The advancement of the Internet of Things (IoT) has paved the way for new requirements for smart cities. This can be represented by a flexible/adaptable operation and an efficient data monitoring system for various real-world applications that may differ in the degree of mobility and performance quality, such as environmental monitoring and intelligent transportation in a smart city [1,2]. IoT-based wireless sensor network (WSN) events may, to some extent, lack flexible network operations, such as dynamic network events that are transient and require a real-time adaptive process. The flexible orchestration and reorchestration of such demanding dynamic networks plays an important role in monitoring the operation of the dynamic physical environment. Newer data gathering methods, processing, and communication frameworks can be arranged to accomplish effective and intelligent process management based on distinct physical process requirements. In this sense, a cloud-based architecture is a potential solution, as it includes a plethora of software-based computational capabilities, such as virtualization, data and knowledge repositories, and more involved operational and analytical tools [3,4]. Virtualization and softwarization are significant components of cloud architecture in this context, as they contribute to network flexibility of certain applications such as vehicular networks (VNs) and can address issues associated with reactions to any operational events [5]. As a means to achieve this, an intelligent, self-organizing network structure, such as VNs and WSN ground networks seen in forests, can be used to enable rapid response, with the topology reorchestrated by software definition. This can be aligned with advances made in software-defined networking (SDN) when operating various operational network phases, such as self-healing mechanisms. This, in turn, encourages the conceptual development of software-defined wireless sensor networks (SDWSN), in which WSN functions are softwarized and integrated into the network. The three core functionalities are represented by the terms “leaf sensor node”, “router node”, and “IoT gateway node”, which can be modelled and tested on a virtual platform before physical deployment [6]. Herein, cloud-based virtualization is used to develop and provide specified functional configuration parameters to physical nodes. As a result, the cloud can assist with performance analysis via the virtual unit, where possible reorchestration of network behavior can be deployed and tested prior to real-world implementation, paving the way for the cyber–physical system [7,8,9].

The system architecture and design criteria for real-world applications such as health and environmental monitoring were evaluated in relation to the WSN testbed. Several researchers have highlighted key testbed elements that can be customized to design specifications, such as hardware deployment capabilities [10,11,12,13,14]. Some researchers have emphasized the use of different hardware components such as Raspberry Pi and Arduino [15], as well as use cases such as sensor node distribution for environmental applications.

Considering a use case of data collection from scattered traps on the ground in New Zealand, the forest vegetation region is separated into four parts based on [16] (Claverley, North Taupo, Leader Valley, Purakaunui). At these four sites, the density of effective traps ranges from two to four traps per hectare. The spacing between traps varies based on the species targeted, according to the New Zealand Department of Conservation [17], and it determines the distance between traps in a trap line for rats, stoats, and possums. According to [17], the initial spacing between possum traps is 20 to 40 m, but can be expanded to 100 m if possum population is low. The work in [18] used 108 brushtail possum monitoring sites to estimate possum occupancy rates in wildlife, and the transect of traps is recognized at a distance of 200 m apart. Therefore, the distance between possum traps needs to be specified. At the node level, trap sensors capture field-level data before sending them to the central gateway via entry points such as unmanned aerial vehicles (UAVs).

According to Faiçal et al. [19], the embedded hardware collects and processes data from each sensor using the Raspberry Pi single-board computer as a gateway. The work in [20] employed ultralow-power TelosB nodes as ground sensor nodes distributed in the field to improve accuracy when spraying pesticides while lowering the risk of human exposure to these products. To reduce energy consumption, a precise and scalable data gathering scheme was designed to enable long-distance communication at low bitrates. In [21,22], the authors employed an Arduino-based data collection implementation for both WSN and UAV, as well as a Raspberry Pi at the base station to collect environmental data.

Owing to the high cost and risk of damage associated with hardware testing without prior operational and functional network design, virtualization has long been a widely accepted solution for performing software-based simulation testing and obtaining adaptations for use in physical networks. Efficient use of underlying physical functions is mainly achieved by abstracting them into logical or virtual functions [23]. Software-driven virtualization offers a testing ground for conducting and analyzing soft trials of network scenarios, such as dynamic behavior. Such parallel co-simulation running in the cloud backend can significantly aid in leaning out the network configuration process by means of obviating the hardware requirement (during the testing process). For example, the network simulator Contiki Cooja was adopted as a virtualization platform for certain target hardware (Motes such as TI CC2538 Evaluation Module) [24]. Acharyya et al. [24] emphasized the importance of virtualization in driving towards a flexible IoT-based WSN organization. Herein, by accessing real physical data for the purpose of modelling and simulating virtual networks, the organization herein benefited from improved flexibility and reduced latency by deriving appropriate feedback generated by the virtualization unit. Cloud-based virtualization has been adopted to plan and test various WSN reorchestration scenarios when a dynamic event occurs prior to actual implementation. Network performance (i.e., packet loss, network downtime, etc.) can be analyzed so that the most appropriate reorchestration structure can be applied to the physical network [3,25]. This can support flexible network operation and lessen the impact of unexpected network behavior [7].

Concepts such as SDN and network virtualization necessitate the use of tools that can model and test the capability of a network to be tested on a virtual platform before the actual implementation to avoid any major adjustments that need to be conducted in a physical network structure. Furthermore, dealing with the virtualization platform can facilitate dynamic planning for possible network reorchestration as demanded. The Contiki Cooja virtualization (network simulator) tool [26] was utilized in the works of [24,25,27,28] to reflect some of the mentioned ideologies. Furthermore, with Cooja acting as a virtualization platform, virtual nodes are created by compiling and configuring the same Contiki operating system firmware that is used to configure the actual targeted hardware platform of the Texas Instruments CC2538 sensor nodes. From utilizing the tool for a given application point of view, Karegar et al. [29] proposed a point-by-point air-to-ground communication system that considers the clustering structure for partitioning the ground network into small clusters of sensor nodes distributed over large spaces, through which efficient communication between sensor nodes and a UAV is supported. Herein, the UAV path flight was relaxed by using the possible dynamics in WSN orchestrations, as suggested in the approach. The Contiki Cooja simulator was utilized to establish a communication dialogue between the UAV and the ground WSN, the communication planning method being based on the distance between the sensor nodes and the UAV highlighted by the RSSI measurements.

From the compatibility of the suggested system’s point of view, the proposed system model offers interoperable compatibility between the virtual and physical system models. The method’s characteristic in [30] is based on defining a range called compatibility rate that can decide the level of strictness and abstractness of the design, embracing SOA (service-oriented architecture) as the base of concept. This can help a service designer to decide on the level of strictness and abstractness of the design by adjusting the compatibility rate. Therefore, this method reduces the effort and time required for designing an IoT service. Researchers' method is also based on designing dynamic and interoperable architecture once and applying that to multiple use cases due to offering flexibility of network orchestration adaptation based on the use case requirements. The proposed network adaptation capability offers full compatibility for the proposed system model.

Also, authors in [31] used an interpretable architecture to offer communication among the SensorThings API and web processing service, enabling interconnection among environmental sensors, data, and applications. Although this integration may offer real-time access to sensor observations simulation results, it lacks the virtualization capabilities that enables a network testing environment prior to physical network implementation.

In summary, limited information is available pertaining to the generic design of a general-purpose testbed to support various applications ranging from low performance to high performance. Furthermore, the various states of network mobility and connectivity can influence the continual data flow relevant to the different applications. Although significant efforts have been made to develop WSN testbed schemes, the real-time configurability of modules that affect system performance, such as ground communication cost and packet delivery rate, as well as network parameters such as node capacity, has received less attention.