IEEE SENSORS JOURNAL, VOL. 22, NO.

18, 15 SEPTEMBER 2022 17497

Design and Testing of Automated Smoke

Monitoring Sensors in Vehicles
Pravin R. Kshirsagar , Member, IEEE, Hariprasath Manoharan , Fadi Al-Turjman , Member, IEEE,
and Kailash Kumar Maheshwari , Member, IEEE

1 Abstract —This article emphases an advanced method for

2 controlling the pollution of vehicles through sensor integra-
3 tion process. All sensor nodes will be connected to a con-
4 trol centre and it will be monitored using Online Monitoring
5 System (OMS) with the help of an Improved Smoke Detection
6 Algorithm (ISDA). The purpose of using ISDA is to monitor the
7 vehicles which are located in different areas under different
8 temperatures where, the ISDA divides the regions into circles.
9 The data obtained in OMS will be aggregated and it will
10 be plotted in MATLAB for better understanding. To test the
11 efficiency of the projected method three test cases which
12 include control of energy, minimization of range and cot have
13 been performed. It is observed that in all three cases ISDA
14 proves to be more efficient in terms of energy consumption, cost and distance when compared with existing methods.
15 Index Terms — Automated vehicles, energy consumption, pollution, range, sensors.

16 I. I NTRODUCTION vehicles which are producing more smoke than the threshold 29

values at a particular place without causing any accidents.

I N RECENT days, there is a huge necessity of wireless
17 30

18 devices to monitor the daily needs of human. Even the The prevention can be done in two possible ways (i) The 31

19 applications of wireless devices have been growing and it is driver must install the sensor and should be able to monitor the 32

20 easy to implement in vehicular systems also. Most of the urban value of smoke in their vehicles by their own (ii) A detailed 33

21 areas are facing the problem of environmental pollution due information report will be sent to the vehicular department 34

22 to the smoke that is emerging from vehicles. Even a lot of in the concerned areas and they should stop the vehicles 35

23 people are unable to breathe due to the presence of smoke automatically where, the safety of the individual should also 36

24 in the atmosphere. Most of the people are still using old be safeguarded. Both ways can be implemented in real time 37

25 vehicles which are producing more smoke than estimated and but more specifically second method can be preferred because 38

26 even the change in type of engines also does not provide any a detailed report will always be maintained by department offi- 39

27 satisfactory solutions. The only way to overcome from the cials and necessary actions can be taken at prescribed times. 40

28 threat that is caused by the smoke from vehicles is to stop those Since, the processor to be used is a high speed processor the 41

values acquired from the sensor will be stored and displayed 42

on an Online Monitoring System (OMS). Therefore, it is 43

Manuscript received 22 November 2020; revised 10 December 2020;
accepted 10 December 2020. Date of publication 14 December 2020; possible to monitor the vehicles in a continuous manner and 44

date of current version 14 September 2022. The associate editor coor- the life time security of each individual will be assured by 45
dinating the review of this article and approving it for publication was this automated systems. Figure 1 shows the implementation 46
Dr. Yin Zhang. (Corresponding author: Pravin R. Kshirsagar.)
Pravin R. Kshirsagar is with the Department of Electronics and of proposed method. 47

Communication Engineering, AVNIET, Hyderabad 501510, India (e-mail: An Up-to date information about traditional methods of 48
[email protected]). pollution monitoring in vehicles have also been provided [1] 49
Hariprasath Manoharan is with the Audisankara College of
Engineering and Technology, Gudur 524101, India (e-mail: where, all different characteristics has been analyzed with 50

[email protected]). their own benefits and difficulties. These characteristic changes 51

Fadi Al-Turjman is with the Research Center for Artificial Intel- will have more update about wireless networks in vehicular 52
ligence (AI) and Internet of Things (IoT), Department of Artificial
Intelligence, Near East University, 99138 Istanbul, Turkey (e-mail: applications for monitoring the contamination level. The level 53

[email protected]). information of information provided in [1] is more insistent 54

Kailash Kumar Maheshwari is with the College of Computing and and it also provides a better way for enhancing the vehic- 55
Informatics, Saudi Electronic University, Riyadh 11673, Saudi Arabia
(e-mail: [email protected]). ular applications. Also, a sustainable monitoring process is 56

Digital Object Identifier 10.1109/JSEN.2020.3044604 needed for enrichment of all public transportation system 57

