Flight Time Optimization in People Identification by Multidrone-Femtocell Systems Roberta Avanzatoa , Francesco Beritellia and Gabriele Nicotrab a Department of Electrical, Electronic and Computer Engineering, University of Catania, Catania, CT, Italy b Department of Mathematics and Computer Science, University of Catania, 95125 Catania, Italy Abstract The paper proposes an extension of a previous algorithm for the geolocation of missing people, which is aimed at a reduction in search times. The proposed technique involves the use of femtocells on board the drone, and therefore offers the possibility for identifying a mobile terminal based on the estimate of the power levels. In particular, a multi-drone system is proposed that allows for better performance in terms of reduction in localization times, which are halved in the case of simultaneous use of 4 drones. Keywords Multi-drone/Femtocell systems, Energy consumption, Mobile terminal positioning algorithm, 4G technologies, Reference signal received power (RSRP) 1. Introduction ger as the only means for covering a disaster area but as a localization system leads to novel studies con- The occurrence of a natural disaster in urban or sub- ducted in [17, 18, 19]. In these studies, the authors urban areas always poses a series of problems in terms propose an algorithm capable of locating any mobile of public safety, social and economic hardship. terminal in a given monitored area through the use of The development in technological innovation is of- UAV systems. Through the femtocell cover, placed on ten able to provide support to the problems that must the drone, it is possible to create a connection with the be faced in the event of a post-natural disaster. For terminals and locate them using the received power example, on a social level it is of crucial importance values. In particular in [19] the authors present a new to connect the areas affected by disasters and cover criterion for classification and geolocation in the pres- them with telecommunications systems [1, 2]. In this ence of non-isotropic radio signal propagation using a regard, many researchers have studied new solutions 4G femtocell aboard a drone system. The authors also based on the use of UAV (Unmanned Aerial Vehicle) present a first study on the capacity and efficiency of systems, proposing audio-video recording systems ba- a time-of-flight optimization and data processing al- sed on technologies for redundant connection in mo- gorithm performed by the drone. The purpose of this bility [3, 4, 5], as well as drone-femtocell system so- algorithm is to reduce rescue times in natural disaster lutions as an alternative to classic radio base stations scenarios as much as possible. when these are out of service [6, 7, 8, 9, 10]. In this article, we propose the extension of the flight Another research field is people identification and time optimization and processing algorithm using a localization [11, 12, 13], in particular the techniques multi-drone-femtocell system. employed searching for missing persons in post-earthquakeThe use of multiple drones with femtocells on board scenarios. allows scanning the monitoring area more rapidly; the Several methods have been proposed to date includ- algorithm is responsible for making the two or four ing the localization of mobile terminals by radiofre- drones cooperate, in order to follow their respective quency (RF) signals, in scenarios where rubble is a sour- paths with the minimum overlap. This mechanism leads ce of significant attenuation to the propagation of the to a considerable reduction in the flight and processing electromagnetic signal [14, 15, 16]. times of each drone and therefore avoids considerable The idea of using the drone-femtocell system no lon- waste of flight energy [20]. The paper is structured as follows: section II de- ICYRIME 2020: International Conference for Young Researchers in Informatics, Mathematics, and Engineering, Online, July 09 2020 scribes the proposed method, i.e. the method of op- " roberta.avanzato@phd.unict.it (R. Avanzato); timizing flight and processing times through multiple francesco.beritelli@dieei.unict.it (F. Beritelli) drone-femtocell systems; section III shows the perfor-  Β© 2020 Copyright for this paper by its authors. Use permitted under Creative mances obtained using this method as the number of CEUR Workshop Commons License Attribution 4.0 International (CC BY 4.0). CEUR Workshop Proceedings (CEUR-WS.org) http://ceur-ws.org ISSN 1613-0073 drones used and the size of the monitoring area