The Impact of LoRa Parameters on UAV-to-X Communications in Emergency Response Scenarios

The Impact of LoRa Parameters on UAV-to-X Communications in Emergency Response Scenarios Maria Karatzia, Nicolas Souli, Panayiotis Kolios, and Georgios Ellinas Abstract—In recent years, emergency response systems have been employed extensively in critical operations, particularly for disaster management scenarios. During these operations, robust and efficient communication is crucial to ensure that first responders can obtain critical information for coordination and adaptation to the dynamic conditions of an emergency. The employment of unmanned aerial vehicles (UAVs) in conjunction with Internet of Things (IoT) devices can provide support and enable the transmission and reception of relevant information in emergency situations. A number of different communication methods have been explored to address communication performance and robustness issues that arise during emergencies. This work tackles these issues by developing and evaluating the performance of an integrated LoRa (long-range communication)-based system for UAV-to-X communications. For evaluation purposes, a LoRa-based prototype system is designed and implemented to achieve robust, real-time, and efficient communication for both static and mobile nodes in indoor and outdoor environments. The prototype of the proposed communication architecture is subsequently tested in a real-world environment, demonstrating the feasibility and effectiveness of the proposed solution in terms of communication range, transmission latency and security for the applications investigated. I. INTRODUCTION Emergency response systems play a pivotal role in safeguarding human lives and property during emergency situations. Such systems aim to facilitate rapid response, efficient coordination, and effective communication, as prompt intervention can significantly minimize the extent of damage or loss of life [1]. However, the effectiveness of these systems is often challenged by various factors inherent to the diverse nature of emergencies, as each presents distinct obstacles. For instance, wildfires often spread rapidly and unpredictably, requiring responders to operate at significant distances due to the intense heat and smoke. In contrast, urban searchand-rescue operations in earthquake-stricken areas necessitate precise localization of victims within collapsed structures. Floods and hurricanes introduce further challenges, such as damaged infrastructure and wide geographical impact, which complicate the coordination and deployment of resources. M. Karatzia, N. Souli, and G. Ellinas are with the Department of Electrical and Computer Engineering and the KIOS Research and Innovation Center of Excellence, University of Cyprus. P. Kolios is with the Department of Computer Science and the KIOS Research and Innovation Center of Excellence, University of Cyprus. Email: {karatzia.s.maria, souli.nicolas, pkolios, gellinas}@ucy.ac.cy This work was partially supported by the Border Management and Visa Policy Instrument (BMVI), co-financed by the European Union and the Republic of Cyprus (BMVI/2021-2022/SA/1.2.1/015) (project REACTION). It was also partially supported by the European Union’s Horizon Europe research and innovation program under grant agreement No 101187121 (EUSOME) and by the DigiPol project of the NextGenerationEU programme under the Republic of Cyprus’ Recovery and Resilience Plan, (collaboration agreement ΚΔΕΕ 05/2022 between KIOS and the Cyprus Police). Nevertheless, irrespective of the emergency situation faced, robust communication is fundamental to the success of any emergency response operation. Efficient communication ensures that responders can convey critical information, coordinate their efforts, and adapt to the dynamic conditions of the emergency environment. However, conventional communication networks are frequently compromised during disasters due to infrastructure damage, network congestion, or lack of coverage in remote areas. In response to these challenges, various communication technologies have been thoroughly investigated. In particular, low power wide area networks (LPWANs) such as ZigBee, LTE-M, LoRa, and LoRaWAN have received significant attention due to their potential to provide resilient and scalable communication solutions. For example, LPWANs are employed in applications that require transmitting a limited amount of information over long-range distances [2]. LoRa, as a wireless communication protocol, is one of the most promising LPWAN technologies, as it is designed for low-power, long-range communications suited for IoT applications [3]. LoRa defines the physical (PHY) and data link layer protocols to enable transceivers that operate in the sub-gigahertz frequency bands, allowing for efficient communication amongst numerous nodes over extended distances, while consuming minimal power by employing a chirp spread spectrum modulation technique [4]. Thus, LoRa provides distinct advantages for embedded systems, delivering reliable connectivity and extended operational lifespan. Additionally, it offers cost-effective applicability for commercial and