Deterministic Wireless Channel Characterization towards the Integration of Communication Capabilities to Enable Context Aware Industrial Internet of Thing Environments

In order to provide interactive capabilities within the context of Internet of Thing (IoT) applications, wireless communication systems play a key role, owing to in-herent mobility, ubiquity and ease of deployment. However, to comply with Quality of Service (QoS) and Quality of Experience (QoE) metrics, coverage/capacity analysis must be performed, to account for the impact of signal blockage as well as multiple interference sources. This analysis is especially complex in the case of indoor scenarios, such as those derived from Industrial Internet of Things (IIoT). In this work, a fully volumetric approach based on hybrid deterministic 3D Ray Launching is employed providing precise wireless channel characterization and hence, system level analysis of indoor scenarios. Coverage/capacity, interference mapping and time domain characterization estimations will be derived, considering different frequencies of operation below 6 GHz. The proposed methodology will be tested against a real measurement scenario, providing full flexibility and scalability for adoption in a wide range of IIoT capable environments.


Introduction
Industrial environments are evolving to achieve Industry 4.0 paradigms, supported by communication, computing and robotic capabilities provided by the Industrial Internet of Things (IIoT) or Cyber physical Systems (CPS), among others. In this context, communication systems play key role providing collaborative, context-aware environments, with high levels of interactivity, whilst considering the specific requirements in terms of data integrity, delay, security or interoperability of industrial applications [1,2]. As a function of application and system requirements, different communication systems can be employed, combining traditional field buses with transport networks providing service to Supervisory Control and Data Acquisition (SCADA) systems or telemetry and tele control applications, combining wired/wireless networks [3]. Wireless communication systems are gaining popularity, owing to rapid deployment, inherent mobility and scalability [4]. The evolution in communication systems is enabling to optimize network device implementation or tactile internet applications based on the use of edge/fog computing approaches [5,6], or the implementation of delay constrained services with the aid of capabilities in 5G such as ultra-reliable low latency communications [7].
However, their adoption is challenging owing to different factors such as high propagation losses and unstable links, owing to large density of scatterers and elements leading to obstruction, interference or security vulnerabilities [8][9][10][11][12]. The expected increase in the number of transceivers leads to a reduction in cost, form factor and energy consumption, posing new challenges in terms of operation cycle design, energy efficient link control and routing and the use of energy harvesting strategies [13]. An increase in node density, as expected from the massive deployment of transceivers in IoT enabled environments, demands a reduction in overall device cost, which can potentially impact in a negative way in system operation, owing to factors such as 1 3 reduced receiver sensitivity or non-optimal antenna configurations, which increase overall interference levels. This is opening the path of exploring new mechanisms to optimize the operation of IIoT systems, such as federated learning strategies [14] or the use of artificial intelligence to optimize wireless link operation [15]. In this way, new capabilities such as multi robot operation [16] or enhanced tracking and localization techniques [17], among others can be adopted. In this context, the analysis of wireless channel characterization and its impact on overall system operation is necessary to provide adequate Quality of Service (QoS) and Quality of Experience (QoE) metrics [18,19]. Wireless system design can hence benefit with the adoption of different strategies, mainly in layers 1 to 3, with the definition of functionalities such as dynamic spectrum access [20], delay aware routing protocols [21], link scheduling and cooperative transmission schemes [22,23], traffic offload mechanisms [24], to name a few.

