LoRaWAN-Based Energy-Efficient Surveillance by Drones for Intelligent Transportation Systems

Urban networks aim at facilitating users for better experience and services through smart platforms such as the Intelligent Transportation System (ITS). ITS focuses on information acquisition, sensing, contrivance control, data processing and forwarding to ground devices via user-specific application-interfaces. The utility of ITS is further improved via the Internet of Things (IoT), which supports “Connectivity to All”. One of the key applications of IoT-ITS is urban surveillance. Current surveillance in IoT-ITS is performed via fixed infrastructure-based sensing applications which consume an excessive amount of energy leading to several overheads and failures in the network. Such issues can be overcome by the utilization of on-demand nodes, such as drones, etc. However, drones-assisted surveillance requires efficient communication setup as drones are battery operated and any extemporaneous maneuver during monitoring may result in loss of drone or complete failure of the network. The novelty in terms of network layout can be procured by the utilization of drones with LoRaWAN, which is the protocol designated for Low-Power Wide Area Networks (LPWAN). However, even this architectural novelty alone cannot ascertain the formation of fail-safe, highly resilient, low-overhead, and non-redundant network, which is additionally the problem considered in this paper. To resolve such problem, this paper uses drones as LoRaWAN gateway and proposes a communication strategy based on the area stress, resilient factor, and energy consumption that avail in the efficient localization, improved coverage and energy-efficient surveillance with lower overheads, lower redundancy, and almost zero-isolations. The proposed approach is numerically simulated and the results show that the proposed approach can conserve a maximum of 39.2% and a minimum of 12.6% of the total network energy along with an improvement in the area stress between 89.7% and 53.0% for varying number of drones over a fixed area.


Introduction
Intelligent Transportation System (ITS) includes strategies for traffic management, navigation, user-tracking and coordinated information processing for better services to the users [1,2]. ITS includes the coalescence of different technologies for achieving involute applications associated with wireless communication, image processing, audio-video monitoring, sensor and vehicular communications in an urban environment. Existing networks can be improved for performance by associating different devices as a transmitting and receiving component in ITS. Evolution of Internet of Things (IoT) has further improved the existing applications of ITS by connecting most of the devices with data-supporting   This paper demonstrates that LoRaWAN can be improvised for including drones as on-demand nodes to fortify energy efficient surveillance in ITS. Further, requisites such as fail-safe connectivity for battery-operated contrivances, high resilience, lower overheads, and lower resource utilization can be procured via LoRaWAN architecture. However, there are limited studies available at the moment, which have actually utilized LoRaWAN for designated application such as surveillance in urban scenarios. Thus, this paper, to the best of authors' knowledge, is the first to exploit the features of LoRaWAN-based drones for energy efficient surveillance in urban-ITS.

Problem Statement
Currently, most of the ITS systems use V2V or V2I communication models for sharing data required to build information of the entire geographical area. This information is transmitted as ad hoc formations and uses on-demand algorithms for data-sharing between the involved entities. However, the prosperity of information depends on the number of sensors deployed in a geographical area and the regions with lesser sensors or conveyances pass little information required for surveillance. Such a situation leads to an extremely high number of isolated areas where surveillance is impractical. On the contrary, urban surveillance can be conducted by the accumulation of sensors with the monitoring drones for understanding the current state of the transportation system. The coalescence of ITS with drones can support urban computing architecture and execute on-demand dissemination of data to the intended end-users. However, the involution of drones and their communication with the ground sensors are subject to energy overheads and may result in obstructed transmissions. Such a situation may expeditiously deplete all the resources of the system and result in a failed setup. It is required that surveillance should be energy efficient, scalable, non-redundant, low on overheads and non-isolated. Further, the surveillance should be conducted at the real time with a direct connection between the end-user and the deployed sensors. In addition, the communication system should be fail-safe and highly resilient in terms of processing and power consumption. Thus, it is required to develop a communication architecture which supports on-demand urban-surveillance via drones with energy efficient setup in ITS.

