Industrial networks for process automation are deployed in sites which can be hundreds of meters wide, hosting very dense networks consisted of hundreds or thousands of nodes. Harsh industrial environments impose a number of challenges for wireless communications: reliability, fault-tolerance and low latency being the biggest ones. Unpredictable variations in temperature, humidity, vibrations and pressure make the industrial environments harsh, as well as the presence of highly reflective (metal) objects and electromagnetic noise. Even though not much data needs to be communicated in an industrial application, reliability and latency are critical, that is, delivery of all data must be guaranteed in real-time. Wired networks have met these requirements and are being used in spite of the high cost of wiring and the often present installation difficulties (see Figure 1
) because wireless solutions are not as robust as their wired counterparts. Industrial automation systems in chemical industry, power plants, oil refineries or underground water supply systems implement complex monitoring and control processes. Thousands of devices send measured values (i.e., temperature, pressure, flow, position) to the actuators that control processes and to the servers that coordinate the production phases. Wiring is generally both challenging and costly (cca. 20 $
/m): flammable, explosive and hot environments have to be avoided (e.g., in the presence of flammable gases in an oil refinery), remote or unavailable locations are hard to reach and mobile nodes can hardly be connected at all. Although wired networks at this time cannot fully be replaced by wireless networks in this domain, supervision and non-critical control with loose enough requirements could be realized over wireless. In addition, significant constrains that limit the practical deployments of wireless networks in such scenarios are battery capacity and power consumption of the devices. Ideally, communication and power cables can be mitigated to enable a fully wireless solution. For that, the devices should be energy efficient and able to power from a battery for years. Moreover, wireless networks introduce logical benefits that could be used in maintenance and commissioning, such as “plug-n-play” automation architectures to reduce downtime and speed-up tests and “hot-swapping” faulty modules. In addition to control and supervision, global wireless plant coverage could enable localization and tracking of parts in production, coordination of autonomous transport vehicles and mobile robots [1
Industrial Wireless Sensor and Actuator Network (IWSAN) are gaining popularity in process industries due to their advantage in lowering infrastructure cost and deployment effort. The advent of Industry 4.0 already resulted in the successful use of IWSANs for monitoring applications and non-critical open-loop control in factory automation. A few new wireless technologies, such as WirelessHP [2
], OFDMA wirelesscontrol [3
], Real-Time-WiFi [4
], Wireless network for Industrial Automation and Process Automation (WIA-PA) [5
], can replace extensive wiring on industrial machinery, providing connectivity between machine parts with
order of magnitude latency. Even though they enable reliable and fast communication, the range of such networks is limited to only a few meters, making them unsuitable for broad usage across an entire industrial site in process automation or for reaching remote areas if infrastructure cost has to be kept low. Ranges up to a few hundred meters are feasible with 802.15.4-based technologies such as WirelessHART [6
], ISA 100.11a [7
], 802.15.4g [8
] with Time Slotted Channel Hopping (TSCH) [9
] and WIA-PA [10
], at the cost of other performance metrics. Sub-GHz wireless technologies, such as LoRa and SigFox, further extend the coverage due to the better signal propagation characteristics (up to 15 km and 50 km respectively) but are not suitable for frequent critical traffic given their low data rates (up to 50 kbps and 0.1 kbps respectively) which lead to very long transmission times in both uplink and downlink. In downlink, long transmission times also limit the gateway to serve many nodes, more so considering the duty cycle limitations [11
]. Moreover, LoraWAN Class A and Sigfox only allow downlink transmissions that immediately follow uplink, resulting in substantial downlink delays due to buffering. NB-IoT experiences downlink delays due to buffering as well, when using the Power Saving Mode (PSM). The existing trade-off between the range and latency varies across different technologies (cf. Figure 2
), aiming to cover a variety of use cases. This paper explores the aforementioned trade-offs and the conditions that enable the use of particular Internet of Things (IoT) wireless technologies in heterogeneous sensor-actuator networks for mid-range communication able to cover an industrial site ranging up to more than one kilometer in diameter.