vary; Proceedings Figure 1: Optimized root in the case 1 drone is used. the final section is dedicated to conclusions. in [19], the main hypotheses for the application of this algorithm are summarized below: 2. Proposed Method β€’ The grid must be an 𝑀 Γ— 𝑀 matrix where 𝑀 = 2𝑛 + 1 and 𝑛 = 2, 3, . . . , 𝑁 ; In this paper, the "Cluster-based Fast Proximity Algo- rithm" proposed in [19] is extended to the use of two or β€’ The matrix must not be 2 Γ— 2 or 3 Γ— 3; more cooperating drones, optimizing flight time and β€’ The number of iteration phases of the algorithm areas to be covered. This mechanism leads to a re- must be given by: duction in the energy consumed by drones and also in rescue times. 𝐹 = 𝑛 = π‘™π‘œπ‘”2 (𝑀 βˆ’ 1) (1) Therefore, by using multiple drone-femtocell sys- tems, the need arises to remodel the algorithm for op- The equations that determine the processing time, timizing the drone flight time, in order to intelligently flight time and energy, respectively, in the case of two cover each sub-area of the monitoring area. To this and four drones are the following: end, once the optimization algorithm is applied, the graphs relating to processing times, flight times and π‘œ π‘‡π‘ƒβˆ’π‘‘π‘œπ‘‘ = 𝑑𝑝 βˆ— (5 βˆ— πΉπ‘šπ‘Žπ‘₯ + 1) (2) energy expenditure are obtained as the number of dron- es used and the size of the matrix that defines the mon- πΉπ‘šπ‘Žπ‘₯ itoring area vary. To apply the algorithm, the follow- π‘‡π‘‰π‘œ βˆ’π‘‘π‘œπ‘‘ = 𝛿𝑑 βˆ— 20 + 2πΉπ‘šπ‘Žπ‘₯ βˆ‘ 𝐹 βˆ’2 3 (3) ing constraints were introduced: [ ( 𝐹 =2 2 )] β€’ coverage radius of the femtocell on board the πΈπ‘‡π‘œ 𝑂𝑇 = 𝑃 βˆ— (π‘‡π‘ƒβˆ’π‘‘π‘œπ‘‘ π‘œ + π‘‡π‘‰π‘œ βˆ’π‘‘π‘œπ‘‘ ) (4) drone equal to half the diagonal of the starting The processing, flight and energy expenditure times grid; in the case of two drones are defined, respectively, by β€’ the terminals hook onto the first femtocell they (2), (3), and (4). However, in the case of 4 drones, the detect; processing, flight and energy expenditure times are defined by (5), (6), and (7), respectively. β€’ the drones depart from the edges of the grid with { } a time lag of one minute, to prevent them from 𝐹 +1 π‘œ π‘‡π‘ƒβˆ’π‘‘π‘œπ‘‘ = 𝑑𝑝 βˆ— 5 βˆ— +4 (5) passing through the same point at the same time; [ 2 ] β€’ uneven distribution of terminals. 3 π‘‡π‘‰π‘œ βˆ’π‘‘π‘œπ‘‘ = 𝛿𝑑 βˆ— [2πΉπ‘šπ‘Žπ‘₯ (2 + 𝐹 βˆ’2 )] (6) 2 π‘šπ‘Žπ‘₯ The drone-femtocell system and the details of the classification and localization algorithms are defined π‘œ πΈπ‘‘π‘œπ‘‘ π‘œ = 𝑃 βˆ— (π‘‡π‘ƒβˆ’π‘‘π‘œπ‘‘ + π‘‡π‘‰π‘œ βˆ’π‘‘π‘œπ‘‘ ) (7) 35 Figure 2: Optimized root in the case of simultaneous use of 2 drones 3. Performance Evaluation divided into 9 rows and 9 columns, the phases of the algorithm and the drone path, respectively, when 1, 2 In this section we will evaluate the performance of the and 4 drones are employed. The differences concern flight time optimization and processing algorithm in the number of processing points and the flight seg- the case of 1 drone, 2 drones and 4 drones. As already ments of each individual drone. In the case only one seen in [19] the "Cluster-based Fast Proximity Algo- drone is used, 19 points are processed during the three rithm" algorithm was applied based on a single drone, phases of the algorithm, whereas using two drones in this paper we will apply it to several drones, com- they are reduced to 16. As for the flight segments, from paring performance, in terms of time reduction and the 68 segments obtained with one drone we pass to 56 energy consumption in three different cases. segments. All this leads to a reduction in the process- To test the performance of the system, a practical ing and flight times of each individual drone. However, example of a matrix of size 𝑀 = 9 will be considered, using 4 drones it is possible to observe that the number i.e. a 9 Γ— 9 matrix (with a resolution of 2 meters, thus of phases each drone must complete decreases while obtaining a monitoring area of 18 Γ— 18 meters). Using maintaining the size of the grid unchanged; this hap- two drones, positioned at opposite edges of the area, it pens because each drone is responsible for scanning is noted that the number of phases that the algorithm a smaller sub area equal to almost half of the origi- runs is given by (1) and remains unchanged compared nal one. This decrease occurs every time 4 drones are to the case of using only one drone, i.e. 3 phases are used, regardless of the size of the grid. There is also a carried out. further decrease in the processing points (equal to 14) Fig. 1, Fig. 2, and Fig. 3 show the monitoring area and in the flight segments (equal to 28). 