Our Contribution
The proposed approach is capable of supporting energy efficient surveillance in ITS by using drones as on-demand LoraWAN gateways. The proposed approach is able to enhance the lifetime of the network with a fail-safe and resilient strategy for using drones in an energy-deficit urban environment. The highlights and the major contributions of the proposed work are: • Incorporation of drones in ITS by using them as LoRaWAN gateways.

•
Drone modeling and traffic estimation in ITS.

•
Intelligent solution for location identification and coverage by drones.

•
Energy efficient surveillance and LoRaWAN-based drone communication.

•
Resilient factor and fail-safe conditions for drone-based transmissions. • Numerical case study for performance analysis of the proposed approach.
The remainder of the paper is structured as follows: Section 2 presents the related works on energy efficient surveillance, use of drones in networks, LoRaWAN models and sensing applications in ITS. Section 3 gives an overview of the network model, drone model, and traffic estimation. Section 4 provides insight into the proposed work with energy efficient setup for allocation, coverage, and surveillance. Performance evaluations are conducted in Section 5 . Section 6 presents discussions of the proposed approach with the state-of-the-art solutions. Finally, Section 7 gives conclusive remarks for the paper along with future directions.

Related Works
Surveillance in ITS is challenging because of a sizably voluminous number of users simultaneously probing for information that causes many overheads. Applications associated with surveillance such as traffic monitoring, location-identification, device tracking, etc, are further affected by the interference caused by the sensors or the locators deployed in an urban setup. These issues of overheads and interference due to overlapping result in wastage of energy resources and a decrementation in the lifetime and resilience of the network. Over the last few years, there have been several solutions proposed by leading researchers from all over the world. Some of them has additionally accentuated on utilizing on-demand drones as a key node (gateway) in the network setup akin to the one considered in this paper, but with divergent perspective and approach. In this section, some of those solutions are discussed that have the potential of being utilized for surveillance in ITS.
Motlagh et al. [40] proposed an Unmanned Aerial Vehicle (UAV)-based IoT platform that can be used for data offloading in mobile edge computing. The authors focused on vision-based data offloading by saving the operational energy of the drones. However, at the moment, mutual tasks and energy consumption during the simultaneous applications are yet to be explored for their proposed solution. This is a limitation as surveillance is a mutually cooperated task that cannot be resolved by direct application of their proposed solution. Wu et al. [41] developed a data dissemination model for drones in urban networks. Their developed approach uses online learning techniques to adaptively balance the broadcast rate and knowledge loss rate while communicating with each other. Their approach is efficient and can be used for enhanced lifetime-based network formation. At the moment, their approach emphasizes much on the use of onboard sensors for collecting information on urban environment and issues related to survivability and energy efficiency are yet not evaluated.
Naqvi et al. [42] discussed the drone-enabled communication for public safety networks. The authors discussed various challenges and issues involved in the utilization of drones for futuristic communication. The authors concluded that the self-adaptive power control of drones depends on the interference and data rate requirements, which are used as key metrics in the proposed work of this paper. Menouar et al. [43] discussed the challenges associated with the use of drones in ITS. The authors discussed that data rate and coverage play a crucial role in surveillance by drone in ITS, which is affected by obstacles and require rigorous path planning and location-allocation strategies.

