance . Finding the optimal tilting angle (which requires + x − 0 (P0 + Pdiff ) = 0. (13)
qB qB
computer modeling,) results in maximum available power of the
line of sight (Popt ). For instance, consider ϕn as the only parame- Concerning Equation (13), MLE results in
ter of VLC system and n = hn . In this case, the maximum LOS (P0 + Pdiff )2 − a1 (P0 + Pdiff ) + a2 = 0 (14a)
AMINI et al.: ACCURATE RANGING ALGORITHM BASED ON RECEIVED SIGNAL STRENGTH IN VISIBLE LIGHT COMMUNICATION 4657

σ02 
In the next step of the algorithm, P β0 obtains from the equation
a1 = 2x + (14b)
qB ˆ
of Pβ0 by replacing with . Subsequently, P  β0 is calculated
σ02 to estimate the receiver range. More precisely, the procedure
a2 = x 2 + x − σ02 (14c) is performed over the allowable rangeof to achieve ˆ which
qB
a1 − a2 −4a2
Solving for d produces the two solutions minimizes the absolute value of
β0 1β
0
β0
−P β0 . Ob-
2
 viously, the diffuse optical power causes estimation error as well
a1 ± a21 − 4a2
P0 = as the noise. As mentioned previously, small changes of β make
2 minor changes in P , thereby running Algorithm 1 recursively is

σ02 ± σ04 + (2σ0 qB)2 irrelevant.
=x+ . (15)
2qB
√
σ 2 − σ04 +(2σ0 qB)2 Algorithm 1: Known Vertical Distance.
Obviously the solution x + 0 2qB which compen- 1
3m 
sates for the effect of the channel diffuse gain, is more reliable. β0 = min(m 7− 80 π4 h0 , π2 )
In other words, we choose the solution xβ0 = Pβ0 + Pdiff + w; 
 
σ02 − σ04 + (2σ0 qB)2  
P0 = x + (16a) for (over the allowable range of ) doP β 0 = Pβ 0  end
2qB  ˆ
=
  1 for;   
(m + 1)S hm+1 m+3   
d0 = R p Pt , (16b)   
2π P0 Pβ 0 = P  a 1β 0 = a 1  a 2β 0 = a 2  ;
  
 β=β0 x=xβ0 x=xβ0
2 
where d0 is the initial MLE of d and so 0 = d0 − h2 . Note a1 β − a21 −4a2β

minimize | 0
2
β0 0
−P β0 | and return the
that d0 > 0 for all possible values of x and x2 . By using Equation position.
(11), d0 maximizes the log-likelihood function which is verified
by examining the second derivative,
 2
∂ 2 ln p(x; d) 2(qB)2 ∂P0 qB ∂ 2 P0
2 = 4 − 2
∂d σ0 ∂d σ0 ∂d2 V. CRAMER-RAO LOWER BOUND
 2 It is proved in [21] that the variance of any unbiased estimator
(qB)2 ∂P0
− 4 6 (x − P0 − Pdiff )2 must satisfy
σ0 ∂d 

2qB ∂ 2 P0 1 σ02
− (x − P0 − Pdiff )(
∂P0 2 qB
) + 4 (x − P0 − Pdiff )2 var(d0 ) ≥ , (19)
4 σ02 + 2(qB)2 | ∂P∂d |
0
σ0 ∂d σ0 ∂d2
 2  2 where | ∂P
∂d | refers to absolute value of
0 ∂P0
qB ∂P0 1 ∂P0 ∂d which is given by
− 2 4 (x − P0 − Pdiff ) − 2
σ0 ∂d σ0 ∂d ∂P0 R p Pt S hm+1
=− (m + 1)(m + 3) m+4 . (20)
1 ∂ 2 P0 ∂d 2π d
+ (x − P0 − Pdiff ) . (17) √ 
2 σ02
σ0 ∂d2 1
CRLB = σ2 +2(qB) 2 ∂P0 ≈ ∂Pσ0
0
, with respect to Equa-
0 | ∂d | | ∂d |
In fact, it must be proved that the second-order differentiation is tion (19).
negative for all values of d. For large values of the SNR, it can Now, we extend the results of [21] to a receiver case with
be approximated by tilting plane. 
Similar to Equation (19), for the tilting technique
  2  
∂ 2 ln p(x; d) ∂P0 2(qB)2 1 case we have var( ) ≥ σ2 +2(qB) 1 σ2
where the derivation
E =− + 2 . (18) 2
| ∂P
∂ |
∂d2 ∂d σ04 σ0 of the channel DC gain is defined as
As it can be seen, Equation (18) is negative for all values of d. ∂P (m + 1)S
Algorithm 1 represents the MLE for the known height prob-  R p Pt (h − ζsinβ)m+1 (b1 + b2 ) (21a)
∂ 2π
lem, taking advantage of the receiver plane tilting technique
introduced in this article. As the first step of the algorithm, b1 = − (m + 3) ( − ζ(1 − cosβ)) ×
we calculate the initial estimation of distance by Equation  − m+5
2
(16). Then, the initial values of β0 and Pβ0 are calculated by (h − ζsinβ)2 + ( − ζ(1 − cosβ))2 ×
given formulas in Algorithm 1. Note that calculating the initial
 