36 Figure 3: Optimized root in the case of simultaneous use of 4 drones Figure 4: Flight and processing time as the number of Figure 5: Energy consumed by a single drone as the number drones varies for a 9 Γ— 9 matrix of drones varies for a 9 Γ— 9 matrix Fig. 4 shows the flight and processing time trends of see that the maximum decrease is obtained by passing a matrix M = 9, as the number of drones used varies. from 2 to 4 drones, with the flight time being halved. As for the processing time, it can be observed that it Regarding the energy, represented in Fig. 5, a net significantly decreases using 4 drones, going from 570 decrease is obtained, passing from the use of 2 drones seconds (9.5 minutes) with one drone to 420 seconds (total energy equal to 50.44Wh) to 4 drones (33.95 Wh). (7 minutes) with 4 drones. Using two drones, how- With one drone, on the other hand, there is an energy ever, the processing time drops to 480 seconds (8 min- consumption of 60.62 Wh. utes). While, the total flight time varies from 680 sec- To generalize the considerations made, additional onds (11.33 minutes) using one drone, to 560 seconds graphs were obtained as the size of the matrix on which (9.33 minutes) with 2 drones, decreasing up to 280 sec- the localization algorithm is applied varies. Fig. 6, Fig. 7 onds (4.67 minutes) using 4 drones. In this case we can and Fig. 8 represent, respectively, the trend of the curves 37 Figure 8: Flight time of a single drone as the size of the Figure 6: Processing time of a single drone as the size of the matrix and the number of drones vary. matrix and the number of drones vary. Table 1 Evaluation of flight times, processing times and energy, as the size of the monitoring area varies using two drones M ToP-tot [min] ToV-tot [min] EoTOT [MJ] 5 5.5 3.66 2.2 9 8 9.33 4.16 17 10.5 20.66 7.48 33 13 43.33 13.52 Figure 7: Energy consumed by a single drone as the size of the matrix and the number of drones vary. relating to processing time, flight and energy consump- tion, based on the use of 1, 2 or 4 drones. These figures confirm what has been said for a 9 Γ— 9 matrix. In fact, as regards the processing time, the greatest reduction is obtained by passing from 1 to 4 drones. Figure 9: Total duration of the journey of a single drone as In terms of flight time and energy, there is a sharper the number of drones and the size of the matrix vary. decrease from 2 to 4 drones. Once the number of drones has been fixed, processing times, flight times and en- ergy increase hand in hand with the increase in the given by the sum of the processing and flight times. As size of the matrix, but with a different trend. The pro- shown in Fig. 9, once the size of the matrix is fixed the cessing time increases linearly, while the flight time total duration of the journey is considerably reduced, and energy grow according to an exponential trend. almost halving going from 2 to 4 drones. For example, having set the use of drones equal to Considering that during a search and rescue oper- 2, the flight, processing and energy consumption times ation of missing persons time is a determining factor, were obtained as the size of the matrix changed. The being able to locate terminals in the shortest possible data is represented in Table 1. timeframe is a major advantage. Another interesting graph that has been obtained concerns the total duration of the journey of each drone, 38 4. Conclusion [5] G. Capizzi, S. Coco, G. Lo Sciuto, C. Napoli, A new iterative fir filter design approach using a In this paper, a study was presented concerning the gaussian approximation, IEEE Signal Processing combined use of drones and femtocells. In particular, Letters 25 (2018) 1615–1619. the paper analyses the cases UAV-femtocell systems [6] M. Deruyck, J. Wyckmans, L. Martens, W. Joseph, are used to create ad hoc emergency networks during Emergency ad-hoc networks by using drone disaster scenarios or to facilitate search and rescue op- mounted base stations for a disaster scenario, erations for civilians missing in post-earthquake sce- in: 2016 IEEE 12th International Conference on narios. Using a real simulation scenario, the following Wireless and Mobile Computing, Networking were considered: and Communications (WiMob), 2016, pp. 1–7. [7] M. Deruyck, J. Wyckmans, W. Joseph, L. Martens, β€’ the number of drones required with respect to Designing uav-aided emergency networks for different parameters (duration of the interven- large-scale disaster scenarios, EURASIP