Network Model
This section presents the network model used as a base for defining the proposed approach. The network comprises all the core components of the LoRaWAN, as shown in Figure 2, along with additional features as explained below: • Application Server: The application server plays the role of central authority and regulates the data flow in the general LoRaWAN. However, in the proposed model, it is used for delivering applications related to surveillance to the end-users. It supports downloading by an interface, which is classified into sensor App, drones App or network App. Every end-user in the network is assumed to possess all the three Apps that help to communicate with each other.
• Network Server: The network server is the connecting entity between the application servers and the LoRaWAN gateways. A series of network servers helps to expand the network and manages connectivity with the public/private cloud infrastructures. It is capable of managing transmissions with the data centers via real or virtual nodes. Network servers are connected through series of routers with information flow management only at the application server. In the proposed setup, the network servers are used as assigning entity that also checks the performance of the end-user devices, sensors, and actuators. • LoRaWAN Gateways: The LoRaWAN gateway operates just like any other gateway and provides a bridge between the network servers and the end-users, sensors or locators. Unlike the traditional LoRaWAN gateways with fixed roles, the proposed model uses on-demand drones as LoRaWAN gateways that provide direct connectivity to the end-users. At the moment, the LoRaWAN gateway is divided into two components, which include a set of drones and a switch. The switch helps to localize the drones in the network and checks for incoming connections from other drones. Generally, a network may have switches for different service providers and all the drones respond to their particular switch. Information broadcasting and sharing between the sensors and the end-users via drones is independent of the service provider and depends on the accessibility roles and conditions between them. • Sensors or Locators: Sensors or locators are the fixed devices in ITS which helps in gathering information across its transmission range and share with the nearest requesting end-user. Sensors or locators also communicate with the drones for long-distance surveillance. This helps in gathering information from any isolated areas. In case, there is no sensor or locator information available, surveillance drones can directly be used for transmissions and accessing geographical information.

•
End devices: These are the users which take services from the entire setup and use the information to know the current state of the transportation system. The usability of information by end-users depends on the type of Apps which are used for evaluation as well as the type of services being provided by the service providers. End-users have pre-installed applications that are provided by the centralized application server. It is to be noted that in the proposed approach, the centralized procedures between the end-users and the application servers are only carried during the network initialization. Once a device is registered, there is a limited role of the application server for information accessibility. This makes LoRaWAN on-demand and scalable.
Mathematically, the network can be expressed as a single application server operating with a set N of network servers, which are connected to a set D of data centers and a set C of cloud servers. Further, the network servers are served by a number of drones represented by a set U . The number of switches is denoted by a set W that may vary for each application server. As explained below, a separate network server is used for each service provider. The entire ITS setup is served by the combination of sensors or locators denoted by a set S, and the end-users are represented by a set E . The details of symbols used in this paper are presented in Table 1. The communication between the drones and the sensors are subject to cause interference with the signals for the end-users. This interference over a particular signal, observed by i th end-user, is computed as signal to interference plus noise ratio, i.e., where T p is the transmission power, K α is the antenna constant for a given path loss exponent α, H is the transmission range and V is the spectral density of the network for a given noise model. Inter-drone communication is assumed to be free from noise and a separate beamwidth is used for the drone to drone communication. It is to be noted that each drone is equipped with a LoRaWAN-antenna as specified for its ISM band. In contrast to drones in 5G setups, LoRaWAN-enabled drones operate on a narrow ISM band in the range of 902-928 MHz (US) or 863-870 MHz (Europe) [61]. Thus, minimization of interference can help in better signal strength to each user or sensor/locator. The entire setup is driven by the received power at a drone as expressed in [61,62]. The proposed approach uses a combination of transmission, which includes a drone and an entity (end-user/sensor or locator), or a sensor/locator and the end-user itself. Both these are driven by Line of Sight (LoS) and Non-LoS (NLoS)-based transmissions. Such transmissions and their signal strengths depend on the received power of the devices. Now, using [62,63], the received power for LoS and NLoS are calculated as: and respectively, where Here, δ k is the path, δ 0 refers to the transmission distance, C S is the shadowing component, and L 0 is the path loss at the reference distance. The received power at a respective location for a particular device helps to maintain a reliability of connectivity R (C) E in the network. This reliability is calculated by defining received power thresholds, which are denoted by P E is calculated for the independently operating components of the network that shares a wireless connectivity, i.e., drones, end-users, and sensors. Each of these devices is considerably operational on multiple channels and may or may not possess redundant connectivity for each of them (In this paper, it is assumed that all the channels over a single entity are identical). Thus, reliability, in any given instance t for X number of concurrent channels, is calculated as: where < entity > refers to drones, end-users or sensors. If, each entity is operational in multiple states or tested after G number of states, then, such that E ,G,<entity> , because state-wise reliability is always less than or equal to the overall reliability of the network. F <entity> is the failure factor, which is calculated as [64]: where f p is the frequency of received power of an entity being lower than the threshold values.
It is calculated as the ratio of instance when P R,TH for LoS mode and NLoS mode, respectively. Equation (7) can also be defined for link availability L A , such that F <entity> (t) = 1 − e −(1−L A ) , where L A = 1 − , and is the outage probability. Now, by definition, failure distribution function for a particular duration, say τ 1 and τ 2 , can be calculated as: This model can be extended for probability of connectivity P (e) C , which can be defined as a function of network load. This load is dependent on the arrival rate λ of incoming packets. For this, P (e) C = f (λ) that can be modeled for r , which is the required rate for the network, and λ , which is the minimum rate that can prevent congestion. From these, P (e) C can be given similar to Equation (8), such that, P (e) , λ ≤ λ. Now, considering that t is the actual time for which the network should be operational to support realistic data transmissions between the end users via LoRaWAN architecture, the probability of failure before this time, is calculated as [65]: where t 0 is the available time equivalent to current time (t 0 = t), and M is the total connected device with the entity which have a high depletion rate. Using these, the expected future lifetime is given as [65]: A , . . . , U (|U |) A are zones for |U | drones. This condition holds as each drone may overlap some of the area covered by another drone. It is to be noted that each drone should be active for a time period t x , such that, there exists a full-time connectivity between the drones and the end-users or the sensors, and between the drones and the LoRaWAN gateways (Backhaul for the drones and the LoRaWAN switches is not evaluated in this paper). Each zone has a particular number of sensors represented by ζ i , where i is the zone number, such that |S| =  (20)). E (0) <entity> is the initial energy of an entity, and E (t) <entity> is the remaining energy of an entity after a time instance t. This same model can be used to calculate the expected number of drones required for a particular zone by trivial calculations for the number of supported uplinks/downlinks on a drone and the requested uplinks/downlinks from a ground-zone.
All the above-described communications are governed by the wait time τ p of a drone over a particular location in the defined range of ground entity. This τ p is measured as V u R u , where V u and R u are the velocity and range of drone transceiver, respectively. τ p can be used to identify the location of a drone at any given instance as well as it can be used to predict network state after a given number of iterations, such that,

