Mobile Communication Systems to Control UAVs: Measurements of QoS Parameters

This paper proposes to identify a propagation model that considers the unmanned aerial vehicles (UAVs) unique characteristics, contemplating two actual wireless technologies, UMTS and LTE, which are theoretically capable of supporting a real-time video service admitting more than one quality index according to the RF conditions. Several measurements were made in a specific outdoor rural scenario in order to understand if the current network infrastructure is prepared to support this type of service using these vehicles, by simulating a real case scenario and considering critical locations where the loss of Quality of Service (QoS) can be significant due to the hole phenomenon that occurs over the antennas/base stations, raising the probability to occur handover.


I. INTRODUCTION
Nowadays, the use of Unmanned Aerial Vehicles (UAV) for civil, military and industrial purposes is growing and communication between the UAV and the human operator must be assured for every type of application. This paper's main goal is to guarantee the quality of service when an UAV flies at certain altitudes where communication is expected to suffer the hole phenomenon when switching antennas/base stations (handover). This phenomenon can be a problem when performing certain operations where a constant and uninterrupted communication is required. The theoretical research for an empirical propagation model that fits into the UAVs unique characteristics was crucial, in order to provide an attenuation estimation based on the transmitted signal, which led to the conclusion that Lisbon University Institute (LUI) model best suits the unique requirements of these type of vehicles, by assuming unusual heights for the base stations' and terminal's antennas, and a wide frequency interval that permits to include the Universal Mobile Telecommunication System (UMTS) and Long-Term Evolution (LTE) frequency bands [1] [2]. A unique spectrum analyzer was used to understand the variation of certain parameters according to the vehicle's simultaneous changes in a 3D coordinates system (latitude, longitude, and altitude). Parameters like signal strength, interference, and channel capacity/quality were analyzed to understand the viability of the network infrastructure from a specific service provider to accomplish the lowest requisites to transmit a real-time video service using a drone. The measurements were made in three different locations close to two base stations from two distinct service providers, in a rural environment, in order to support empirically the previously referred propagation model. These measurements took in consideration the areas above the base station since it is where it is more common to see a significant drop of signal strength and quality, based on the hole phenomenon caused by the lack of coverage from the antennas. The Figure 1 demonstrates the proposed flight plan for every measurement done that only considers 10 meters above the base station due to the windy conditions at the time of the trial, that could put at risk the expensive equipment. The samples captured by the spectrum analyzer and its respective parameters were monitored, recorded and saved into a .csv file, using ROMES software provided by Rohde & Schwarz. Afterward, these files were filtered to supply only the necessary information to design 2D and 3D graphics that relate diverse parameters that are essential to understanding whether cellular networks are trustable to support a video streaming service using these unique vehicles in a rural environment.