industrial purposes [5]. Therefore, during emergency scenarios, seamless LoRa networks have proven to be an extremely valuable technology. However, the effectiveness of LoRa is greatly influenced by various factors such as: (i) LoRa configuration parameters, (ii) interference, (iii) hardware, (iv) topology, and (v) deployment environment [6]. These factors directly impact the system’s reliability, transmission times, communication range, and packet loss. Specifically, LoRa parameters such as the spreading factor (SF), bandwidth (BW), coding rate (CR), and transmission output power (OP) can directly affect the system’s performance and thus need to be thoroughly investigated for the specific application setups. Our focus is on developing a LoRa network that maintains UAV-toX communication, while optimizing communication range, transmission latency, security, and power consumption. In the state of the art, theoretical approaches have been investigated to achieve the optimal selection of LoRa parameters, detailing both the benefits and drawbacks of configuration adjustments. However, the association between theoretical models and real-world applications varies significantly due to the impact of rapid environmental changes during an emergency. Thus, it is crucial to understand the LoRa parameters’ impact on the system and propose a methodology for optimizing these parameters in order to fully leverage the benefits of LoRa technology in emergency response operations. In accordance, this work proposes a LoRa-based practical implementation for UAV-to-X communications in emergency response scenarios, while also presenting an extensive parameter analysis of the LoRa-connected network that employs both static and mobile nodes in indoor and outdoor setups. The setup is based on our previous works in [7] and [8] that implemented an integrated prototype LoRa-ROS-based communication system used in collaborative UAV positioning applications. The main contributions of this work are: (i) The design, development, and implementations of a LoRabased UAV-to-X communication system that can be used for both indoor and outdoor emergency response operations; (ii) A methodical investigation of how each LoRa configuration parameter affects the proposed system’s performance through numerous outdoor and indoor experiments. Specifically, for a LoRA-based UAV-to-X communication system, that is developed to distribute information, LoRa parameters that can directly affect the system’s performance, such as the SF, BW, CR, and OP are thoroughly examined. The goal is to develop a system that effectively meets emergency response needs in indoor and outdoor conditions in terms of communication range, transmission rate, transmission latency, and reliability; and (iii) An experimental evaluation of the LoRa-based system’s performance in an actual search-and-rescue operation, that took place in an open field, with two UAVs deployed to cover a search area. Results demonstrate the importance of taking into account additional emergency response operational factors to ensure effective system performance. II. RELATED WORK Numerous works in the literature have investigated different methods to effectively select the optimal parameter configuration for LoRa networks. Various parameters have been examined either individually or collectively, including their impact on different aspects of the overall system performance. For example, several studies have focused on algorithms and mathematical methodologies to ascertain the optimal selection of LoRa parameters. In [9], a probing regime is employed to provide an optimal choice for the LoRa parameters. In that study, the proposed algorithm determines the subsequent probing configuration based on transmission energy considerations. Also, in [10], a mathematical optimization formula is introduced, that specifically considers SF and CR and is subsequently tested using the LoRaSim open-source simulator, with the proposed method demonstrating promising results, including reductions in data extraction rate. Moreover, in [11], a theoretical assessment of symbol and bit error probabilities is presented and the findings, based on various spreading factor values, are confirmed through a LoRa simulator. However, for all three studies, there is no practical deployment to validate the results of the proposed methodologies and the approaches have not been verified in a real-world environment. The majority of experimental research efforts have focused on examining the effects of different LoRaWAN configurations on communication performance (e.g., transmission range and energy efficiency). In [12], experimental evaluations are carried to examine multiple factors that could influence the performance of a LoRaWAN network. In terms of LoRa PHY layer parameters, the coding rate and payload length have been studied to demonstrate their relationship with the packet delivery rate. Further, a large-scale LoRaWAN test is investigated in [13], utilizing 5mobile devices and 24 different gateways. The received signal strength indicator (RSSI) and signal-tonoise ratio (SNR) values were analyzed for different SF values, while the BW and CR were set