Wireless Sensor Networks (WSNs) have been evaluated from different perspectives in the state of the art literature. However, significantly smaller amount of research is conducted in the context of IWSANs that have much more strict application requirements. An overview of key issues and challenges of wireless technologies in industrial networks is surveyed in References [1
]. Communication requirements and a general profile of a wireless fieldbus for low level short-range factory automation systems are discussed in Reference [20
]. References [13
] discuss security and Quality of Service (QoS) perspectives of IWSAN in industrial automation. Furthermore, an in-depth review of recent advances in real-time IWSANs for industrial control systems is given in Reference [22
], with a focus on WirelessHART. Reference [22
] reviews real-time scheduling and analytic techniques for achieving real-time performance in Reference IWSANs. An extensive survey on wireless network design for control systems is presented in Reference [17
], briefly reviewing a few of the existing wireless technologies in that context but mostly focusing on the joint design considerations of both control systems and wireless networks. A comparative examination of ZigBee, WirelessHART, ISA100.11a and WIA-PA in terms of network architecture and protocol design in the context of iwsan is given in Reference [19
]. State of the art in Low-PowerWide-Area Network (LPWAN) solutions for Industrial Internet of Things (IIoT) services is explored in Reference [23
This paper takes a different approach and interprets the wireless standards from the practical standpoint, offering the readers concrete values on achievable sampling rates, energy consumption, scalability and coverage in practice. Both standard specifications and product datasheets provide extensive low level data, and are insufficient on their own without additional empirical research. This paper quantifies the existing trade-offs in wireless technologies for wireless sensor and actuator networks with coverage of at least a couple of hundred meters, able to cover a production site or at least a large part of it. Range is crucial in process automation for all slave nodes to be able to reach a master node, considering that control is typically done by one or few master nodes (controllers) and a large number of slave nodes (sensors and actuators) that take part in bidirectional communication with the controller and are spatially distributed over the entire site. This paper presents cost, scalability, latency, reliability, range and energy consumption evaluation of wireless technologies with promising range and latency potential, including LoRa, IEEE 802.11ah (Wi-Fi HaLow), Narrowband-IoT (NB-IoT), WirelessHART, ISA100.11a, Bluetooth Low Energy (BLE) and 802.15.4g physical layer with 802.15.4e TSCH on data-link layer. These technologies offer the possibility of a dense heterogeneous wireless network deployment able to serve both actuators and sensors in critical applications, as well as provide an infrastructure for supervisory traffic. Along with the practical limitations of each technology with respect to the existing trade-offs between latency, throughput, coverage and scalability, a direct projection of the aforementioned wireless technologies to their key performance indicators is made, aiming to enable adequate network design in particular industrial applications.
The remainder of this article is organized as follows. The requirements and challenges that industrial networks must comply with are summarized in Section 2
. Section 3
presents a general discussion of the key trade-offs in wireless network design in the context of the requirements, while in Section 4
the discussed trade-offs are quantified for each particular technology. Overall discussion and the experimental evaluation of energy consumption is presented in Section 5
. Finally, Section 6
presents the conclusions.
2. Requirements and Challenges
The International Society of Automation (ISA) classified industrial systems into six classes [13
] on the basis of data urgency and operational requirements. These classes range from critical control systems to monitoring systems, from the strictest requirements to the most relaxed ones respectively:
Safety systems—require immediate actions on events (usually in the order of tens or hundreds of or a few ms).
Closed loop regulatory systems - control the system via feedback loops operating either periodically or based on events. They may or may not have stricter timing requirements than safety systems.
Closed loop supervisory systems—similar to regulatory systems with the difference that the feedbacks are usually non-critical and event-based, for example, collecting statistical data and reacting only when a certain trend is observed by issuing a notification or alarm.