estimation of the distance is necessary to find the proper tilting − ζ(1 − cosβ)
angle and for simplicity, we employ Equation (16). However, it cosβ + sinβ (21b)
h − ζsinβ
is possible to consider a known initial tilting angle and in this  − m+3
2
way, the designers of a positioning system enjoy the freedom to 2 2
b2 = (h − ζsinβ) + ( − ζ(1 − cosβ)) ×
design the detector according to their own optimization.
4658 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 39, NO. 14, JULY 15, 2021

TABLE I
SOME TYPICAL PARAMETER VALUES

Fig. 2. The optimal tilting angle against its suboptimal one.

 
sinβ
. (21c)
h − ζsinβ
By knowing the CRLB for , we can easily obtain it for d or in
general for any function
√ of . As shown in [31], if it is desired
to estimate d = 2 + h2 , then the CRLB is

2
∂ )
( ∂d
var(d) ≥ , (22)
∂ 2 ln p(x;)
−E ∂2

∂ = d implies that knowing h constrains a minimum

where ∂d 

value for d and subsequently decreases CRLB for d. Since

 
 2
1 d σ
var(d) ≥ . (23)
σ 2 + 2(qB)2 | ∂P
∂ |
Note that by setting β = 0, Equation (23) results in Equation √
Fig. 3. CRLB against the order of Lambertian emission with several values
(19). of horizontal distance and vertical distance.
To study the impact of the Lambertian emission order on
σ0
CRLB, it must be proved that the first-order derivation of ∂P0 | ∂d |
equals to zero if m = mopt and its second-order differentiation is
positive for all m amounts. Finding such an amount is extremely
complicated and requires numerical calculations. In order to
have a closed form expression, we suggest two sub-optimum
scenarios. Firstly, suppose that σSS is negligible.
In this case, it
1 1
is proved in [30] that msub1 = −(2 + lncosφ ) + 1 + ( lncosφ )2 .
In the case that σSS is not negligible,
 it is straightforward to prove
that msub2 = −(2 + lncosφ ) + 89 + ( lncosφ
1.5 1.5
+ 13 )2 . Fig. 4. The performance of the proposed range estimator against the state-of-
the-art estimator with several values of  and h.
VI. NUMERICAL RESULTS AND DISCUSSION
√
In comparison with optimal tilting angle, suboptimal tilting In Fig. 3, the value of CRLB is evaluated by varying the
angle is straightforward to calculate and follows the optimal horizontal
√ and vertical distances. Fig. 3 indicates that values of
curve perfectly as it is shown in Fig. 2 where h = 2.5 (m). The CRLB rise, by increasing horizontal distance
√ for the given
simulations parameters are listed in Table I. Here, the key idea height. Moreover,
√ this figure shows that CRLB is a convex
is that βopt = π4 under the circumstances which are explained in function, i.e. CRLB has a minimum value as it is presented in
section III. Since the tilting angle increases if the receiver is set Table II. Note that msim resulted by simulations
√ is a numerical
in the corner of the room, we would expect that multiplying the value for the amount of m that minimizes CRLB. As it could
parameter h gives a closer approximation to the optimal tilting be seen, msub2 is perfectly matched with msim followed very
angle. Finally, knowing the optimal tilting angle for different well by msub1 .
scenarios make help to experimentally introduce the parameter In Fig. 4, mMLE resulted by simulations is a numerical value
1
3m for the amount of m that minimizes distance estimation error.
m 7− 80 which is the result of curve fitting. Reasonably, the tilting The y-axis of the figures represents the proportion of distance
angle could not be chosen more than 90 degrees and so βsub = estimation error to the true value of distance. Due to properties
1