Traffic and State Estimation
This setup can be formulated into the number of steps required for predicting the state of the network after a required time. These states are marked as transitions between the zones into which the sensors are deployed. Such an overview of predicting network state via drone modeling and traffic conditions is presented in Figure 3 (This same model can be extended for any type of ground entity, be it the LoRaWAN switches, end-users or even other drones). The details of network state prediction are as follows: At first, the network is accounted for the number of allocated drones. These are checked for all the zones with communicating entity. The communicating entity can be a sensor, end-user, or another drone. (ii) Next, the evaluations are conducted for required transmission time by using the above-given formulations. This helps to understand the exact duration for which each drone must be active over all the given zones. (iii) Next, the details of incoming drones are marked. This is done by understanding the directional maneuvering of each drone and its present location. (iv) The incoming drones are useful only if they possess a certain number of free channels to support extra sensors from the zones they are entering.
Once the evaluations are done for the number of channels supported by incoming drones, the number of left out devices can be identified on the basis of no connections. Note that failed communications and no-connectivity also refers to a state of being left out. (vi) This helps to calculate excessive drones required to maintain the connectivity to all and that too at all the instances. (vii) Now, the excessive requirements for channels can either be resolved by deploying more drones or by understanding the maneuvering cycles of already deployed drones.
The maneuvering cycles of incoming drones are also identified as this identification helps to understand the state of drones which will leave the current zone during one set of communication (uplink or downlink). (ix) These cycles help to identify the network delay as well as the allocation delay, which tells about the time required by a drone to reach a particular zone for connectivity. (x) Next, the number of current transitions on each drone is logged and the current traffic is marked. (xi) Finally, the remaining traffic is calculated with a stipulated delay after a given number of states. This helps to decide the performance as well as the state of the network by considering new deployment or by improving the maneuvering cycle of each drone.