as constant parameters. In a similar vein, as [14] describes, LoRaWAN performance is analyzed based on changes in the CR parameter, with metrics such as RSSI, SNR, and packet error rate used to evaluate the communication performance. Further, in [15], a visual line-of-sight experimental evaluation is conducted using an Arduino and LoRa shield as a transmitter, while a singlechannel gateway is used as a receiver. The outcomes of that study present the impact of the distance and of the SF on packet loss, data throughput, and data transfer effectiveness. Additionally, an experimental evaluation is presented in [16], aiming to identify the optimal LoRa configuration for an RSSI-based positioning system. Although that work presents promising results, only two different configurations of OP, BW, and SF are tested. Also, the experiments take place over relatively small distances in an open-air space and without taking into consideration the impact of physical obstacles. In [17], the authors propose a mathematical model for optimizing the LoRa network performance in terms of time-on-air (TOA), received power, and RSSI. Even though the outcomes demonstrate an improvement in performance as compared to the existing LoRa network, performance evaluation is solely based on simulation. Several studies have also highlighted the critical need for reliable and effective UAV-to-X communication, while also prioritizing energy efficiency and cost-effectiveness. In [18], a thorough analysis is conducted to evaluate various UAV-based communication technologies (including Wi-Fi, cellular, and LoRaWAN). However, these technologies are limited by short communication ranges and are often significantly affected by signal interference. Further, [19] proposed a water monitoring system using UAVs, which employs a 4G network for data communication. Even though this approach demonstrated promising results, it cannot be utilized in emergency situations where network infrastructure could be disrupted (as it heavily relies on stable connectivity). Moreover, the authors in [20] proposed an intelligent reflecting surface (IRS)-assisted UAVto-X communication system. The results indicated transmission latencies ranging from 75.27s to 120s, depending on the number of federated learning rounds. However, it is important to note that the system was only evaluated in a simulated environment. Finally, in [21], a communication architecture that combines various wireless communication technologies (such as Wi-Fi, 3G, 4G, and software-defined radio (SDR)) is proposed to ensure interference-free and simultaneous data transmissions, while addressing mobility concerns with the utilization of time-division multiple access (TDMA) and frequency-division multiple access (FDMA) techniques. How- ever, this approach is based on access to public networks that may not be available in disaster scenarios. Again, the proposed system is only evaluated through simulations. As previously mentioned, this work builds upon our previous works in [7], [8] that developed and implemented an integrated LoRa-ROS-based communication system. It complements those research attempts by providing a prototype implementation of a LoRa-based communication system, developed for both indoor and outdoor operations. The main differences between this work and our previous research attempts are (i) the network topology configuration (i.e., the UAVs in our previous works were able to communicate with each other and the GCS using a mesh network topology), (ii) the extended communication coverage (i.e., the maximum distance between the UAV agents and the GCS in the previous study was limited to 800 m), and (iii) the communication architecture of the proposed system (i.e., our previous research efforts integrated the Robot Operating System (ROS) with LoRa to acquire telemetry data directly from the UAV, rather than relying on the UAV’s onboard LoRa hardware for data acquisition, as done in this work). Also, in our previous works all the LoRa parameters were kept constant, while in this study a thorough evaluation of the parameters has been conducted to assess their impact on network performance based on numerous outdoor and indoor experiments. Contrary to other works, the proposed evaluation method demonstrates the applicability of these networks in realworld environments, where factors such as physical obstacles, dynamic UAV movements, and interference are present, in contrast to simulated environments or theoretical approaches [10], [11]. A variety of LoRa configuration parameters is also considered collectively, rather than individually [12]– [15], to achieve the aforementioned results and ensure a more comprehensive evaluation. III. LORAPARAMETERS In general, the communication performance of the LoRa protocol is heavily influenced by various key configuration parameters (i.e., OP, BW, SF, and CR). These parameters have a significant impact on data rate, communication range, and power consumption, and can be configured based on each application’s requirements [3], [4]. (i) Output Power (OP): Amount of power that a LoRa device utilizes to send signals. Higher transmission power allows for increased