1558-1748 © 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

thorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on December 26,2024 at 03:32:37 UTC from IEEE Xplore. Restrictions app
 17498 IEEE SENSORS JOURNAL, VOL. 22, NO. 18, 15 SEPTEMBER 2022

been incorporated in the projected formulations. The major 93

drawback that has been observed in [6] is that the source 94

of smoke cannot be identified and also it is not possible 95

to send the information that is carried by the sensors in 96

a combined way because the authors have conveyed that 97

each different sensor will be located at different positions. 98

To overcome the aforesaid drawback a Zigbee based module 99

has been developed where, data visualization decision has been 100

provided. In addition, the online monitoring system is also 101

used because the main motive is to detect the fire that has 102

happened in forest areas. The authors have explained the use 103

of monitoring system by integrating random forest algorithm 104

but fail to include the placement cost of sensors. Also, the area 105

of forest will be very large and it is difficult to place the 106

sensors in a particular location and to aggregate the data 107

in a synchronized manner. The authors have considered the 108

implementation method only for large area of smoke detection 109

which is explained with basic formulations. In [7] the authors 110

Fig. 1. Operation configuration of the proposed method.

have conversed that the safety will be improved but the values 111

during data processing is not much secured which leads to 112

failure of the process and a detailed security measurements 113

58 and to reduce the hazards of environmental pollution [2]. are needed. 114

59 More flexible directions with low emission at high traffic The aforementioned works can be applied in all outdoor 115

60 congestion can be provided to all vehicles by following the environments and to larger areas. In line with the above con- 116

61 strategy of distributed sensors. The same distribution method cern, an application for vehicles using sensor technology has 117

62 is followed in the proposed method but instead of random been initiated. The prime objective is to reduce the accident of 118

63 distribution the area is divided in circles. Even it is possible to vehicles by placing a highly reliable sensor in any part of the 119

64 monitor vehicles using stationary sensors where more quality vehicle where, a mobile environment will be considered for 120

65 of communication will be present at low cost [3]. But the installing the smoke sensors. It is noted that in [8], the major 121

66 areas cannot be divided and ranges to particular areas cannot problem is that the sensors can never be calibrated and it will 122

67 be allocated properly. not work together when installed in all parts of the vehicles. 123

Even if the sensors are installed at different places in the 124

vehicle, the life time and the capacity will be decreased until 125

68 A. Related Works proper calibration is made. Also, the security an assessment 126

69 Some existing literatures on Internet of Things (IoTs) have which includes the mobile number of the vendors has never 127

70 been analyzed for monitoring the smoke that has been released been discussed. A camera based detection system has evolved 128

71 by the vehicles. Among those literatures an embedded technol- for capturing the exact image of the location when any natural 129

72 ogy which is designed to monitor the air quality index has been phenomenon exists. This technique uses the IP address of a 130

73 considered for validating the proposed formulations. In [4] particular location where, the same disadvantage of providing 131

74 the fundamental concept regarding the health of environment security has been neglected. If security enhancement is pro- 132

75 is discussed for one particular area whereas, the authors fail vided then, with high end processor it is very easy to detect 133

76 to extend the proposed method to all other remaining areas. the smoke with the help of an image and the workers can 134

77 Therefore, the aforementioned method does not suit well when immediately go to the desired place. Another disadvantage 135

78 the vehicles travels to different areas at a particular time is that there is no latitude and longitude positions which are 136

79 period. Another method with basic model formulations has attached with the image which makes the worker to identify 137

80 been deliberated for monitoring the detectors response. For the desired location within a short span of time [9]. 138

81 the purpose of data aggregation the role of detector is much Similar to [9] a smart sensor system for two wheelers has 139

82 essential therefore, a time index model has been designed been developed by keeping different mechanisms in obser- 140

83 where the threshold values are determined. But, the method vance. The authors in [10] have stated that most of the world 141

84 uses only a particular zone model which cannot be compared population is using two wheelers and there is a need to monitor 142

85 and obtained for different zones [5]. However, the model can the environmental pollution which acts as a major threat to 143

86 be extended to different zones by using an online monitoring all humans. So an automated system has been developed but 144

87 system which forms the base of the projected work. this development has been directly installed in the engine of 145

88 An alarm technique with an intelligent framework has been vehicles which creates complexity for the driver of a particular 146

89 proposed in [6]. The vendor can be able to directly get the vehicle because if the bike does not ignite properly then it 147

90 required sensor output values by using the douser framework. is very difficult to find which affects the real time situation. 148

91 But, this method forms the base on detecting the smoke Therefore this automated mechanism is very difficult to be 149

92 caused by fire where, some useful threshold values have implemented in real time and it is very costly. To reduce the 150

thorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on December 26,2024 at 03:32:37 UTC from IEEE Xplore. Restrictions app
 R. KSHIRSAGAR et al.: DESIGN AND TESTING OF AUTOMATED SMOKE MONITORING SENSORS IN VEHICLES 17499

151 critical issue on life time of humans a smart home monitoring implementation where even installation of GPS based sys- 207

152 system has been developed for smoke detection in case of tems with LoRA technique has been discussed [18]. But 208

153 occurrence of fire inside the home. A unique fire dynamic when compared to sensor devices GPS will be much costly 209

154 simulator has been used for the purpose of detecting smoke. therefore a decentralized remote monitoring system has been 210

155 However, this method provides high consumption of energy developed which is expected to yield precise results at low 211