Proposed Approach
This section proposes a LoRaWAN-based drone transaction model which helps in fail-safe network formation with high resistance towards overheads. The proposed approach uses energy efficiency function for controlling the connectivity and maintaining a non-redundant as well as non-isolated surveillance. The proposed approach first uses location identification procedures to account for the location of each drone in the network. Then it evaluates the model via energy conditions and forms the drone transactions for efficient communication. The details are presented in the following subsections.

Location Identification and Coverage
The proposed approach uses area-stress modeling, inspired by stress-mechanics [66], for calculating the location as well as the load of the network w.r.t. deployed number of drones. For this, the area stress is calculated as: where f N (e) S is the function for network stress, which is calculated as [67]: such that Now, the resultant value is obtained as: where X 1 and X 2 are the number of entities operating in a given drone's zone and the actual number of entities a drone can support, respectively. R average is the average transmitting range of each ground entity and N (e) where Here, O and D are the number of overlaps (The coverage area is considered as a circular ring for each drone) and the distance between the centers of overlapping zones for any drones, respectively. The entire system can be used for location identification of drones in the LoRaWAN setup and to decide whether new drones are required or surveillance can be completed by utilizing the existing infrastructure. The details of these procedures are presented as steps in Algorithm 1. The algorithm operates by taking inputs from the network components as well as the operational time of each entity. Next, it initializes all the components and checks for continuity while finding coordinated information, incidence and adjacency matrices. After these computations, a decision on reshuffling or reallocation is taken by evaluating the area stress of the network. The algorithm is operated continuously until the required communication is not halted. The complexity of the algorithm depends on the number of drones and is equivalent to the complexity involved in building incidence and adjacency matrices for communication and collision avoidance.

Decision Modeling for Energy Efficient Surveillance
The proposed approach performs surveillance by mapping energy requirements and overheads associated with the zone, which comprises sensors or end-users looking for uplinks or downlinks. The proposed approach takes surveillance as a decision problem and uses energy-conservation mechanism for improved connectivity and long-term transmissions. The entire system can be modeled into energy requirements of each link formed in the network. Let Y (U ) e be the energy consumed by each drone to remain in-flight during transmissions. Now, the energy requirements for each zone can be calculated as: and the hazard functions will be given by respectively. Note that for efficient communication during any given instance, the network must fulfil the following conditions: By using the above defined model, the proposed approach is able to select the nodes, which will result in highly resilient and fail-safe network over the properties of LoRaWAN. The selection of nodes can be further understood from the steps presented in Algorithm 3. The algorithm helps in creating a subset of nodes and links that have very minute likelihood of failing on the basis of present depletion rate. The algorithm is iterative and continuously monitors the network for any changes in properties and the defined system model. Further, by using its steps, every node can maintain a connectivity with the above maneuvering drone leading to zero-isolation and fail-proof connectivity over LoRaWAN properties and infrastructure. Calculate ω (P ) c using Equation (27) 12: Calculate