communication range, albeit with the cost of increased power consumption. [22]. (ii) Bandwidth (BW): Spectrum of frequencies designated for data transmission (proportional to the bit rate). A finetuned BW parameter increases transmission rate, which is significantly more durable to noise. LoRa typically operates at 125,250, or 500 kHz [5], [22]. (iii) Spreading Factor (SF): It determines the rate at which data is transmitted over the air. For LoRa devices, it is expressed as a numerical value ranging from 7to 12. Higher SF indicates increased communication range and improved signal sensitivity but leads to decreased data rate [9], [22]. (iv) Coding Rate (CR): It determines the extent of error correction applied to the transmitted data. Higher CR enhances resilience to signal interference and improves the chances of successfully decoding the transmitted message at the cost of reduced data throughput. On the other hand, lower CR provides higher data rates but may result in decreased reliability in challenging radio frequency environments [9]. (v) Received Signal Strength Indicator (RSSI), Signalto-Noise Ratio (SNR) & Transmission Rate (TR): RSSI measures the power level of the received signal, with a higher RSSI value implying a stronger signal, while SNR measures the ratio of signal power to noise level in the communication channel, with a higher SNR denoting improved signal quality. Further, TR represents the speed at which the data is transmitted over a communication channel [measured in bits per second (bps)]. In essence, TR describes how quickly information can be exchanged between devices (e.g., from the Tx to the Rx) [6] and can be calculated as TR =P ayload size (bits) T ransmission T ime (sec), where Transmission Time = Time of Reception - Time of Transmission. IV. SYSTEM ARCHITECTURE AND IMPLEMENTATION A. System Overview To conduct a comprehensive investigation of the LoRa network’s performance, multiple configurations are designed to emulate operations for different scenarios and at different scales, including both indoor and outdoor setups with static and mobile nodes, respectively. Nodes are strategically distributed across designated areas in and around the University of Cyprus premises, with a ground control station (GCS) maintained at a fixed location for real-time monitoring. The GCS emulates the command center in emergency operations, while the nodes represent first responders or victims. Off-the-shelf embedded boards are employed as the transmitting and receiving nodes, with the boards being equipped with LoRa transceivers, GPS modules, onboard environmental sensors, and storage capabilities. For the indoor experiments, the nodes are static, while in the outdoor experiments the nodes are mounted on UAVs to increase coverage area and enhance mobility. Without loss of generality, in all experiments, the GCS remains static and acts solely as a receiver. To meet the emergency scenarios requirements, UAVs equipped with LoRa devices (to enable long-range, low-power communication over vast or obstructed terrains) can be deployed to offer a highly effective solution for communication and data transmission in environments where traditional communication infrastructure may be unavailable or compromised. LoRa nodes within the system can exchange positional and environmental data. Specifically, the data packet from each node includes the following information: ID, date (dd/mm/yyyy), time (hh:mm:ss), latitude (6decimals), longitude (6decimals), altitude (1decimal), temperature (2decimals), pressure (2decimals), and humidity (2decimals). Upon reception of a data packet, the recorded RSSI and SNR values are also included for subsequent analysis. B. Hardware Implementation For the experimental setup, as illustrated in Fig. 1, two DJI M300 UAVs are utilized, each equipped with a LilyGo TBeam Supreme board [23]. These boards include an ESP32S3 microcontroller unit (MCU) and an SX1262 LoRa module, as well as a GPS unit and a BME280 environmental sensor. LoRa firmware for each board is programmable and provides flexibility to configure the LoRa parameters based on the application scenario considered. At the GCS, another LilyGo T-Beam Supreme board is connected to a computing device via serial communication. The board collects data from the rest of the nodes, passes the data through the serial port on the computer, and then stores them in a local database. Figure 1. UAV hardware setup. C. LoRa Firmware Implementation Throughout the experiments, a LoRa star topology is implemented to streamline the process, utilizing the software from the RadioLib library [24], that is specifically designed for wireless communication in embedded systems. To optimize hardware capabilities, code from the LilyGO-Lora-Series [25] is also integrated. Focus is placed on essential operations of the boards, including GPS, environmental sensors, inertial measurement unit (IMU), and LoRa communication. Initially, firmware configuration parameters are set to their default values. These parameters are easily programmable, allowing for subsequent adjustments and reconfiguration. Updated firmware is flashed to each board prior to each experimental