156 when the nodes are installed at different places [11]. A visual cost [19]. Even though the expected results are achieved 212

157 sensor network at a low cost has also been implemented for consumption of energy is much higher. Therefore, the authors 213

158 increasing the accuracy of measurement in overlap areas. This in [20]–[24] have made a survey for addressing the gaps that 214

159 method provides a high quality image where, the workers are present for low quality solutions and to reduce the cost of 215

160 can exactly predict the cause of smoke incidence. But if implementation where, vehicular networks are integrated with 216

161 such overlapping methods are used then, a high quality pi Internet of Things for providing high safety with consistent 217

162 camera is necessary which is very expensive for a small operation [21]. Also a new biological application platform is 218

163 home. Therefore, this method [12] is not suitable for all formed for ignoring the normal operation in machine learning 219

164 indoor and outdoor environments. Apart from these sensor procedures [22] to perform classification and for improving 220

165 based technologies [1]–[12], a new cognitive radio method the performance of vehicular operation. Contrariwise, for 221

166 has been implemented for monitoring the vehicles under directing the humans through desired direction an artificial 222

167 crucial conditions. All the discussed sensor methodologies will intelligence technique is formed [23] where real time data is 223

168 have mobility issues which can be solved by using cognitive sensed and reported to control center. The authors in [24] have 224

169 networks [13]. Even though the mobility issues are solved discussed that offline sharing of data is also possible by using 225

170 by cognitive methods, the number of required users at both a migration mechanism. But in all cases the implementation 226

171 transmission and reception end will be very much higher cost or energy consumed by nodes will be much higher. 227

172 which is observed as a major drawback when compared to the Henceforth to overcome the aforementioned drawbacks top 228

173 solutions provided by the sensors. All the necessary solutions feature sensors have been calibrated and installed in vehicles. 229

174 for implementing cognitive networks have been described In addition, a gap has been assembled by dividing the regions 230

175 in [14]. But the cognitive radio test bed provides low quality in different groups which forms the circle. Therefore, under 231

176 solutions and the major parameter which is defined as fast different temperatures the smoke created by the vehicles can 232

177 information processing has been missing in both methods be monitored and the distance covered by the sensors can also 233

178 that are implemented in [13], [14] when compared to the be maximized that covers the entire area in the circle at low 234

179 methodologies discussed above. cost. 235

180 Even though the sensors are calibrated and designed accord-
181 ing to the requirements it is much important to calculate the
C. Research Objectives 236
182 number of nodes which has been well-defined in [15]. A pre-
183 defined understanding of the relay nodes is much essential The prime objective of the proposed method is to reduce 237

184 for sensor installation and if the nodes are reduced then, the smoke that is caused by the vehicles. The reason behind 238

185 the entire cost of installation can be reduced to a much higher choosing this objective is that the smoke which is exhausted 239

186 extent. This survey helps to design the appropriate relays for by the vehicles is observed as one of the major threat of 240

187 this projected work. Also, another method for optimal sensor pollution in urban areas. Therefore, the objective can be 241

188 placement has been designed for positioning the nodes are solved by integrating a sensor technology with an efficient 242

189 proper locations. These two methods [15], [16] if incorporated ISDA algorithm which produces substantial results in reducing 243

190 then, it will be more useful for choosing the optimal location the smoke of vehicles under different temperatures. Further, 244

191 and nodes which in turn reduces the cost of installation with the objective is also solved by minimizing the cost of sensor 245

192 huge benefits. A global survey from International council of implementation and energy consumption with maximization 246

193 green transportation [17] has reported that in 2017 nearly of distance. 247

194 1 billion people suffered due to various diseases and this

195 transpired due to the presence of air pollution. Since the II. P ROBLEM F ORMULATION 248

196 numbers of vehicles are grown rapidly during the years in 2020 Since, the smoke which is produced by the vehicle has to be 249

197 nearly 1.9 billion people have suffered huge loss due to air monitored it needs to be compared with maximum amount of 250

198 pollution which is emitted by vehicles. Many people cannot air quality index which is prevailing in the environment. If the 251

199 able to inhale the natural air and it is very difficult for drivers smoke that is produced by the vehicles crosses the threshold 252

200 during smog spell. Even many ozone attributes are at the verge limit of environment then, it an alarm will be indicated where, 253

201 of vanishing due to smoke created by vehicle. Also the amount the vehicles will be stopped. This formulation of air quality 254

202 of emission crosses nearly 400,000 PM2 which is much higher index will use the resistance of sensor for exact prediction 255

203 than expected emissions. which is given by the formula, 256

Ci × l
Ri = − li (1) 257
204 B. Research Gap and Motivation Oi
205 All the existing methods that are discussed under literature where, Ci represents the maximum input voltage to the air 258

206 works grieves from any one drawbacks like high cost of quality sensor l and li denotes the maximum load offered to a 259

thorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on December 26,2024 at 03:32:37 UTC from IEEE Xplore. Restrictions app
 17500 IEEE SENSORS JOURNAL, VOL. 22, NO. 18, 15 SEPTEMBER 2022