4 h , 2 ).
3m π  π
min(m 7− 80 of MLE it is expected mMLE to follow the values of msim .
AMINI et al.: ACCURATE RANGING ALGORITHM BASED ON RECEIVED SIGNAL STRENGTH IN VISIBLE LIGHT COMMUNICATION 4659

TABLE II √
THE RESULTS OF LAMBERTIAN EMISSION ORDER FOR MINIMIZING CRLB

Fig. 5. The distance error estimation versus the order of Lambertian emission Fig. 6. The results presentation of height uncertainty where the true value of
with several values of h, where  is fixed at 1.75 (m). vertical distance is h = 1.75 (m) and  is fixed at 2 (m).

By comparing Table II and Fig. 4, one can find a good match in Fig. 5, mopt is about 10. Although the Lambertian model
between mMLE and msim . In this view, it is reasonable to conclude for LED lighting (Equation (1)) is quoted over and over in
that the specific system model utilized in this paper results in literature, it may not properly reflect the beam shaping in a
a more accurate ranging formulas for VLC. Fig. 4 shows that typical LED product. Therefore, concerning the beam shaping,
mMLE is closer to values of msub2 than msub1 . Moreover, the mopt could change. Note that the Lambertian order is an in-
performance gain of the proposed range estimator is evaluated in trinsic property of LEDs and is independent of the receivers
this figure by comparison with the state-of-the-art estimator that position. Once a specific type of LED is chosen, the value of
ignores RSN. The state-of-the-art estimator (which is utilized in the Lambertian order cannot usually be changed dynamically.
previous RSS-based estimation works,) employs P0 = x instead In this view, the distance error estimation is plotted versus the
of Equation (16). As it could be seen, our proposed estimator order of Lambertian emission to just provide more information
has reduced the distance estimation error. All in all, not only about the performance and constraints of the ranging system.
our formulas for CRLB and mopt are more accurate than [30], It is important to mention that in the simulations, it has been
but also considering RSN and its relation to distance parameter tried to cover a wide range (of different types and amounts of
results in a more reliable formula for MLE. parameters) for a light source in order to compare them with
The performance of algorithm 1 is investigated by Fig. 5 each other. In real modes, for large values of , an array of light
with several values of h, where is fixed at 1.75 (m). Once sources have to be installed to ensure light coverage. As a result,
again it is observed that the MLE distance error falls by choos- interference from other LEDs may not be neglected either. With
ing msub2 = 6 where h = 1.75 (m) and by choosing msub2 = 8 regard to σSS = 2q(P + Pdiff )B by increasing the received
where h = 2 (m). In order to observe the efficiency of the re- power which can be the result of interference, the noise related
ceiver plane tilting technique, the receiver is tilted at an angle to the received power also increases.
1

As the next simulation, the results of height uncertainty are
of β0 = min(m 7− 80 π4 h0 , π2 ) towards the center of the LED
3m
presented in Fig. 6 where the true value of vertical distance is
transmitter whereby 0 is itself calculated by MLE estimator. h = 1.75 (m) and is fixed at 2 (m). In this work, we assumed
As it can be observed, the MLE estimator has a low estimation that vertical distance is known. In fact, the measurements may
error without using tilting technique. However, when higher have a degree of error and have not reported, precisely. Fig. 6
accuracy is desired for a specific scenario, it would be preferable demonstrates that the vertical distance uncertainty has the same
to employ the tilting technique as it is described in algorithm 1. effect on range estimation by using the tilting technique and
Furthermore, it is illustrated in Fig. 5 that the distance error without using it.
estimation based on proposed MLEs is a complex function of Finally, the results of the power uncertainty are presented in
h and under certain conditions, it can be reduced by setting the Fig. 7 where h = 1.75 (m) and is fixed at 2 (m). In this work,
receiver farther in the vertical axis. we assumed that transmitted optical power is known. In fact, due
We demonstrated that the value of m changes the distance to subjects such as voltage fluctuations, temperature drifts in the
estimation accuracy and amount of the received optical power. electronics and the LEDs, dust on the fixtures, or ageing of the
This matter itself can affect the capacity and security rate of LEDs the measurements may have a degree of error and have
VLC. Therefore, the results of this article could be used to not reported, precisely. This figure demonstrates that the power
calculate the security rate of the physical layer more accurately. uncertainty has the same effect on range estimation by using the
It is worth mentioning that according to the scenario chosen tilting technique and without using it.
4660 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 39, NO. 14, JULY 15, 2021