procedure. V. EXPERIMENTS PERFORMED A. Categorization of Emergency Scenarios As previously stated, during any emergency response scenario, the reliability and efficiency of communication systems is crucial. The indoor and outdoor experiments described below are conducted in order to provide valuable insights into how LoRa parameters can be optimized for different emergency situations. Specifically, during each experiment, one parameter is maintained constant and the others are varied, in order to gain valuable insights into their impact on performance. Additionally, the payload length and transmission intervals are kept consistent across all experiments. In essence, by systematically modifying the configuration parameters, specific settings that enhance performance under different conditions can be identified. The unique communication needs for different emergency scenarios are categorized into two distinct groups (Emergency Indoor Scenarios (EIS) and Emergency Outdoor Scenarios (EOS)): (i) EIS: Reliable and penetrating indoor communication able to reach through obstacles (e.g., building materials and debris in urban search-and-rescue operations); (ii) EOS: Long-range outdoor communication with non-line-ofsight (NLOS) conditions covering vast areas with minimal infrastructure, stable and consistent signal transmission, and immediate and reliable communication in unpredictable environments (e.g., for wildfire monitoring and management, flood response and monitoring, earthquake aftermath conditions). B. Experimental Setup In the indoor experiments, two transmitters are placed within the laboratory, with the receiver connected to a local computer acting as the GCS. One transmitter node is positioned at the same level (ground floor) as the GCS but in NLOS conditions, while the second transmitter node is placed on the first floor of the building, almost above the first transmitter node, with multiple obstacles present between the transmitters and the receiver. The presence of obstacles is crucial for testing purposes, as they can significantly impact the signal’s strength and quality, providing a more realistic assessment of the system’s performance in typical indoor environments. In the outdoor experiments, two UAVs are utilized that act as transmitters, while a single receiver is connected to the GCS. To ensure consistency, the UAVs follow the same path for each outdoor experiment. Initially, they are placed at a horizontal distance of 800m from the GCS and they progressively increase their distance, reaching a maximum distance of 1100m. UAVs 1and 2are flying at constant altitudes of 80 and 90m, respectively, while the distance amongst them is also kept constant to avoid collisions. Each indoor experiment typically lasts between 1.5to 2 hours, whereas outdoor experiments last around 15 min, due to the limited flight time of the UAVs. The default configuration of the parameters is set as follows: OP=22 dBm, BW=125 kHz, SF=9, and CR=4/7. Further, each board transmits a data packet with an average length of 65 bytes every 2seconds throughout all experiments. VI. PERFORMANCE EVALUATION A. Evaluation Metrics Even though each category requires different communication performance criteria, uniform metrics are employed for the system’s performance evaluation, i.e., RSSI, SNR, and TR. Packet loss (PL) is also computed, providing insights into network reliability across all experimental scenarios. PL is calculated as the difference between the expected and actual number of data packets from all nodes over the specified duration of the experiment (in seconds), i.e., PL=(RXe −RXa RXe)·100. Additionally, the fraction of packet corruption (PC) in received messages is calculated. This metric is significant as it provides crucial information for the reliability and integrity of the communication system under test. In this case, all messages are considered collectively rather than individually for each node, as it is not possible to identify the source node of the corrupted message. Thus, PC is calculated based only on RXa from both nodes and is determined as PC= RXc RXa . By combining the number of lost and corrupted packets (that is, RXe-RXa and RXc), the amount of unusable data received is identified as total loss, LT=(RXe −RXa)+RXc RXa . Finally, reception latency is calculated (LRX ) to assess reliability in scenarios requiring immediate communication. Table I INDOOR EXPERIMENTS’RESULTS Conf. Parameters (BW, SF, CR) RSSI (dBm) SNR (dB) PL (%) PC (%) LT (%) TR (bits/sec) 125, 9, 7 -83.39 0.3 13.18 2.74 15.56 3120.88 250, 9, 7 -79.0 0.42 16.64 4.08 20.01 4437.77 500, 9, 7 -84.42 -1.38 11.01 7.76 17.92 4300.83 125, 7, 7 -78.98 3.29 13.78 5.67 18.67 3816.08 125, 9, 5 -80.11 2.29 37.73 5.19 4.96 4151.71 125, 9, 8 -77.04 6.09 5.95 0.14 6.08 4229.21 RSSI and SNR are investigated both individually for each node and collectively (across all nodes). Firstly, RSSI and SNR measurements for each node are investigated separately, taking into account variations in communication range and environmental conditions. The analysis is based exclusively on non-corrupted packets, as only these packets allow us to identify the originating node. The