260 distinct sensor and i t h sensors Oi signifies the output voltage If the energy consumption of all nodes is reduced then, 310

261 that is obtained from each sensor after each iterations the transmitter can be fitted in the vehicles suitably and the 311

262 The resistance of sensor is a basic parameter that needs receiver can be fitted at the control station where, the corre- 312

263 to be calculated at preliminary stage. Once the threshold sponding departments are available. So to properly attach the 313

264 values are established then, should never be changed. So while transmitter and receiver it is necessary to maximize the range 314

265 designing, the output voltage during each iteration has to of sensors which can be given from Equation (7). 315

266 be carefully calculated. In the next step the concentration of 

267 carbon-di-oxide needs to be monitored where it can be given Ri = 1.2 πr 2 (7) 316

268 mathematically as, Since, there is an option for dividing the areas in a circular 317

1/emini way the range will be calculated by using the formula of 318
269 Conci = (2) area. Once maximum distance is achieved by the sensor then,
ai v i 319

the cost of placement of sensors will be minimized which is 320

270 where, emini denotes the minimum emission of CO2 content
given in Equation (8). 321
271 ai represents the exchange of smoke content from vehicles in
272 air v i signifies the volume of outdoor environment. 
n

273 From Equations (1) and (2) the concentration content in Mi n costi = Ri x i (8) 322

274 air can be found out. Then, the temperature will be calculated i=1

275 because during summer and winter seasons the air content will where, x i represents the weight of individual sensors 323

276 vary so three different values (Including fog season) needs to From the above Equations (1-8) the objective function can 324

277 provided as threshold levels to the Online Monitoring System be formulated as, 325

278 (OMS). Since the OMS is used the output temperature can be 
n 
n
279 calculated by using Equation (3), O Bi = mi n E i , cost i &max Ri (9) 326

Tmax i=1 i=1

280 Ti = (3) If the above objective function is solved then, the data
C Ti 327

aggregation will be properly achieved between transmitter and 328

281 where, Tmax represents the maximum amount of outdoor
receiver with minimized energy, cost and maximized coverage 329
282 temperature C Ti denotes the critical temperature that can be
range. 330
283 tolerated in the implemented sensor
284 The role of maximum outdoor temperature is much impor-
285 tant because mostly under humidity conditions the smoke III. O PTIMIZATION A LGORITHM 331

286 released by the vehicle will never dissipate which causes the It is necessary to integrate the proposed formulation 332

287 emergent of severe breathing problem to all humans. There- with an efficient algorithm for making the entire system 333

288 fore, under humidity condition which occurs during winter and to work properly. Therefore, an Improved Smoke Detection 334

289 fog seasons the designed OMS will continuously monitor the Algorithm (ISDA) with deep learning technique [25] has 335

290 maximum value and it will compare it with critical values. been implemented for detecting the smoke of vehicles. This 336

291 Even if the area has more industries then, the presence of dust algorithm will provide effective solutions during different tem- 337

292 particles also plays a major role. Therefore the humidity and peratures to an OMS. The activation function of the projected 338

293 dust particles in the air can be monitored using Equations (4) ISDA can be given as, 339

294 and (5). 

n

273K AF(i ) = S(i )max(0, L(i )) (10) 340

295 Hi = (4) i=1

log10 Pi
where, S(i ) represents the angle of sensor placement L(i ) 341
296 where, Pi denotes the pressure of outdoor vapor denotes the optimal location of sensors inside the vehicle 342
297 The dust particle can be calculated as, The performance evaluation of ISDA which is expressed 343

Ri in percentage is very much higher when it is used for the

Di =
344
298 (5)
SVi detection of smoke. This method is compared with [12] where, 345

299 where, Ri represents the control value of each sensors SVi no deep learning algorithm is used. For both the methods the 346

300 denotes the sample value of all the installed sensors performance can be evaluated using the following formula. 347

301 The next parameter to be considered for calculating the 

n
302 vehicle smoke consumption is the energy consumption of the P(i )(%) = o(i )/T (i ) × 100 (11) 348

303 connected nodes in the vehicles. The energy consumption by i=1

304 both transmitter and receiver should be always lesser which where, o(i ) denotes the number of positively detected exces- 349

305 can be established by Equation (6). sive smoke vehicles T (i ) represents the total number of 350

vehicles tested
E i = (E max + E min ) − β(i )− ∝ (i )
351
306 (6)
From Equation (11), it can be seen that the ISDA will 352

307 where, E max and E min denotes the maximum and minimum check the number of smoke samples at a higher speed when 353

308 energy β and ∝ represents the degenerated and amplified compared to [12] and the number of positive samples will be 354