[9] N. Stevens, “Bias introduced by true radiation patterns in RSS-based

visible light positioning,” in Proc. 2019 IEEE Sensors, Montreal, QC,
Canada, 2019, pp. 1–4.
[10] A. Sahin, Y. S. Eroglu, I. Guvenc, N. Pala, and M. Yuksel, “Hybrid 3-D
localization for visible light communication systems,” J. Lightw. Technol.,
vol. 33, no. 22, pp. 4589–4599, Nov. 2015.
[11] M. F. Keskin, S. Gezici, and O. Arikan, “Direct and two-step positioning in
visible light systems,” IEEE Trans. Commun., vol. 66, no. 1, pp. 239–254,
Jan. 2018.
[12] C. Amini, A. Taherpour, S. Gazor, and T. Khattab, “Theoretical accuracy
analysis of indoor visible light communication positioning system based on
time-of-arrival,” in Proc. IEEE Can. Conf. Elect. Comput. Eng. (CCECE),
Vancouver, British Columbia, pp. 1–5, 2016.
Fig. 7. The results presentation of the power uncertainty where h = 1.75 (m) [13] B. Zhu, J. Cheng, Y. Wang, J. Yan, and J. Wang, “Three-dimensional
and  is fixed at 2 (m). VLC positioning based on angle difference of arrival with arbitrary tilting
angle of receiver,” IEEE J. Sel. Areas Commun., vol. 36, no. 1, pp. 8–22,
Jan. 2018.
VII. CONCLUSION [14] J. Yan et al., “Error analysis on indoor localization with visible light
communication,” Remote Sens., vol. 11, no. 4, pp. 1–17, Feb. 2019 Art.
In order to maximize received optical power, an optimal and no. E427.
suboptimal tilting angle are proposed in this work. It is showed [15] L. C. Mathias, L. F. De Melo, and T. Abro, “3-D localization with multiple
that the computational complexity of finding suboptimal tilting LEDs lamps in OFDM-VLC system,” IEEE Access, vol. 7, pp. 6249–6261,
Dec. 2018.
angle is low with high accuracy. Considering received shot noise [16] Y. Cai, W. Guan, Y. Wu, C. Xie, Y. Chen, and L. Fang, “Indoor high
and its relation to distance parameter, we proposed an RSS-based precision three-dimensional positioning system based on visible light
ranging in VLC by taking into account the tilting technique. We communication using particle swarm optimization,” IEEE Photon. J.,
vol. 9, no. 6, pp. 1–20, Dec. 2017, Art. no. 7908120.
calculated the corresponding CRLB and MLE. Our simulation [17] S. Arnon, Visible Light Communications. Cambridge, U.K.: Univ. Printing
results showed that the models and proposed estimator have led House, 2015.
to substantially more accurate range estimation. The resulting [18] N. Stevens and H. Steendam, “Influence of transmitter and receiver orien-
tation on the channel gain for RSS ranging-based VLP,” in Proc.11th Int.
distance estimation error for the proposed tilting technique is Symp. Commun. Syst. Netw. Digit. Signal Process. (CSNDSP), pp. 1–5,
about 16 percent lower than the case of not considering the afore- Jul. 2018.
said technique. In addition, we have undertaken a detailed study [19] D. Plets, S. Bastiaens, L. Martens, and W. Joseph, “An analysis of the
impact of led tilt on visible light positioning accuracy,” Electronics, vol. 8,
of the effect of Lambertian emission order on the CRLB since no. 4, 2019. [Online]. Available: [Link]
the estimator can be improved by using optimum Lambertian 389
emission order. DC intensities can very well be measured by a [20] S. Shen, S. Li, and H. Steendam, “Simultaneous position and orientation
estimation for visible light systems with multiple LEDs and multiple PDs,”
cheap camera, which also can measure angles of arrival. Thus a IEEE J. Sel. Areas Commun., vol. 38, no. 8, pp. 1866–1879, Aug. 2020,
camera would deliver an accurate extra input, and angles are not doi: 10.1109/JSAC.2020.3000805.
subject to intensity fluctuations. In our future works, we would [21] C. Amini, P. Azmi, and S. S. Kashef, “Theoretical accuracy analy-