Performance Evaluation
The proposed approach is analyzed for its performance over a scenario that considered hovering drones in a smart city environment through system simulations in Matlab TM . The drones are assumed to be transmitting traffic locations over LoRaWAN connections. The proposed solution is tested on an area of 2500 × 2500 m 2 with number of drones varying between 10 and 20. A total of 10 zones is considered each having number of sensors equal to 50 and the number of users less than or equal to 500. The drones are set at a cruising speed of 30 kmph and can operate continuously for whole simulations provided that the depletion rate of resources is not too high. The other parameters and their values used for analysis are provided in Table 2. Initially, the network is set as per the configured entities and Algorithms 1 and 2 are used for allocation of drones for the given number of sensors and users. The proposed setup operates with an average degree of connectivity varying between 1 and 6 for each of the users that have at least one of their channels operational all the time. The plot for the degree of connectivity for an active number of users is presented in Figure 4. This plot helps to understand the connectivity of the network and also assists to predict the traffic which will be generated once all the users are online and transmit through the same network.
The proposed approach operates for defining a system which can withstand the low bit rate pressure of LoRaWAN as well as can ensure high reliability of connections with less redundancy and almost zero-isolation. Figure 5 presents the plot for reliability for time-based as well as state-based network with varying frequency of received power to its thresholds. The thresholds are set at a range equivalent to an average of the minimum and mean value of received power during one cycle. It is evident from the graph that with increasing limits for thresholds, the reliability of the network decreases, as with time the received power for each entity in the network decreases. This decrease is due to the depleting energy for each link operating from a zone of users and sensors. However, the reliability lowers to a minimum value only if the f p reaches its maximum ratio of 1, which is a harsh scenario and usually, all the networks are recharged for their resources before reaching this limit. It is observable that the proposed approach can offer high reliability in a scenario where drones are used as LoRaWAN gateways. Also, the trends in the reliability graphs justify the conditions imposed on Equation (6). Thus, verifying the defined system model.  The entire model is efficient if it imposes lower stress on its resources. This is calculated in terms of area stress in Equation (12). The area stress is directly proportional to the network stress and the number of overlaps in the network (O). The number of overlaps causes excessive interference in the network and wastes much of the network resources in counterfeiting the effects of signals from other simultaneously communicating entities. From the density plot in Figure 6, it is noticeable that as the area stress of the network follows a logarithmical increase, the entire network observes a high number of overlaps. Further, this graph can be reversely studied, which means that the area stress of a network will increase with the increasing number of overlapping zones that will lower the transmission rate of the network and in lieu, the reliability of the entire system will decrease. There are certain durations in the network when the network stress is unaffected by the increasing number of overlaps. Such a situation includes more drones than normal setup over a single communicating zone.
From the system modeling, it is understandable that the transmission range plays a crucial role in sustaining of the network and its survivability, which is studied in form of the resilient factor in Equation (26). A network with better connectivity, lower consumption of energy, high yield in terms of remaining energy, better capacity and coverage area, possesses high resilient factor along with the reliability. Further, these values increase with an increase in the radio range of drones as a single drone can serve multiple users from different zones provided that the incoming requests are within the limits of connections supported by it. The results for reliability and resilient factor with an increasing radio range are presented in Figure 7. The resilient factor can attain a high value depending on the factors controlling it. The values can also be presented as normalized output by controlling the range of η 1 , η 2 and η 3 . However, at the moment, the real values are included for actual evaluation of the system.  The proposed approach is analyzed for its energy consumption throughout the transmission. With an increasing number of overlaps, the energy wastage increases, which decreases the overall available energy of the network as shown in Figure 8. The results in the figure show the confidence bounds at 95% interval and illustrate that a generalized quadratic equation can be used to analyze this decrease in available energy with an increasing number of overlaps. However, this lowering can be improved by subdividing the zones or by deploying more number of drones at varying altitude. Such an approach can lead to a longer duration of the network, but at an expensive operational cost.
Apart from the above results, the system is also tested for results with the varying number of drones as shown in Figures 9 and 10. The plots in Figures 9 and 10 illustrate the outcomes for area stress and available energy of each link, respectively, for a different set of drones. The result for area stress shows that the setup with 20 drones causes 89.7% and 53.0% lower burden on the network than the setups with 10 and 15 drones, respectively. This suggests that increasing drones can help managing network connectivity as well as support multiple LoRaWAN gateways than can support surveillance even at a lower bit rate. Further, the per-link available energy of the network increases with an increase in the number of LoRaWAN gateways (drones). It is noticed that a network with 20 drones observes 39.2% and 12.6% more available energy for each active link in comparison with the networks operating with 10 and 15 drones, respectively. The decrement in the observed values in Figure 10 even for a higher number of drones is due to reallocation and reshuffling operations conducted through Algorithms 1-3. Further, it is noticed that the proposed approach operates with a no failure probability (ω (P ) c ) equal to 0.99, which is sufficient enough for demonstrating the capabilities of the proposed model. Thus, from the results presented in this paper, it is evident that the proposed approach of using drones as LoRaWAN gateways for energy-efficient surveillance in ITS provides a fail-safe and highly resilient connectivity even at a low rate of transmissions. Further, the observed overheads and energy consumptions are lowered with no-isolations and low-redundancy even at the higher overlapping of communicating zones.