309 energies accurate which makes the performance of ISDA to achieve 355

thorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on December 26,2024 at 03:32:37 UTC from IEEE Xplore. Restrictions app
 R. KSHIRSAGAR et al.: DESIGN AND TESTING OF AUTOMATED SMOKE MONITORING SENSORS IN VEHICLES 17501

356 a rate of 76.8% when compared to [12] which produces

357 only 64.2%. This evaluation has been made this algorithm
358 unique from all smoke detection vehicles. For the implemented
359 algorithm, it is also necessary to detect the number of negative
360 samples so that the misunderstanding in false detection of
361 smoke vehicles can be avoided in case if there is any fault
362 in the sensor.
363 F(i )(%) = n(i) /T (i ) × 100 (12)
364 where, n(i ) denotes the number of negatively detected smoke
365 vehicles that occurred during failure of sensors
366 After a careful check the number of false detection which
367 is calculated from Equation (12) for ISDA is only 1.34%
368 but for [12] it is much higher which resides within 3.7%.
369 Therefore, as a result the accuracy rate can be calculated as a
370 sum of positive and negative samples which can be given as,
o(i ) + n(i )
371 Accur acy(%) = × 100 (13)
T (i )
372 The major reason for employing ISDA with projected model
373 is that sensors are assimilated for observing runway vehicles
374 where, more number of data will be surveyed in everyday and
375 effectiveness of data processing will be improved only when
376 ISDA is integrated. In addition, decision on monitoring smoke
377 data from vehicles will be improved at much higher extent
378 and ISDA is having the ability on improving the efficiency of
379 overall process within a short span of time. Also, it is found Fig. 2. Flow chart of ISDA.
380 from Equation (13) that the accuracy remains very close to
381 98.5% when ISDA is used. This is taken as one of the best
projected method cluster complexity is much lesser due to the 405
382 advantage that is provided by deep learning algorithms. The
high stand still distance and cluster areas. 406
383 flow chart of ISDA is explicated in Figure 2.

B. Step-by-Step Implementation of ISDA 407

384 A. Computational Complexity of ISDA
Step 1: Initialize the calibrated sensor data according to 408
385 In the proposed method computational complexity of ISDA
capability of vehicles. 409
386 is calculated by taking positive samples of corresponding
Step 2: Identify maximum threshold value (Strength) of 410
387 objects. Since, the level of smoke is calculated using maximum
sensors. 411
388 capability of vehicles it is always necessary that Ri and
Step 3: Integrate the objective function from Equation (9) 412
389 Hi should be positive constants. Both Ri and Hi should
and monitor precise measurement values. 413
390 not be inclined by any input parameters. Therefore, ISDA
Step 4: Calculate positive and negative sample values and 414
391 computational properties will be totally calculated using big-O
categorize the successive samples. 415
392 notation which can be given as,
Step 5: Direct successive to online monitoring system and 416

393 f (i ) = Og(i ) (14) check for normalized smoke value. 417

Step 6: If the normalized value is accurate then, go to step 8. 418

394 Equation (14) defines the upper bound of the function Otherwise go to step 7. 419
395 where, complexity analysis of ISDA will be under complex Step 7: Observe all outdoor environments and check the 420
396 model that should satisfy the following constraint, false detection values from Equation (12). 421
⎧
⎪ Step 8: Resurrect the loop and go to step 5
⎨ Ri ≥ 0
422

Step 9: If normalized values are accurate then stop the loop.

og(i ) = 1 i f Hi ≥ 0
423
397 (15)
⎪
⎩ Otherwise go to step 2. 424
0 ≤ f (i ) ≤ Ri g(i ) Step 10: Stop the iteration and print monitored values. 425

398 From Equation (15) it can be implied that gi is the asymp-

399 totic upper bound for the given function. To compute the IV. R ESULTS AND D ISCUSSIONS 426

400 complexity analysis for ISDA in corresponding vehicles a The proposed method using ISDA has been evaluated using 427

401 complete data set should be defined clearly then similarity an OMS where, all the values that have been attained are 428

402 matrix should be calculated using clustering functions which plotted in MATLAB. To check the efficiency of the projected 429

403 is denoted by n. Since, number of iterations is larger initial smoke detection methods using ISDA the following test cases 430

404 cluster center will be defined during each time period. For the have been performed. 431

thorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on December 26,2024 at 03:32:37 UTC from IEEE Xplore. Restrictions app
 17502 IEEE SENSORS JOURNAL, VOL. 22, NO. 18, 15 SEPTEMBER 2022

Fig. 4. Comparison of node energy consumption.

Fig. 3. Percentage of observed Positive and Negative samples.