investigation, which includes RSSI and SNR measurements from both nodes (i.e., the average is calculated), also incorporates corrupted packets, as the originating node does not affect the analysis. In this case, degraded RSSI and SNR performance is expected, as the corrupted packets are now also taken into account. To simplify the presentation of the experimental results, the average RSSI and SNR values for each node, as well as the averages from the combined analysis, are used to present the performance metrics. In particular, 6indoor and 6outdoor experiments are conducted using the described setup. As a result, the average values for each node are calculated individually for each experiment and then followed by the average between the two nodes for each scenario. B. Performance Results 1) Indoor Experiments: For the indoor experiments, performance degradation is expected, particularly for the node placed on the first floor, due to interference between the transmitter and receiver. In the first experiment, the OP parameter is examined, showing that any value below the maximum (i.e., 22 dBm) is ineffective. Consequently, only that OP value is used for further analysis for all experiments (including outdoor), allowing uniform comparison throughout all setups. To investigate the impact of the BW parameter, the default value is set as BW=125 kHz and subsequently 250 and 500 kHz values are also investigated, while to examine the impact of the SF parameter, the default value is set to 9and the next values chosen are 7(minimum) and 12 (maximum). Generally, it is shown that high data extraction rate (DER) is indicative of low collision packet rates and good network behavior. Even though higher SFs (9-12) offer greater coverage distance, a decreased DER is expected at high device densities due to increased packet collisions and extended TOA [26]. Finally, the CR parameter is also examined to ascertain its impact on data quality (i.e., increasing CR leads to higher SNR but the transmission time also increases). For the LoRa module employed, the default value is set to 7and subsequently set to 5(minimum), and 8(maximum). In essence, by focusing on CR with values of 5,7, and 8, the outdoor field tests are expected to demonstrate a trade-off between data reconstruction reliability and error correction capabilities [27]. The results depicted in Table I indicate that all SNR values are sub-optimal (i.e., SNR <10 dB). Given the extent of penetration through the building, such results are anticipated. Nevertheless, amongst the three tested BW values (even though their performance differences are small), BW=125 kHz is the preferred BW value, as it yields the lowest LT, and, although the SNR value remains sub-optimal, it offers the best performance considering the balance between PL and SNR. Further optimization may be needed to enhance SNR without compromising PL rates. Further, a lower SF (i.e., SF=7) results in better SNR performance, which is crucial for mitigating noise, e.g., originating from building debris. However, total packet loss is slightly higher when compared to a higher SF. Such observation aligns with the theory that higher SF improves signal sensitivity, reducing packet loss. Thus, a trade-off exists between achieving optimal SNR and minimizing packet loss, necessitating a balance in SF selection for specific applications. Moreover, the CR parameter investigation provides valuable insights. A high CR value (i.e., CR=8) outperforms other scenarios in terms of both SNR and LT. This can be attributed to the enhanced resilience to signal interference provided by higher CR values. The improved performance in both metrics suggests that increasing CR is an effective strategy for enhancing communication quality, especially in environments with significant interference. Besides the previously mentioned metrics, the calculated TR also shows relatively good results for all scenarios. Based on all aforementioned results, it is concluded that the final combination of configuration parameters — BW=125 kHz, SF=9, and CR=8— is a highly suitable choice for indoor emergency scenarios where interference and noise are prevalent. This configuration allows an optimal balance between SNR and total packet loss, ensuring reliable communication in challenging indoor environments. Table II OUTDOOR EXPERIMENTS’RESULTS Conf. Parameters (BW, SF, CR) RSSI (dBm) SNR (dB) PL (%) PC (%) LT (%) TR (bits/sec) 125, 9, 7 -96.04 -9.25 39.08 8.18 44.06 2488.46 250, 9, 7 -95.69 -9.58 27.57 9.14 34.19 3046.25 500, 9, 7 -95.43 8.08 35.84 15.81 45.98 4910.85 125, 7, 7 -89.34 -0.11 27.80 27.23 47.47 3092.86 125, 9, 5 -89.96 2.13 1.53 1.73 3.23 3116.63 125, 9, 8 -89.64 -0.06 28.62 6.77 33.45 4119.71 2) Outdoor Experiments: Table II presents the results for the outdoor experiments. As expected, the general performance of the system decreases with larger distances and increased interference. This is evident from the higher values of LT, compared to the indoor experiments. Nevertheless, there is a notable improvement in RSSI values in outdoor conditions. Specifically, for the BW parameter, the optimal performance (averaged over all experiments) in terms of