Open loop control systems—where sensors collect data and store it to the central database. An operator (human) analyzes the data and acts upon it if needed.
Alerting systems—send periodical or event-based alerts indicating different stages, for example, heating up the boiler and alerting every once in a while to indicate the progress.
Information gathering systems—collect the data (logging) and forward the logs to a server. These systems have no immediate operational consequence.
Wireless coverage of the entire industrial site may benefit classes 2–6, whereas class 1 requires a solution combining both ultra-high reliability, redundancy and ultra-low latency, which is infeasible with long range wireless considering the trade-offs. Performance requirements of different classes are depicted in Figure 3
. For site-wide coverage, a range of at least a few hundred meters is needed. Site-wide coverage would enable multicasting measurements to several destinations, for example, actuators, supervision systems, databases, enabling the support of different services for several classes of industrial systems, making the network heterogeneous. Thus, site-wide IWSANs need to be scalable enough to accommodate new nodes and provide QoS, considering that they are expected to run for several decades. Different applications require different performance and services, as examples in Table 1
Besides the key performance requirements illustrated in Figure 3
, deployment cost, energy consumption, interoperability, QoS and service differentiation come to focus especially when considering heterogeneous networks. A wired fieldbus network is very expensive to deploy because of tens of kilometers of cables needed to connect devices to their master nodes, the time needed for deployment and the maintenance of such deployment. Lower deployment and implementation costs are the prime motivations for the transition from wired to wireless solutions wherever possible. Among wireless solutions, subscription fees for operator based networks also vary. Operator based solutions are generally not ideal for industrial purposes as the dependence on the operator in case of failure increases the repair time. However, reliable full-duplex operation of wired industrial networks is a large advantage over wireless technologies that are the subject of this article. Namely, reliability inherently suffers in the full-duplex wireless solutions because of the self-interference and increased interference from the neighbours [27
]. Opting for half-duplex instead causes the inability to send and receive at the same time on the same channel, which in turn largely increases the latency in wireless networks. IWSANs must operate in real time to serve class two systems. Specifically, closed loop regulatory systems require IWSANs to sample, process and exchange the data between a sensor and an actuator in a time frame that is less than the cycle time of the loop, with typical values ranging from microseconds to hundreds of milliseconds (depending on the concrete process being controlled). Critical applications (classes 1 and 2) also require redundancy, resistance to noise and robustness against failure as they must ensure timely and successful delivery. In addition, the failure of one or a few nodes must not compromise the operation of the network as whole.
Many of the stated requirements are interconnected and there is no single technology that covers all of them simultaneously. The inevitable trade-offs, their causes and consequences are elaborated in the next section.
3. Trade-offs in Wireless Network Design
Providing wireless communication to heterogeneous applications, including the time-critical ones, over a wide industrial site is a conflicting task. For example, l All three are partly determined by the choice of frequency band and bandwidth but also with other design choices that create additional interlocks between the performance parameters. These trade-offs, illustrated in Figure 4
, complicate the design of wireless solutions.
3.1. The Transmission Range
The transmission range is mainly determined by the transmission power, typically limited by regulations [11
], the radio and propagation properties, as well as coding and modulation complexity. If a radio transmits at a constant power, lowering this complexity rate permits the correct decoding of a weaker or more distorted signal by a receiver, thus extending the transmission range. Also, higher frequency bands with more bandwidth available enable higher data rates and faster data transmission but they also have worse penetration capabilities which reduces the range in an industrial site full of obstacles. Range is largely determined by topology as well. Multi-hop topologies extend the range at the expense of latency, design complexity and energy consumption because of the need for synchronization of nodes, routing and so forth. In conclusion, low data rates at low frequencies and multi-hop topologies are prolonging the range but they all increase latency.