432 Case 1: Consumption of energy

433 Case 2: Maximization of range (Circle analysis)
434 Case 3: Minimization of cost
435 Before discussing about the case studies it is important Fig. 5. Comparison of distance covered in circles.
436 to note the number of positive and negative samples that
437 has been obtained at different temperatures. It is required
438 that minimum positive samples should be obtained on an energy consumption of the sensor nodes will be 30 whereas, 464

439 average of about 60-70%. If the samples are not obtained for the existing method [12] it is found to be 45 Joules per bit. 465

440 within the aforementioned rate then, all the case studies will Similarly if data transfer rate is increased to an extent 466

441 result in error and it will provide conflicting results such as of 500 Mbps the energy consumption between proposed and 467

442 maximization of energy, cost etc. So to avoid the errors in existing method is higher where for existing method it is 468

443 measurement the number of positive and negative samples has observed to be 36 Joules per bit but for proposed method 469

444 been observed in OMS and is shown in Figure 3. it is equal to 25 Joules per bit. Thus energy is also saved 470

445 From Figure 3, it can be observed that the positive samples when proposed algorithm is integrated. It is always indorsed 471

446 are more where, successive samples of smoke in all vehicles that if a system produces low energy level at high data rate 472

447 have been obtained for testing. Now after a careful investiga- then in future it provides more support for monitoring battery 473

448 tion, the case studies needs to be performed for validating the operated vehicles. This proves that ISDA is more efficient for 474

449 proposed formulations with ISDA. monitoring the smoke produced by vehicles at a low energy 475

450 Case 1: The pre-installed nodes at both transmitter and rate. 476

451 receiver should be able to minimize the energy which is Case 2: In this case the most important parameter which is 477

452 calculated using Equation (6). The amplified energy of the denoted as range (Ri ) covered by the sensors is determined. 478

453 nodes plays an important role in this case. If there is any need For an efficient working sensor, the distance should be as long 479

454 of sensor where, energy is maximum then the corresponding as possible. This range is calculated using Equation (7) where, 480

455 energies can be converted to amplified form. In this way the area of circle of different zones has been taken under 481

456 the consumption of energy by the sensors can be minimized. consideration. The areas are divided into necessary regions 482

457 Figure 4 shows the consumption of energy by sensor nodes which are covered in a larger circle. Therefore, adequate 483

458 when the data is transferred to the receiver (per second). number of sensors will be placed in the regions that are divided 484

459 From Figure 4, it can be seen that for the maximum data as circles. Figure 5 shows the distance covered by each sensor 485

460 transfer rate which is varying between 100- 500 Megabits and it varies according to the divided circles. 486

461 per second (Mbps), the energy consumption for the proposed From Figure 5, it can be seen that a maximum of 100 sen- 487

462 ISDA method will be varying on an average of 32 Joules sors will be placed and the region circles will also be divided 488

463 per bit. For example, for the transfer rate of 300 Mbps the accordingly. For the existing method [12], the range covered 489

thorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on December 26,2024 at 03:32:37 UTC from IEEE Xplore. Restrictions app
 R. KSHIRSAGAR et al.: DESIGN AND TESTING OF AUTOMATED SMOKE MONITORING SENSORS IN VEHICLES 17503

Fig. 6. Calculation of cost ($). Fig. 7. Trailing rate of ISDA.

490 by each sensor is very lesser when compared to proposed

491 ISDA method because the area has not been divided in
492 circles. For example, if the number of sensors is 80, then
493 the distance covered by the existing method [12] is approxi-
494 mately 10 Kilometers whereas, the proposed method will cover
495 nearly 13 Kilometers. The range of Kilometers will be highly
496 augmented if the numbers of sensors are increased. This can
497 be proved by considering the case when number of sensors
498 is equal to 100 where, the covered range is much higher
499 for proposed method which is equal to 19 Kilometers but
500 the existing method can able to cover only 14 Kilometers.
501 If the range covered by large sensors is minimum then in
502 future it cannot be installed on highways where length of
503 transportations is much extensive. Also the installed large Fig. 8. Robustness characteristics of ISDA.
504 number of sensors will provide high resolution for larger
505 images. This shows that the ISDA performs well and it covers
506 a larger range than the existing method [12]. of projected sensor based technical method. Therefore, 531
507 Case 3: This case study depends on the previous case that is two basic parametric evaluations such as capacity of 532
508 discussed above where, if the distance is higher then, the cost trailing and associated robustness of proposed ISDA is 533
509 of sensor to be installed will be very much less. It should calculated. 534
510 always be typical that even for autonomous driving of vehicles Figure 7 shows the trailing rate of ISDA where numbers 535
511 the cost of installation should be much lesser with long ranges. of iterations are considered to be 100. It can be observed 536
512 Therefore, in this case maximum of 100 sensors will be taken from Figure 7 that at starting when number of sensors is low 537
513 under consideration. The cost of each sensor is calculated the proposed algorithm cannot able to converge at a faster 538
514 using Equation (8). After implementing the corresponding rate. But if number of sensors is high then capacity of trailing 539
515 equation the cost values are obtained in OMS and are plotted starts decreasing to original level and finally a constant rate 540
516 in Figure 6. is reached. This constant rate is also achieved at iteration 541
517 From Figure 6, it can be seen that the cost implemented between 30 and 35 where unnecessary wastages are dropped. 542
518 for the proposed method is always lesser. For example, if the Therefore this performance metric proves that the proposed 543
519 number of sensors are 60 then, the cost of implementation method can able to hold maximum capacity at minimum time 544
520 will be 134.5$ whereas, for the existing method [12] with periods where effective operation is guaranteed. Similarly the 545
521 same number of sensors the cost will be around 139.7$. proposed algorithm can be verified by using another metric 546
522 Even if number of sensors are much higher which is equal which is termed as robustness. Robustness denotes the strength 547
523 to 100 the proposed method utilizes much less cost where it of the proposed algorithm where examination of all parametric 548
524 can be observed as 221$. If installed sensors grow rapidly it values will be considered. 549
525 is possible to achieve low cost by connecting infrared image Figure 8 shows the robustness characteristics of proposed 550
526 sensors in highways. This proves that the ISDA is more ISDA which is observed for maximum of 30 trial runs. 551
527 efficient for implementation in real time at reduced cost. It can be observed in Figure 8 that base strength of ISDA 552