LTis observed at BW=250 kHz, while in terms of SNR, BW=500 kHz provides the best results. Additionally, it can be seen that the difference between the three BW values is now notable. Nevertheless, at larger distances (e.g., >1km), that are of interest for outdoor scenarios, communication at BW=125 kHz outperforms the rest of the configurations in terms of packet Figure 2. PL vs. distance from the GCS for varying BW values. Figure 3. PL vs. distance from the GCS for varying SF values. loss, as depicted in Fig. 2. Further, in terms of SF, as can be seen from both Table II and Fig. 3, a higher SF performs better especially over larger distances in terms of PL and consequently, LT. Moreover, Fig. 4, presents the transmission intervals for each combination of parameters, highlighting that, although there are variations in performance with respect to losses, the transmission intervals remain quite similar across the different scenarios (i.e., information is transmitted in a range of 1-4seconds, demonstrating an average latency of 2seconds). The calculated TR also demonstrates consistently good results across all scenarios. Finally, contrary to the results obtained from the indoor experiments, in outdoor experiments the lowest CR yields the best performance among all tested configurations, as it provides the lowest PC and PL (see Fig. 5). This observation can be attributed to the fact that the outdoor setup experienced minimal interference due to a clear LOS environment, which facilitated the achievement of higher data rates as necessitated by emergency outdoor scenarios. Figure 4. Moving average of the interval transmission frequency for the outdoor experiments. To extend the evaluation, a real-life emergency response operation is also conducted, which includes the search-andrescue of three missing people, and with the operation now located in a relatively open area with longer distances. For Figure 5. PL vs. distance from the GCS for varying CR values. (a) (b) Figure 6. Actual paths with LoRa messages received at different waypoints for (a) UAV 1and (b) UAV 2. this operation, the LoRa parameters used are OP=22 dBm, BW=125 kHz, SF=9, and CR=4/7. Moreover, the data that in this case each UAV transmits includes the “remote identification" packet, as required by the European Union Aviation Safety Agency (EASA). This packet comprises of the UAV operator registration number, unique serial number, timestamp, current latitude, longitude, altitude and heading (relative to north), geographical position (latitude, longitude, altitude) of the take-off point, and emergency status. As the message contains a maximum total of 106 bytes (rather than the 65 bytes used previously in the experiments), a degraded performance is expected compared to the previous experiments, due to the extended length of data being transmitted. Finally, in this experimental setup, the system is now integrated with our existing multi-drone platform (AIDERS platform - https://www.kios.ucy.ac.cy/aiders/aiders-ai-toolkit/) which rejects corrupted packets. Thus, PC and LTmetrics are not calculated. Also, SNR measurements from the nodes are not recorded. Consequently, only the average RSSI value for each node, as well as the overall RSSI average across both nodes, are calculated. Figure 7. Average packet loss vs. distance from the GCS for both UAVs. Figures 6(a) and 6(b), illustrate the actual paths along with the points where a message is received, for UAVs 1and 2, respectively. For these paths, Fig. 7 presents the average packet loss (for both UAVs) versus their distance from the GCS, while Fig. 8 presents the moving average of the interval transmission frequency. The results indicate that PL is significantly high across all considered distances, mainly due to increased interference caused by the presence of multiple active devices (e.g., control station wireless equipment) near the receiver being used during the operation. It should be noted that the missing values in Fig. 7 reflect a packet loss of 0%. Additionally, transmission latency is also slightly increased (in the range 2-5sec), as the distances between the UAVs and the GCS are extended (>1.05 km), compared to the previous outdoor scenarios. Finally, the degraded performance can also be observed from the average RSSI values of the two UAVs which is calculated at 77.48 dBm. Overall, the system showcases robust and accurate performance, thereby demonstrating its ability to be employed in both indoor and outdoor emergency response operations. The integration of LoRa with UAVs enhances coverage and extends communication range, especially in cases where the traditional network infrastructure (i.e., using Wi-Fi or cellular networks) is unreliable or inaccessible [18], [19], [21]. Figure 8. Moving average of the interval transmission frequency. VII. CONCLUSIONS This work demonstrates the feasibility, effectiveness, and robustness of an integrated LoRa-based prototype communications system for data transmission in UAV-to-X emergency response operations. Extensive experimental testing of this system in the field provides valuable insights on its performance in real-world scenarios. Results demonstrate that the