Comparison with State-of-the-Art Solutions
The proposed approach provides fail-safe as well as resilient connectivity between the sensors and the drones during surveillance leading to high conservation of energy. The proposed approach uses drones as LoRaWAN gateways and allows extensive monitoring, coverage and communication with the underlying users. With the help of area stress metric, the proposed approach is able to identify the amount of energy that will be consumed by the involved entities during their operations. The novelty of the proposed approach is depicted in terms of the network model, energy-efficient surveillance and low-failure probability. Over the years, drones or UAVs have been exhaustively used for surveillance but there are limited studies which actually emphasizes the role of drones in ITS. Also, the ones with energy efficient architecture, are unable to support fail-safe communications, and the others with effective surveillance are unable to provide low-power and long-range support as that of LoRaWAN setup. Further, there are limited works that exploited the features of LoRaWAN setup for surveillance purposes.
Most of the existing works focused on capacity, scalability, cooperation and task allocations for drones in ITS [69,70], but advancement to energy-efficient surveillance is little. Despite limited works on the applications of LoRaWAN, some of the key solutions which emphasize on surveillance are identified and compared with the proposed approach as shown in Table 3. The comparisons are drawn for different parameters such as ideology, use of drones, support for cellular and wireless communications, surveillance, energy-efficient deployment, resilience, fail-safe operations, and use of LoRaWAN features. From these comparisons (to the best of authors' knowledge), it is evident that none of the existing works has utilized the features of drones as well as LoRaWAN for energy-efficient surveillance, and the proposed approach is successful in supporting energy-efficient, fail-safe and highly resilient surveillance by drones for ITS.

Conclusions and Future Remarks
Energy efficient surveillance in IoT-ITS is a major issue which needs effective strategies as most of the devices involved in monitoring are battery operated. Majority of the devices involved in urban-surveillance in IoT-ITS are always at an edge of depleting resources that may result in the complete failure of the network. To resolve such a problem, a drone-assisted strategy was proposed in this paper which utilized the concept of LoRaWAN and its architecture for conserving energy via an efficient deployment of drones between the end users and the ground sensors. The proposed approach emphasized on using drones as LoRaWAN gateway and support communication without letting it travel to the application server, thus, saving unnecessary dependence on a centralized server. Further, drone modeling and traffic estimation were performed over the defined network model to form a base for energy efficient connectivity. Next, an intelligent solution for location identification and coverage by drones was proposed followed by energy efficient surveillance and LoRaWAN-based drone communication. Area stress, resilient factor, and energy conservation were the key metrics used for the operations of the proposed approach. The numerical results suggested that the proposed approach can conserve a maximum of 39.2% and a minimum of 12.6% of the total network energy and can improve the area stress by a maximum of 89.7% and a minimum of 53.0% for varying number of drones over a fixed area. It was also observed that the proposed approach can support continuous connectivity with no failure probability as high as 0.99. Thus, it can be concluded that the proposed approach is capable of providing energy efficient surveillance by using drones as LoRaWAN gateway in energy-deficit urban IoT-ITS.
At the moment, the proposed model does not account for the backhaul formed with the drones and LoRaWAN switches. This is marked as a future work and will be developed as an extension to the proposed work. Communication strategies and protocols for data dissemination over the defined network model are also required. Being energy constrained setup, the LoRaWAN infrastructure can be tested with different types of hardware that can facilitate the practical utilization of drones for LoRaWAN-based urban surveillance.