is 180 millilitres where after integrating the proposed for- 553

528 A. Performance Analysis of Proposed Method mulations the strength varies between 174 to 180 millil- 554

529 In addition to all case studies discussed above it itres. If the corresponding level goes below 174 millilitres 555

530 is much important to determine performance analysis then an alarm sound will be produced inside the vehicle 556

thorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on December 26,2024 at 03:32:37 UTC from IEEE Xplore. Restrictions app
 17504 IEEE SENSORS JOURNAL, VOL. 22, NO. 18, 15 SEPTEMBER 2022

557 which indicates the strength of monitoring process is lesser. [6] L. Andrea, R. Abirami, M. Diviya, and J. S. Nancy, “Framework for fire 605

558 Conversely, the proposed method achieves maximum strength detection and mitigation using IoT,” Int. J. Pure Appl. Math., vol. 118, 606
no. 18, pp. 1801–1811, 2018. 607
559 of 180 within 15 trials where more efficient monitoring process [7] Q. Wu et al., “Intelligent smoke alarm system with wireless sensor 608
560 is achieved. Also the data strength is checked for same number network using ZigBee,” Wireless Commun. Mobile Comput., vol. 2018, 609

561 of trials and it has been observed that within short run the no. 2, pp. 1–11, 2018. 610
[8] K. Arjun, P. Prithviraj, and A. Ashwitha, “Sensor based application 611
562 data quality is much higher and there is no need for an for smart vehicles,” Int. J. Latest Trends Eng. Technol., vol. 8, no. 1, 612
563 individual to change the corresponding sensor for a period of pp. 526–532, Jan. 2017. 613

564 10 years. [9] R. Pi, “An IoI based forest fire detection using raspberry Pi,” Int. J. 614
Recent Technol. Eng., vol. 8, no. 4, pp. 9126–9132, 2019. 615
[10] N. Pawne, M. Studies, and G. Deo, “Auxiliary safety systems for two,” 616

565 V. C ONCLUSION Int. J. Novel Res. Develop., vol. 2, no. 6, pp. 104–115, Jul. 2017. 617
[11] F. Saeed, A. Paul, A. Rehman, W. Hong, and H. Seo, “IoT-based 618
566 In this article, the issue on rising pollution by vehicles has intelligent modeling of smart home environment for fire prevention and 619
567 been controlled by a new flanged procedure. The focal differ- safety,” J. Sens. Actuator Netw., vol. 7, no. 1, p. 11, Mar. 2018. 620

568 ence between the existing method and the projected method is [12] W. Li and W. Zhang, “Sensor selection for improving accuracy of target 621
localisation in wireless visual sensor networks,” IET Wireless Sensor 622
569 that the ISDA algorithm will divide the necessary regions into Syst., vol. 2, no. 4, pp. 293–301, Dec. 2012. 623
570 circles. This separation has been applied in order to monitor [13] A. Paul, A. Daniel, A. Ahmed, and S. Rho, “Cooperative cognitive 624

571 the vehicles under different temperatures. This method will intelligence for Internet of vehicles,” IEEE Sensors J., vol. 11, no. 3, 625
pp. 1–10, 2017. 626
572 create awareness among the people who are living in metropol- [14] K. D. Singh, P. Rawat, and J.-M. Bonnin, “Cognitive radio for vehicular 627
573 itan area. When the vehicle crosses the threshold contamina- ad hoc networks (CR-VANETs): Approaches and challenges,” EURASIP 628

574 tion value, automated information will be sent to the necessary J. Wireless Commun. Netw., vol. 2014, no. 1, pp. 1–22, Dec. 2014. 629
[15] M. Younis and K. Akkaya, “Strategies and techniques for node place- 630
575 department where, the workers will monitor the vehicle and ment in wireless sensor networks: A survey,” Ad Hoc Netw., vol. 6, 631
576 the engine will be switched off automatically. The purpose no. 4, pp. 621–655, Jun. 2008. 632