Latency is reduced by increasing the data rate, in turn enabled by more complex codings and larger bandwidths. Furthermore, multi-hop topologies increase the latency considering that forwarding and routing introduce additional delays. In addition, computing a new route when a link fails also introduces delay which can render multi-hop topologies useless in low-latency time-critical applications. Medium Access Control (MAC) design has a significant impact on latency as well, especially in IoT technologies where devices aim to sleep as long as possible to save energy, therefore delaying transmissions and receptions. MAC protocols can be classified into four classes: (1) Fixed Assignment Protocols where resources are divided among the nodes for a defined time duration, (2) Demand Assignment Protocols where resources are provided to a node on demand, (3) Random Access Protocols where resources are divided randomly and (4) Hybrid Protocols that combine fixed or demand assignment with random access. Fixed Assignment Protocols such as Time Division Multiple Access (TDMA) introduce determinism and achieve lower latency than random access protocols under very high load, but under low load they waste resources by inefficient usage of the channel time, where random access protocols achieve lower latency. Demand-based protocols are not suitable for low-latency time-critical communications given that explicitly asking for resources every time takes up bandwidth and adds up to latency. For heterogeneous industry applications, hybrid approaches are the most promising given that they aim to combine the benefits of both fixed assignment and random access protocols, while surpassing their limits at the same time and adapting to the network conditions. In addition, retransmissions need to be kept a minimum as they also increase the latency.
Reliability is determined by topology, MAC design and Modulation and Coding Scheme (MCS). One of the major setbacks of wireless technologies in terms of reliability, in comparison to their wired counterparts, is the inter- and intra-technology interference on air which can cause collisions and increase packet loss. Technologies that operate in licensed bands reserve a part of the spectrum for themselves, mitigating the issue. However, spectrum is a scarce resource and it comes at a high price. Private deployments are not possible in reserved spectrum, disabling the possibility of local control over a network. Shared spectrum, on the other hand, can be shared by any number of technologies which can try and mitigate interference by channel hopping or using some MAC layer mechanisms such as Listen Before Talk (LBT). Furthermore, in a single-hop networks, the success probability is entirely dependent on a single link, opposed to multiple links in multi-hop networks. Reliability can be improved by employing both retransmissions and repetitions at the MAC layer, which also add to the latency. To reduce the number of retransmissions, error control techniques such as Forward Error Correction (FEC) can be used. Coding schemes and modulation largely define reliability. Coding rates create extra error checking bits that make modulation more reliable. Modulation schemes are more reliable as they have fewer points on the constellation diagram but also slower. That makes Binary Phase Shift Keying (BPSK) the slowest and the most reliable modulation compared to Quadrature Amplitude Modulation (QAM) and Quadrature Phase Shift Keying (QPSK), given that it only accommodates two points (one bit per burst). QPSK uses four constellations, whereas QAM can have any number of points. Any increase in the number of points on the constellation diagram reduces the space between them, leaving fewer margins for error. This makes QAM the fastest modulation scheme but more unreliable over longer distances.
3.4. Data Rate
Data rate is directly correlated with the available bandwidth and thus frequency band, on one hand, and with modulation and coding scheme on the other. More bandwidth enables higher data rates, while modulation techniques and coding schemes can further contribute to the achievable data rate by encoding more data into the signal. Unlicensed wireless technologies operate either in sub-GHz frequency bands (400 MHz, 800–900 MHz), in 2.4 GHz or in 5 GHz. Sub-GHz technologies generally (although not universally) use narrower channels (few hundred kHZ) than those in GHz frequency bands (22 MHz Wi-Fi, 2 MHz 802.15.4) and thus have more limited data rates than the GHz ones.
3.5. Energy Consumption
Energy consumption depends on data rate, topology and MAC design, as well as the hardware design of course. Low data rates result in long transmission times, which increases the energy consumption of the node and reduces the battery lifetime. Topology wise, nodes in multi-hop networks consume more energy than in single-hop networks given that, besides their own transmissions and receptions, they also need to forward other nodes’ packets. Energy-efficiency of data forwarding paths give the routing protocols a strong influence over energy consumption as well. Complex coding and decoding operations also contribute to energy consumption. For example, FEC has been omitted in commercial 802.15.4 based networks due to the energy consumption of the decoding operation. Nevertheless, employing FEC could reduce the overall energy consumption as less energy would be spent on retransmissions and rescheduling [28
]. Besides, MAC design has a significant impact on energy consumption as it defines scheduling and hence the radio on and off times.