II. LUI MODEL
Considering the large quantity of propagation models that exist, it is important to choose one that fulfills the requirements inherent to the UAV's unique characteristics. Firstly, it is necessary to reduce the number of possibilities by defining the type of propagation model: empirical, theoretical or hybrid. In this case, empirical is the best option since it is based in measurements or experimental trials. It is also adequate to identify the environment, scenarios, the base stations' and terminal's station heights, and a frequency range that includes UMTS and LTE frequency bands, ensuring higher data rates to overcome or guarantee the minimum requisites for real-time video. The most known and used empirical models like Okumura-Hata, Cost 231-Hata, Walfish-Ikegami, Erceg and SUI model were developed for specific scenario, assuming a limited frequency range and showing the incapability to consider simultaneously the UMTS and LTE frequency bands. However, LUI model demonstrates the opposite by assuming a wider frequency spectrum from 800 to 2600 MHz. Besides that, most of these models are used in scenarios where the base stations' and terminal stations' height are between 0 and 200 meters and 3 and 10 meters, respectively. These heights are ideal for the general user equipment like smartphones and notebooks, but it represents a limitation which, once again, LUI model is able to overcome since it considers infinite heights for the base stations and terminal stations. Nonetheless, this model can assume one of two formulas depending on the height of the terminal station, since one of the factors related with the angles of the antenna ( ) attenuates significantly the signal strength results when the terminal station's height (ℎ ) is below the base station's height (ℎ ) , which is proved by the Figure 1. However, this factor does not affect that parameter when terminal station's height is above the base station's height, resulting in two distinct formulas to calculate the average path loss in each one of these cases: Where d is the distance using a 3D coordination system and can be calculated by using (2).
L0 represents the free path loss, stands for path loss exponent and it can assume different values according to the type of environment, d0 is the reference distance, in meters, that vary according to the technology in use which, in this case, assumes picocell characteristics that is represented by d0=1 meter. Δ ,ℎ is the correction factor associated to the BS effective height and it is usually multiplied by a rectangular function [ (ℎ ) − (ℎ − ℎ )]) that can result in 0 if: hTS < 0 or hTS > hBS; or 1 if: 0 < hTS < hBS. describes the path loss exponent (3) and it varies according to the terrain category reflected by the parameters a, b and c values, the base station antenna effective height, and the expected result from rectangular and unit step functions (ℎ − ℎ ), where the last function result in 0, if: hTS < hBS; or 1, if: hTS ≥ hBS.
is another correction factor that rely on the characteristics of the antennas like azimuth ( ), elevation angle ( ), tilt of the antenna ( ) and the angle that determines which sector is being used ( ), corresponding to the area from one or more antennas on the base station that provides coverage to the terminal station (4).
is one of the two correction factors necessary to determine the result taking into account the elevation and the tilt of the antenna (5). where, Where the elevation angle ( ) can be determined by the formula (7): + is the remaining correction factor to be able to calculate , that considers the azimuth ( ) and angle that determines which used is being used ( ), by using the formula (8): The previous correction factors represented by the formula (6) and (9) use the inverse of the Dirac Delta Function (DDF), which are (1 − δ( + Ψ) ) and (1 − δ( + ) ). This function can result in 0 or 1, depending on the condition (10): Graphic 1. Signal strength variation w/ or w/o the χangles correction factor. [1] III. EQUIPMENT Beforehand, it was necessary to take a look at the market to understand which spectrum analyzer would and type of drone would fit in such specific demanding. Firstly, there was the chance to test Spectran HF-60100, that permitted to verify that is capable of determining the signal strength for every individual signal in a certain frequency range defined by the user, using its own software to study the power variations while measuring the signal. However, there are several cons like the autonomy (~20 minutes), limited memory size and the lack of identification from the antennas or base stations unless the directional antenna is pointed exactly to one of them. Based on the previous statements, it wouldn't be possible to obtain sustainable and reliable results. R&S TSME is also a spectrum analyzer able to measure up to eight different technologies simultaneously in the 350 MHz to 4.4 GHz. It is compact, lightweight, low power consumption and it has an internal GPS. Unlike the first, it provides information related to base station ID, signal strength/quality, SINR and several codes (MCC and MNC) that permit to identify the service providers. The only defect that affects the final decision is the fact that it needs a full-time physical connection with a host PC, which makes this combination extremely (close to 5 kg, considering the use of a regular laptop) heavy to be lifted by a light/medium caliber drone. Finally, R&S TSMA is similar to TSME but the main and crucial difference between the two is that TSMA is battery powered with rechargeable batteries and charging function, ensuring that is always ready to operate. With its functions, it is possible to analyze and detect radio dead zones (e.g., hole phenomenon) or locations with too much interference. It comes with ROMES software that permitted the analysis of the diverse signal related parameters while measuring it and save that progress into a file for future data treatment/filtering. Besides that, it is possible to control the software by establishing a wireless connection with a smartphone to provide the user interface for configuration before starting the measurement campaign. The only defect is the weight (~2.5 kg) but, in this case, it is possible to overcome it by using a medium/heaving weight drone (e.g. octocopter). After studying the pros and cons of each spectrum analyzer, it was clear that R&S TSMA was the only able to accomplish the challenges of this measurement campaign, pointing out three characteristics: battery autonomy, lightweight and independence. Furthermore, it was used an octocopter, which is considered to be a medium/heavy caliber drone that was configured to lift 2.5 kg related to the spectrum analyzer together with its battery pack unit which it is possible to verify it in the Figure 3

IV. MEASUREMENTS
The main goal in this measurement campaign is to understand how the signals provided by each antenna on the base station behaves during the flight and try to understand if there is any location where the reference sector, which is the sector where the terminal station is located, might not be able to provide sufficient throughput for a video streaming service considering more than one video quality option since the capacity is not the same for UMTS and LTE. If this sector is not able to provide the required quality and strength, the network might consider handover if the cellular infrastructure is prepared to support it since these vehicles behaviors are not common when compared to the ones from general equipment. Considering the importance of throughput in video streaming service, it is necessary to verify the recommended bit rates associated to a specific video quality.  [4] The video quality parameter varies from 296 kbps (VQ1), which represents the minimum requisites to ensure that video streaming is maintained with the lowest quality, to 7196 kbps (VQ6). However, in this case, the highest quality considered is VQ5 since only PCs and tablets were assumed for supporting the streaming service, so the minimum requirement for highest quality is 3246 kbps according to Based on the fact that all the measurements realized have similar characteristics and goals that are illustrated in Figure  1, only one scenario is presented here as a reference to the other two.