577 of providing information to the corresponding department is [16] M. Maksimović and V. Milošević, “Evaluating the optimal sensor 633
placement for smoke detection,” Yugoslav J. Oper. Res., vol. 26, no. 1, 634
578 that the consultant should inform the driver to stop the vehicle pp. 33–50, 2016. 635
579 at a proper place. This policy is usually followed to prevent [17] Status of the Vehicular Pollution Control Programme in India, Minist. 636

580 accidents in urban areas. Further, the results obtained by ISDA Environ. For. Govt. India, CPCB, New Delhi, India, 2010, pp. 1–114. 637
[18] R. Salazar-Cabrera, Á. P. de la Cruz, and J. M. M. Molina, “Sustainable 638
581 prove to be more effective during this monitoring process with transit vehicle tracking service, using intelligent transportation system 639
582 less cost of implementation. Also, ISDA achieves an optimistic services and emerging communication technologies: A review,” J. Traffic 640

583 performance rate of 76.8% which is higher than existing Transp. Eng., vol. xxx, pp. 1–19, Nov. 2020. 641
[19] J. Wei, C.-H. Chiu, F. Huang, J. Zhang, and C. Cai, “A cost-effective 642
584 methods. In future the proposed smoke detection method decentralized vehicle remote positioning and tracking system using 643
585 can be extended by installing multiple sensor nodes at each BeiDou navigation satellite system and mobile network,” EURASIP J. 644

586 highway lanes for detecting the smoke of vehicles in addition Wireless Commun. Netw., vol. 2019, no. 1, p. 8, Dec. 2019. 645
[20] M. Bernas, B. Płaczek, W. Korski, P. Loska, J. Smyła, and P. Szymała, 646
587 with detection of new parameters like direction and speed of “A survey and comparison of low-cost sensing technologies for road 647
588 vehicle. traffic monitoring,” Sensors, vol. 18, no. 10, pp. 1–27, 2018. 648
[21] S. Karmore, R. Bodhe, F. Al-Turjman, R. L. Kumar, and S. Pillai, 649
“IoT based humanoid software for identification and diagnosis of 650
589 R EFERENCES covid-19 suspects,” IEEE Sensors J., early access, Oct. 21, 2020, doi: 651
590 [1] D. Kandris, C. Nakas, D. Vomvas, and G. Koulouras, “Applications 10.1109/JSEN.2020.3030905. 652
591 of wireless sensor networks: An up-to-date survey,” Appl. Syst. Innov., [22] B. Misganaw and M. Vidyasagar, “Exploiting ordinal class structure in 653
592 vol. 3, no. 1, pp. 1–24, Mar. 2020. multiclass classification: Application to ovarian cancer,” IEEE Life Sci. 654
593 [2] S. Ullo et al., “Application of wireless sensor networks to environmental Lett., vol. 1, no. 1, pp. 15–18, Jun. 2015. 655
594 monitoring for sustainable mobility,” in Proc. IEEE Int. Conf. Environ. [23] F. Al-Turjman and J. P. Lemayian, “Intelligence, security, and vehic- 656
595 Eng. (EE), Mar. 2018, pp. 1–7. ular sensor networks in Internet of Things (IoT)-enabled smart-cities: 657
596 [3] S. Kaivonen and E. C.-H. Ngai, “Real-time air pollution monitoring with An overview,” Comput. Electr. Eng., vol. 87, Oct. 2020, Art. no. 106776. 658
597 sensors on city bus,” Digit. Commun. Netw., vol. 6, no. 1, pp. 23–30, [24] S. Ramasubbareddy, S. Ramasamy, K. S. Sahoo, R. L. Kumar, 659
598 Feb. 2020. Q.-V. Pham, and N.-N. Dao, “CAVMS: Application-aware cloudlet 660
599 [4] A. Biswal, J. Subhashini, and A. K. Pasayat, “Air quality monitoring sys- adaption and VM selection framework for multicloudlet envi- 661
600 tem for indoor environments using IoT,” in Proc. AIP Conf., vol. 2112, ronment,” IEEE Syst. J., early access, Oct. 26, 2020, doi: 662
601 Jun. 2019, Art. no. 020180. 10.1109/JSYST.2020.3029807. 663
602 [5] Y. Yamauchi, A. Mammoto, M. Dohi, H. Ebata, and M. Morita, “A cal- [25] A. Namozov and Y. I. Cho, “An efficient deep learning algorithm for 664
603 culation method for predicting heat and smoke detector’s response,” Fire fire and smoke detection with limited data,” Adv. Electr. Comput. Eng., 665
604 Sci. Technol., vol. 24, no. 4, pp. 179–210, 2005. vol. 18, no. 4, pp. 121–128, 2018. 666

thorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on December 26,2024 at 03:32:37 UTC from IEEE Xplore. Restrictions app