Scalability is primarily determined by MAC design. Scheduling, contention resolution and other MAC mechanisms work together to provide maximum network capacity. In single-hop networks, the network capacity upon reaching the upper limit can only be extended by deploying more base stations. However, in practice the density of such base stations is limited. Multi-hop networks address this issue by allowing for wireless data forwarding, at the expense of overall throughput. In TDMA-based protocols, network density is limited by the need for synchronization and time division in combination with QoS requirements.
3.7. Spectrum Regulations
Another tackling design choice is the one between unlicensed Industrial, Scientific and Medical (ISM) and licensed bands. On the one hand, worldwide permitted unlicensed operation reduces the runtime costs but has no regulatory protection against interference by other wireless networks operating in the same frequency band. On the other hand, even though licensed bands prevent interference, they typically depend on an external operator. Therefore, network issues cannot be immediately resolved on site, the external operator needs to resolve them. This introduces administrative delays which are unaffordable in time-critical industrial applications. Communication technologies operating in the unlicensed spectrum are maintained and managed locally. However, several co-located or overlapping wireless networks operating in the same frequency band will interfere with each other and can experience decreased QoS and extensive packet loss [14
]. In an effort to alleviate coexistence issues in unlicensed spectrum, regulatory bodies have issued a number of norms such as a Clear Channel Assessment (CCA) check before each transmission by all devices, that is, a device has to sense if the channel is free by energy detection or other types of Detect And Avoid (DAA) mechanisms [14
]. Although the DAA mechanisms improve the coexistence between the contending wireless nodes and networks, collisions can still occur. Apart from collisions, medium sensing adds to the latency and introduces non-determinism due to the medium congestion. Aforementioned facts significantly limit the use of wireless solutions in closed loop control applications in automation industry. A limiting regulation is present in unlicensed sub-GHz spectrum as well. Devices with an operating range of 863–868 MHz in Europe, 916.5–927.5 MHz in Japan and 902–928 MHz in the US must comply with the maximum duty cycle limit of 2.8% and 10% for the, Access Point (AP) provided that they support LBT and Adaptive Frequency Agility (AFA) features, 1% otherwise [11
Key specifications of each technology introduced in Section 4
are listed in Table 2
and Table 3
. Note that not all numbers in Table 2
and Table 3
can be taken for granted as they do not depict the trade-offs, but only present theoretical limits. NB-IoT operates in licensed spectrum. It can reach as far as its cellular infrastructure goes, and it is inherently more reliable than technologies that operate in shared spectrum. LoRa, IEEE 802.11ah and 802.15.4g operate in unlicensed sub-GHz frequency bands, which makes their long range a feature of their physical layers. LoRa operates in 863–870 MHz frequency band in Europe, offering three sub-bands at 864 MHz, 867 MHz and 868 MHz. Five 125 kHz channels are defined in 867 MHz band and three in 864 MHz and 868 MHz band [54
]. The three default channels in 868 MHz band are to be implemented by every node, whereas the rest are optional. Wi-Fi HaLow operates in 863–868 MHz band in Europe and defines five 1-MHz channels and two 2-MHz channels [56
]. Considering this, it is clear that Wi-Fi HaLow and LoRa’s optional channels overlap in 3 out of 7 Wi-Fi HaLow channels and do not interfere with each other otherwise, which makes their parallel deployments possible. However, Wi-Fi HaLow can severely interfere with 802.15.4g [57
], given that they operate in the same bands. The 2.4 GHz unlicensed frequency band is widely utilized today and the technologies that operate in this band must pay special attention to coexistence. WirelessHART, ISA100.11a and 802.15.4e TSCH are all based on the 802.15.4 2.4 GHz PHY layer that operates in sixteen 2 MHz-wide and 5 MHz-spaced channels in 2.4 GHz frequency band. They all employ frequency-hopping to improve the reliability of their transmissions, as they make use of exactly the same spectrum. Given their shorter single-hop ranges, these technologies employ multi-hop topologies to extend their range.