Wireless channel characterization in IIoT scenarios
When considering the limitation in the use of wireless communication systems in industrial applications, the impact of interference is one of the main issues, decreasing coverage/capacity relations and delay metrics mainly due to an increase in retransmissions. As the number of coexisting technologies and the density of transceivers increase, interference impact is more significant [25][26][27], so characterizing interference is relevant to envisage potential mitigation mechanisms [28]. In this sense, different approaches have been proposed, such as the use of cognitive radios [29] or enhanced interference identification aided by fuzzy neural networks [30]. Moreover, large signal variability owing to shadowing effects given by large clutter densities can also degrade system performance, with larger percentage of nonline of sight and partial line of sight communication links. Given therefore limitations imposed by interference and signal variability, especially in the case of indoor scenarios, wireless channel modelling and characterization is compulsory to perform radio planning tasks, related with wireless network/node topology and configuration. Different models have been proposed within the framework of 3GPP [31][32][33], following different approaches such as 3D geometric stochastic models in industrial environments [34], modelling in mm-wave frequency ranges [35], the consideration of human body motion within indoor scenarios [36], or polarization diversity [37,38].
In this work, the characterization of complex indoor scenarios amenable to industrial environments will be performed to obtain accurate estimations of coverage/capacity frequency/power values, interference mapping and time domain characteristics. Full details of the indoor scenario will be taken into account and results for the complete volume of the scenario under test will be derived, thanks to the use of deterministic volumetric in-house implemented channel modelling approach. In this way, it is feasible to provide a full volumetric characterization of the wireless channel behavior, in terms of desired transmitted signals as well as mapping of arbitrary interference sources. This enables to provide coverage/capacity estimations within scenarios with high clutter densities, such as industrial environments, which in turn aid in device as well as network planning phases. The proposed methodology enables to perform wireless channel characterization with high accuracy, enabling the consideration of any required transceiver location and variation in the characteristics of the scenario, such as variable densities and distributions of objects, such as stocked goods or raw materials. The rest of the work is organized

Simulation methodology and scenario description
Industrial scenarios in which wireless communication systems operate are generally characterized by exhibiting high density of multiple scatterers, high node density and arbitrary locations of interference sources. In order to provide accurate wireless channel characterization considering the impact of the aforementioned conditions, a deterministic approach based on geometric optics and uniform theory of diffraction (GO-UTD) is adopted in this work. An in-house simulation code has been implemented in Matlab, based in 3D Ray Launching (3D RL) approximation, in which transmitter sources are described by equivalent rays which depart from them in solid angular distribution. A comprehensive description of the code operation and mathematical  formulation is given in several references: a description of the 3D-RL computational technique and acceleration based in the use of feed-forward neural network interpolators is presented in [61]; in [62] and [63] the computational cost derived from the consideration of diffraction is reduced by applying electromagnetic diffusion principles; collaborative filtering techniques in combination with the 3D RL code are considered in [64], to increase simulation resolution parameters in terms on number of launched rays and considered reflections; finally [65] provides a comprehensive description of convergence analysis criteria in terms of angular resolution, launched rays and maximum number of reflections considered. The simulation scenario is implemented in a 3D matrix, which is divided in cuboids. The elements within the scenario are described by their shape and size, as well as by the dispersive electromagnetic parameters of the constitutive materials, given by the dielectric constants and the electric conductivities. The material parameters employed are shown in Table 1. Accuracy (related with average errors between estimation of simulated/measured received power levels as well as with time domain characterization) is mainly determined by the relation between cuboid size, subtended arc between transmitter and receiver points, angular resolution and the maximum amount of reflections allowed until ray extinction [65]. Additionally, edge diffraction as well as diffuse scattering can also be considered in the calculation. In this sense, extensive convergence analysis has been performed in relation with the 3D RL simulation code to obtain optimal values in terms of accuracy vs computational cost for simulation parameters such as angular resolution, cuboid size and maximum number of reflections until ray extinction [65]. The different parameters employed for the 3D RL simulations are given in Table 2. Computational complexity increases with scenario size as well as with the consideration of effects such as diffraction. To reduce computational cost for any given general scenario, the 3D RL code has been coupled to perform hybrid simulation, based on neural network ray interpolators [61], simplification aided by the consideration of electromagnetic diffusion [62,63] or the use database enhanced estimations based on collaborative filtering techniques [64]. Parameters are dynamically adapted as a function of the scenario under analysis, which for the specific case of this work have made use of edge diffraction detection, owing to the type of clutter within the scenario.
In order to perform the analysis of wireless channel performance, the Luis Mercader laboratory, at the Universidad Pública de Navarra has been considered as the scenario under test. The schematic representation of this scenario is given in Fig. 1 The laboratory has a surface of 156 m 2 and a volume of 624 m 3 , in which multiple work benches are present. The scenario has a large density of scatters, owing to constructive elements such as lighting and ducts, as well as obstruction owing to furniture and walls, qualitatively providing a scenario that resembles industrial operating conditions in terms of the number of metallic scatterers, overall number of elements within the scenario and the proportions of line of sight, partial line of sight and non-line of sight channel condition. The lab is divided in two different sections by plaster/cement walls with large window openings. The schematic representation corresponds to the detailed simulation implemented within the 3D RL simulation code, in which the shapes, dimensions and material properties of all the constitutive elements have been mapped, following the information shown in Table 1. Different potential wireless transceiver locations and frequencies of operation will be considered within the volume of the scenario under test, as described in the following section.