sis of RSS-based range estimation for visible light communication,”
combine AOA and RSS information in order to reach a hybrid 3- arXiv:2011.14080, vol. 1, pp. 1–14, 2020.
D localization for VLC systems based on the results of this paper. [22] S. M. Moser, “Capacity results of an optical intensity channel with
input-dependent Gaussian noise,” IEEE Trans. Inf. Theory, vol. 58, no. 1,
pp. 207–223, Jan. 2012.
REFERENCES [23] T. Komine and M. Nakagawa, “Fundamental analysis for visible-light com-
munication system using LED lights,” Consum. Electron. IEEE, vol. 50,
[1] T.-H. Do and M. Yoo, “An in-depth survey of visible light communication
no. 1, pp. 100–107, 2004.
based positioning systems,” Sensors, vol. 16, no. 5, pp. 1–40, 2016, Art.
[24] J. Gancarz, H. Elgala, and T. D. C. Little, “Impact of lighting requirement
no. E678.
on VLC systems,” IEEE Commun. Mag., vol. 51, no. 12, pp. 34–41,
[2] X. Liu, X. Wei, L. Guo, Y. Liu, Q. Song, and A. Jamalipour, “Turning the
Dec. 2013.
signal interference into benefits: Towards indoor self-powered visible light
[25] C. Amini and A. Taherpour, “Adaptive threshold device for detection of
communication for IoT devices in industrial radio-hostile environments,”
reflections based visible light communication,” Opt. Commun., vol. 388,
IEEE Access, vol. 7, pp. 24978–24989, Feb. 2019.
pp. 1–11, 2017.
[3] H. Kim, D. Kim, S. Yang, Y. Son, and S. Han, “An indoor visible
[26] X. Zhang, J. Duan, Y. Fu, and A. Shi, “Theoretical accuracy analysis
light communication positioning system using a RF carrier allocation
of indoor visible light communication positioning system based on re-
technique,” J. Lightw. Technol., vol. 31, no. 1, pp. 134–144, Jan. 2013.
ceived signal strength indicator,” J. Lightw. Technol., vol. 32, no. 21,
[4] E.-M. Jeong, S.-H. Yang, H.-S. Kim, and S.-K. Han, “Tilted receiver
pp. 4180–4186, Nov. 2014.
angle error compensated indoor positioning system based on visible light
[27] T. Komine and M. Nakagawa, “Fundamental analysis for visible-light
communication,” Electron. Lett., vol. 49, no. 14, pp. 890–892, Jul. 2013.
communication system using LED lights,” IEEE Trans. Consum. Electron.,
[5] G. Shi et al., “Accuracy analysis of indoor visible light communication
vol. 50, no. 1, pp. 100–107, Feb. 2004.
localization system based on received signal strength in non-line-of-sight
[28] C. H. Edwards and D. E. Penney, Calculus. 6th Ed. Toronto, Canada:
environments by using least squares method,” Opt. Eng., vol. 58, no. 5,
Pearson Educ., 2002.
pp. 1–11, May 2019.
[29] Z. Ghassemlooy, W. Popoola, and S. Rajbhandari, Optical Wireless Com-
[6] H. Steendam, T. Q. Wang, and J. Armstrong, “Theoretical lower bound for
munications: System and Channel Modelling With MATLAB, 2nd ed. Boca
indoor visible light positioning using received signal strength measure-
Raton, FL, USA: CRC Press, May 2019.
ments and an aperture-based receiver,” J. Lightw. Technol., vol. 35, no. 2,
[30] C. Amini, A. Taherpour, T. Khattab, and S. Gazor, “On the more accurate
pp. 309–319, Jan. 2017.
channel model and positioning based on time-of-arrival for visible light
[7] E. Gonendik and S. Gezici, “Fundamental limits on RSS based range
localization,” Opt. Eng., vol. 56, no. 1, pp. 1–12, 2017. [Online]. Available:
estimation in visible light positioning systems,” IEEE Commun. Lett.,
[Link]
vol. 19, no. 12, pp. 2138–2141, Dec. 2015.
[31] M. Steven Kay, Fundamentals of Statistical Signal Processing, Vol. I,
[8] N. Stevens and H. Steendam, “Magnitude of the distance estimation bias
Estimation Theory. New Jersey, NJ, USA: Prentice Hall, 1993.
in received signal strength visible light positioning,” IEEE Commun. Lett.,
vol. 22, no. 11, pp. 2250–2253, Nov. 2018.