Wireless networks have the benefit of mitigating cabling for communication but the question of power cables still remains. Ideally, power cables could also be mitigated when the devices can live long enough drawing the power from the batteries available today. This is feasible for sufficiently large cycle times. Therefore, it is important to also consider the energy consumption. To evaluate the possibility of entirely wireless deployments that do not need power cables, an experimental energy efficiency comparison between LoRa, NB-IoT, IEEE 802.11ah and IEEE 802.15.4g (Wi-SUN with TSCH) was performed. The experiments combine energy consumption values of state-of-the-art off-the-shelf radios obtained from their data sheet (cf. Table 4
), with simulated performance analysis. At the time of this study, no off-the-shelf radio was available for Wi-Fi HaLow and the same radio was assumed as for 802.15.4g (Atmel AT86RF215), as it supports the required modulation and coding schemes. The simulation experiments were performed using the ns-3 network simulator for LoRa [58
], NB-IoT [59
] and Wi-Fi HaLow [60
]. Using the 6TiSCH simulator, 802.15.4g was evaluated [61
]. A single device and a single gateway/AP is considered in the simulations. Hence, the results do not take into account scalability or contention. To maintain simplicity, predictability and comparability of the results, the experiments did not take into account any packet loss due to propagation errors, interference or collisions. As such, the PHY and channel contribution to the energy consumption is limited to the time-on-air of the radio for the reported packet sizes (including preamble and PHY/MAC header). The main contribution of the results pertains to the energy consumption of the MAC layer protocols of the considered technologies. For this reason, we used a simplified linear battery discharging model.
Wi-Fi HaLow is significantly more energy efficient than 802.15.4g with TSCH. This improvement is due to the higher supported data rate that leads to shorter transmission times. This result is also reflected in the predicted battery lifetime shown in Figure 5
. Given a sufficiently large battery of 2000 mAh, the lifetime of Wi-Fi HaLow is expected to be above 10 years for a 10-min transmission interval, without battery replacement. For 802.15.4g, battery life expectancy is under 3 years in the same scenario. For the long-range contenders, it is under a year, as shown in Figure 6
b, except the LoRaWAN sf7 that can live up to 7.5 years (cf. Figure 6
a). The compared configurations of 802.15.4g and 802.11ah do not represent the best case scenario regarding energy consumption, as both technologies are configured to use low data rates. However, the chosen settings result in similar coverage range. In a line-of-sight scenario, 802.15.4g FSK-50 reaches a good PDR up to 700 m, while FSK-200 goes up to 420 m [62
]. For 802.11ah [63
], mcs10 (150 kbps) has a range of 700 m to more than 1000 m. For a higher data rate, 802.15.4g will achieve better energy efficiency but it’s achievable range would be very different from that of mcs10 of 802.11ah. Similarly, 802.11ah will achieve better energy efficiency for higher data rates (mcs10 represents the lowest possible data rate and thus the worst case).
illustrates the benefits of low power modes in the design of a technology. As shown in the Figure, the lifetime of LoRaWAN devices is much higher than that of NB-IoT. Power consumption of LoRaWAN devices in transmission (TX)/reception (RX) mode is the dominant factor of the overall consumption, considering the very low power consumption in the idle/sleep mode (cf. Table 4
). This results in a large difference in lifetime between LoRaWAN devices that use SF7 and SF12. The difference between SF7 and SF12 is especially large when the transmissions are frequent. The rarer the transmissions, the smaller the difference between SF7 and SF12 as idle/sleep time prevails and the TX/RX time becomes less significant in comparison to the idle/sleep time.