system’s performance is based not only on the distance between the UAVs and the GCS, but also on the LoRa’s configuration parameters. Specifically, a certain configuration of parameters (BW=125 kHz, SF=9, CR=8) is a highly-suitable choice for indoor emergency scenarios where interference and noise are prevalent. Moreover, the outdoor experimental testing presents that a different set of configuration parameters (BW=125 kHz, SF=9, CR=5) is highly-effective for outdoor scenarios over long-range distances. Finally, examining a real-life search-andrescue operation revealed that additional latency and packet loss were observed, primarily due to larger distances and increased interference. This highlights the need to account for additional factors in operational scenarios to ensure the system performs as intended, necessitating the inclusion of performance margins. Future research includes the investigation of deep-learning frameworks and the employment of sensor data fusion techniques to improve communication performance. Further, research will also be conducted to enhance the proposed system’s security (e.g., using chaotic-based encryption [28]). REFERENCES [1] H. Jaldell, “How important is the time factor? Saving lives using fire and rescue services,” Fire Technology, vol. 53, no. 2, pp. 695–708, 2017. [2] R. S. Sinha et al., “A survey on LPWA technology: LoRa and NB-IoT,” ICT Express, vol. 3, no. 1, pp. 14–21, 2017. [3] J. P. S. Sundaram et al., “A survey on LoRa networking: Research problems, current solutions, and open issues,” IEEE Commun. Surv. Tutor., vol. 22, no. 1, pp. 371–388, 2019. [4] S. Devalal and A. Karthikeyan, “LoRa technology-an overview,” in Proc. ICECA, 2018, pp. 284–290. [5] C. Li and Z. Cao, “LoRa networking techniques for large-scale and longterm IoT: A down-to-top survey,” ACM Comput. Surv., vol. 55, no. 3, pp. 1–36, 2022. [6] P. Gkotsiopoulos et al., “Performance determinants in LoRa networks: A literature review,” IEEE Commun. Surv. Tutor., vol. 23, no. 3, pp. 1721–1758, 2021. [7] M. Karatzia et al., “Implementing mission-critical UAV swarm coordination through the integration of LoRa and ROS frameworks,” in Proc. IEEE ICT-DM, 2023, pp. 1–7. [8] N. Souli et al., “Mission-critical UAV swarm coordination and cooperative positioning using an integrated ROS-LoRa-based communications architecture,” Computer Communications, vol. 225, pp. 205–216, 2024. [9] M. Bor and U. Roedig, “LoRa transmission parameter selection,” in Proc. IEEE DCOSS, 2017, pp. 27–34. [10] E. Sallum et al., “Performance optimization on LoRa networks through assigning radio parameters,” in Proc. ICIT, 2020, pp. 304–309. [11] G. Ferré and A. Giremus, “LoRa physical layer principle and performance analysis,” in Proc. ICECS, 2018, pp. 65–68. [12] R. Sanchez-Iborra et al., “Performance evaluation of LoRa considering scenario conditions,” Sensors, vol. 18, no. 3, p. 772, 2018. [13] F. Turˇ cinovi´ cet al., “Analysis of LoRa parameters in real-world communication,” in Proc. ELMAR, 2020, pp. 87–90. [14] M. J. Faber et al., “A theoretical and experimental evaluation on the performance of LoRa technology,” IEEE Sensors Journal, vol. 20, no. 16, pp. 9480–9489, 2020. [15] E. D. Widianto et al., “LoRa QoS performance analysis on various spreading factor in Indonesia,” in Proc. ISESD, 2018, pp. 1–5. [16] A. Vazquez-Rodas et al., “Experimental evaluation of RSSI-based positioning system with low-cost LoRa devices,” Ad Hoc Networks, vol. 105, 2020. [17] G. Kaur et al., “An approach to optimize LoRa network performance for efficient IoT applications,” Wireless Personal Communications, vol. 128, no. 1, pp. 209–229, 2023. [18] S. Javaid et al., “Communication and control in collaborative UAVs: Recent advances and future trends,” IEEE Trans. Intell. Transp. Syst., vol. 24, no. 6, pp. 5719–5739, 2023. [19] P. D. Dai et al., “Efficient water quality monitoring using unmanned aerial vehicles and Internet of thing technologies,” in Proc. ECAI, 2024. [20] K. T. Pauu et al., “IRS-aided federated learning with dynamic differential privacy for UAVs in emergency response,” IEEE Internet of Things Magazine, vol. 7, no. 4, pp. 108–115, 2024. [21] R. Rukaiya et al., “Communication architecture and operations for SDRenabled UAVs network in disaster-stressed areas,” Ad Hoc Networks, vol. 160, p. 103506, 2024. [22] A. Augustin et al., “A study of LoRa: Long range & low power networks for the Internet of Things,” Sensors, vol. 16, no. 9, p. 1466, 2016. [23] “T-Beam SUPREME,” https://lilygo.cc/products/t-beam-supreme, 2024. [24] J. M. Solé et al., “Implementation of a LoRa mesh library,” IEEE Access, vol. 10, pp. 113 158–113 171, 2022. [25] “LilyGo-LoRa-Series,” https://github.com/Xinyuan-LilyGO/LilyGoLoRa-Series, 2024. [26] F. Cuomo et al., “EXPLoRa: Extending the performance of LoRa by suitable spreading factor allocations,” in Proc, IEEE WiMob, 2017. [27] I. P. Manalu et al., “Performance analysis of LoRa in IoT application of suburban area,” in Proc. IEEE ICT, 2023, pp. 1–4. [28] N. Souli, et al., “Performance evaluation of a prototype UAV-based secure communication system employing ROS and chaotic communications,” in Proc. IEEE ICUAS, 2024, pp. 1034–1041.