• Base Station A (BS A):
o  However, there is a quality regression between 70 and 130 seconds, which is the interval when the terminal station is above the antenna that covers the reference sector, where is not able to fulfill the requirements to sustain a high-quality video streaming. Nonetheless, Graphic 3 demonstrates that the adjacent sectors from the reference base station are able to compensate the lack of quality to maintain the high demanding for video quality in real-time video service, leading to a softer handover. In LTE's case, there is no point in referring the lowest quality (VQ1) due to the fact that any of the presented sectors is able to fulfill the minimum requirements during the entire flight and that is one of the main reasons why the highest quality threshold is considered in this technology. Graphic 4 illustrates the relation between throughput and signal strength, which leads to the conclusion that they are not directly proportional due to the interference factor used in Shannon's Theorem [5][6] to calculate the throughput results (11). However, even if in the reference sector looks that way, the adjacent sectors prove it wrong by assuming higher signal strengths than the reference sector in certain locations, but with higher interference that leads to worst quality signals, which is also proved by the order 2 polynomial trendlines for each sector.  The three previous graphics demonstrate how signal strength behaves according to simultaneously changes in drone's movement like distance to BS and its height from the ground level. The Graphics 5 and 6 are related to the adjacent sectors of the BS A, demonstrating that the spectrum analyzer wasn't receiving any signal strength from the antennas covering the adjacent sectors when close to 40 meters height, which corresponds to the antennas' height in BS A. Unlike the previous cases, the reference sector results that are illustrated in Graphic 7, demonstrate that, close to the same location, it is capable of providing greater signal strength results due to the fact that terminal station (drone) is located in front of the main lobe of the antenna covering the present sector.
Graphic 8. CDF vs Throughput; LTE: 796 MHz; Sc: B Graphic 8 exhibits the probability of each sector reach a certain signal strength/throughput value. In this case, the reference sector represented by the orange line is predominant, assuming the highest probability to reach higher results in both parameters when comparing to the remaining sectors from the same base station. The adjacent sectors assume almost the same probabilities. However, the sector represented by the blue line (CI: 177) has a higher probability to assume superior results when compared to the ones from the remaining sector (CI:179), but this difference is not significant. In this scenario, the spectrum analyzer did not capture any information about sectors from adjacent base stations using the same channel frequency, considering this technology and the same service provider. In UMTS, lower throughput results are expected according to the theoretical limits related to this technology and this is illustrated in Graphic 9, where only one sample from the reference sector is above the highest quality threshold. Based on the previous statement, it is expected that the main goal for UMTS is to guarantee the minimum requisites to support a real-time video due to its limitations. However, the reference sector does not provide enough throughput during the entire flight and that is demonstrated in the interval from 70 to 120 seconds described in Graphic 10 but, once again, the adjacent sectors from the same base station are able to provide enough capacity to guarantee the minimum requisites to maintain a video streaming service, even if in the lowest quality, assuming the existence of a softer handover event in these cases.
Graphic 11. Order 2 polynomial trendlines for every cell in reference BS The trendlines expressions relating the signal strength and throughput represented in Graphic 11 have some similarities to the ones from LTE but, in this case, the one corresponding to the reference sector does not assume such a constant growth. However, in the remaining sectors, they are similar. broadcasting might stop while this event occurs. The hole phenomenon occurs at the top of the base station/antennas in both technologies. However, if the main goal is to guarantee the minimum service requisites, the reference sectors achieve it using LTE technology. So, if LTE technology is available, it is more reliable to use it under these circumstances. According to the Portuguese law, it is possible to fly a drone to a limit of 120 meters height when in user's line of sight. Besides that, assuming a new flight plan where the drone would have to go from one BS to another from the same service provider, by also assuming the highest possible altitude according to the legislation and verify if the infrastructure is ready to guarantee the Quality of Service (QoS) for the same or other services that require higher data rates.