Jour- tion, user coverage, flight height of the drone, nal on Wireless Communications and Network- level of service provided to the user); ing 2018 (2018) 79. β€’ reduction of processing and flight times of each [8] K. Gomez, A. Hourani, L. Goratti, R. Riggio, individual drone; S. Kandeepan, I. Bucaille, Capacity evaluation of aerial lte base-stations for public safety commu- β€’ optimization of battery life (if, for example, a nications, in: 2015 European Conference on Net- single drone has a discharged or low battery to works and Communications (EuCNC), 2015, pp. cover the entire monitoring area, there are other 133–138. drones to support it to cover the areas it cannot); [9] J. Lyu, Y. Zeng, R. Zhang, T. J. Lim, Placement op- timization of uav-mounted mobile base stations, β€’ possibility to replace a drone in case any of them IEEE Communications Letters 21 (2017) 604–607. is damaged; [10] E. Kalantari, H. Yanikomeroglu, A. Yongacoglu, β€’ localization of more terminals, through the se- On the number and 3d placement of drone lection of a greater number of sub-areas, where base stations in wireless cellular networks, in: greater quantities of terminals with greater power 2016 IEEE 84th Vehicular Technology Confer- are detected. ence (VTC-Fall), 2016, pp. 1–6. [11] C. Napoli, G. Pappalardo, E. Tramontana, An agent-driven semantical identifier using radial References basis neural networks and reinforcement learn- ing, in: Proceedings of the XV Workshop ”Dagli [1] R. Giuliano, F. Mazzenga, A. Vizzarri, Integration Oggetti agli Agenti”, volume 1260, CEUR-WS, of broadcaster and telco access networks for real 2014. time/live events, IEEE Transactions on Broad- [12] J. A. del Peral-Rosado, R. Raulefs, J. A. LΓ³pez- casting 66 (2020) 667–675. Salcedo, G. Seco-Granados, Survey of cellular [2] F. Bonanno, G. Capizzi, G. Lo Sciuto, C. Napoli, mobile radio localization methods: From 1g to Wavelet recurrent neural network with semi- 5g, IEEE Communications Surveys Tutorials 20 parametric input data preprocessing for micro- (2018) 1124–1148. wind power forecasting in integrated generation [13] R. Giuliano, G. C. Cardarilli, C. Cesarini, systems, in: 5th International Conference on L. Di Nunzio, F. Fallucchi, R. Fazzolari, Clean Electrical Power: Renewable Energy Re- F. Mazzenga, M. Re, A. Vizzarri, Indoor lo- sources Impact, ICCEP 2015, 2015, pp. 602–609. calization system based on bluetooth low energy [3] F. Beritelli, A. Gallotta, C. Rametta, A dual for museum applications, Electronics 9 (2020) streaming approach for speech quality enhance- 1055. ment of voip service over 3g networks, in: 2013 [14] H. Liu, H. Darabi, P. Banerjee, J. Liu, Survey of 18th International Conference on Digital Signal wireless indoor positioning techniques and sys- Processing (DSP), 2013, pp. 1–5. tems, IEEE Transactions on Systems, Man, and [4] F. Beritelli, A. Spadaccini, The role of voice ac- Cybernetics, Part C (Applications and Reviews) tivity detection in forensic speaker verification, 37 (2007) 1067–1080. in: 2011 17th International Conference on Digi- [15] S. M. M. Dehghan, S. Haidari, H. Mordi, Toward tal Signal Processing (DSP), 2011, pp. 1–6. aerial simultaneous target localization and obsta- 39 cle estimation using rssi observations, in: 2015 3rd RSI International Conference on Robotics and Mechatronics (ICROM), 2015, pp. 517–522. [16] T. Golubeva, Y. Zaitsev, S. Konshin, I. Duisen- bek, A study on the wi-fi radio signal attenua- tion in various construction materials (obstacles), in: 2018 Tenth International Conference on Ubiq- uitous and Future Networks (ICUFN), 2018, pp. 718–723. [17] R. Avanzato, F. Beritelli, M. Vaccaro, Identi- fication of mobile terminal with femtocell on drone for civil protection applications, in: 2019 10th IEEE International Conference on In- telligent Data Acquisition and Advanced Com- puting Systems: Technology and Applications (IDAACS), volume 1, 2019, pp. 269–273. [18] R. Avanzato, F. Beritelli, An innovative technique for identification of missing persons in natural disaster based on drone-femtocell systems, Sen- sors 19 (2019) 4547. [19] R. Avanzato, F. Beritelli, A smart uav-femtocell data sensing system for post-earthquake localiza- tion of people, IEEE Access 8 (2020) 30262–30270. [20] F. Bonanno, G. Capizzi, C. Napoli, Some remarks on the application of rnn and prnn for the charge- discharge simulation of advanced lithium-ions battery energy storage, in: SPEEDAM 2012 - 21st International Symposium on Power Electronics, Electrical Drives, Automation and Motion, 2012, pp. 941–945. 40