The larger difference in battery lifetime between NB-IoT mcs9 and mcs4 occurs due to the better efficiency of mcs9 which reflects not only in data TX/RX but also in signaling between the transmissions which adds up to the difference. The difference between mcs9 and mcs4 also reduces as the traffic interval increases, similarly as the difference between spreading factors of LoRaWAN. NB-IoT has higher data rate than LoRaWAN, thus it is more efficient in terms of TX/RX. However, the NB-IoT hardware is less efficient in terms of energy consumption in all four modes, so LoRaWAN becomes more energy efficient relative to NB-IoT. Higher data rates of NB-IoT cannot compensate for the energy efficiency of LoRaWAN in the evaluated scenarios.
It is important to note that the shelf life also affects the battery life. The lifetimes illustrated in Figure 5
and Figure 6
represent the ideal cases which only take into account the consumption of a radio and a microcontroller. Hence, the presented results do not take into account energy consumption of peripherals, self-discharge of battery nor other power drains. For example, when considering only radio and microcontroller, NB-IoT platform using mcs4 and mcs9 would need around 230 mAh and 155 mAh respectively to run for three years. For the same time period, LoRaWAN radio and microcontroller using SF7 and SF12 would consume 8 mAh and 47 mAh, when ignoring other power drains and employing sleep mode (45 mAh and 83 mAh with idle mode). However, the above mentioned factors need to be taken into account when estimating very long battery lives as they make the largest part of energy consumption over a long time.
This paper focused mostly on PHY and MAC layer features of the considered wireless technologies, even though some technologies also address commissioning, security, roaming and other higher layer features. In addition to PHY and MAC, LoRaWAN defines a complete network architecture, various device types, commissioning and security (network and application). Parameter configurations and network management are generally implementer specific for all the considered technologies. LoRaWAN is configurable in terms of many parameters such as:
spreading factor (and thus data rate): fixed choice or adaptive,
reliability: ACKs or multiple transmissions of the same packet without downlink ACKs,
higher layer logic: raw payload or Internet Protocol (IP) compliant stack.
Choosing the network setup for a certain scenario is not always straightforward, which motivated using optimization techniques to choose the optimal parameters [64
]. Moreover, LoRaWAN does not impose the coordination of transmissions of class A devices, which might impact scalability in real deployments. Similarly, Wi-Fi HaLow defines PHY and MAC layer, as well as layer 2 security upon connecting. On top, IP compliant stack is to be used. Wi-Fi HaLow is highly configurable, defining a multitude of different parameters to be configured such as MCS, RAW, Traffic Indication Map (TIM) and others [37
]. The configurable parameters can significantly influence the performance and the lifetime of the network. Higher layers of NB-IoT can be both IP- and non-IP compliant. NB-IoT also defines some configurable parameters such as extended Discontinuous Reception (eDRX) and PSM timers. However, other settings are under control of the network operator. The standard IEEE 802.15.4g/e TSCH defines PHY and MAC, with TSCH MAC layer specifying how to execute the schedule but not how to define it. Given that a schedule significantly impacts the network performance, there are standardization efforts in Internet Engineering Task Force (IETF) on scheduling functions but there is also a choice between centralized and decentralized scheduling. WirelessHART and ISA100.11a share a similar concept as TSCH. They have self-configuration capabilities greatly simplifying the deployment. They define the full stack and are centrally managed by the network manager that makes use of commands defined by the standard for network management [67
]. They are out there for a long while and are quite well understood. BLE also defines the entire stack as well as other aspects such as commissioning. BLE has configurable parameters such as advertisement interval and connection interval that influence its performance. Different combinations of the configurable parameters can result in a different performance of any technology in various scenarios, thus the interdependence of the configurable parameters needs to be empirically examined.