Wireless channel characterization in complex indoor scenarios
Once the scenario under test has been defined, estimations of frequency/power distributions, coverage/capacity estimations, interference analysis and time domain parameters have been obtained. Different frequency bands have been analyzed, with the aim of considering LPWAN, WLAN and 5G NR FR1 system deployments. Simulation results for the complete volume of the scenario under analysis are obtained with the aid of 3D RL simulation code. It is worth noting that the complete topo-morphological details have been included in the recreated scenario, in order consider precisely the impact of all the elements within the scenario. received power levels are presented for different cut-plane heights for the sake of clarity. Figure 2 represents the results obtained for all the frequency bands under consideration for a height of h = 1.2 m from the floor reference. As it can be seen, as frequency increases, received power levels decrease owing to inherently higher path loss. Variations within the received power distribution are given by shadowing conditions in the case of NLOS paths, as well as by strong multipath components, which will be described in terms of time domain results later on. These effects can be seen by varying the observation cut-plane height, as depicted in Fig. 3, from 0.6 m to 2.4 m in regular 0.6 m intervals. Validation of the estimations provided by the 3D RL simulation code have been obtained by performing continuous wave radio channel measurements within the scenario. To this extent, a wide band voltage-controlled oscillator (Mini-Circuits ZX95 VCO) has been connected to a wide band transmitter antenna (Antenova Omni LOG up to 8 GHz, depicted in Fig. 4), as transmit source.

Coverage/capacity estimations within the indoor scenario
Received power level measurement have been obtained with the aid of a portable spectrum analyzer (Rohde Schwarz FSH20, up to 20 GHz) directly connected to the same antenna model at the equivalent receiver end, which are depicted in Fig. 5. In all cases, simulation and measurement results are in good agreement, with average errors in the order of 3-5 dB for all cases, showing that the simulation approach provided accurate results for wireless channel estimation tasks.
The scenario has also been characterized in terms of preexistent interference conditions. For this purpose, measured spectrograms have been obtained for each one of the measurement frequencies, with the receiver configuration previously described. A CW carrier for each frequency band of interest was to provide a reference value within the spectrograms, which can be clearly seen for the 4 frequency bands under consideration, depicted in Fig. 6. Average spectral power densities increase particularly in the case of 2.4 GHz and 3.5 GHz, owing to pre-existent wireless systems (WLAN, WBAN and PLMN).
To gain insight in QoS and QoE performance, coverage/capacity estimations can be obtained, as a function of receiver sensitivity for adaptive modulation and coding schemes, thus providing variable sensitivity levels. The corresponding variations in coverage radius as a function of the system employed considering the effect of the surrounding indoor environment can be seen for different linear radial distributions, as shown in Fig. 7. Different systems have been considered: LoRa/LoRaWAN (@433 MHz, @868 MHz), BLE (@1 Mbps, @ 2 Mbps), WLAN (802.11 G @ 6 Mbps, @ 48 Mbps) and 5G NR FR1 3.5 GHz (@ BW 10 MHz, @ BW 50 MHz). Link balance results indicate that an increase in transmission rate, coverage radius decrease as a result in reduction in receiver sensitivity threshold. Moreover, significant variations, in excess of 20 dB for distance variations in the range of 10 cm owing to fast fading can be observed, with relevant link variations in the 1 m to 7 m range.

Interference analysis
Coverage/capacity relations are further constrained by variations in sensitivity owing to the pre-existent noise level within the scenario under analysis. Overall noise power spectral density is given by intra-system interference (i.e., additional transceiver operating within the host system), inter-system interference (operation and co-existence of external communication systems) and external interference sources (e.g., EMI related interference sources, such as motors or machines). The impact of interference has particular relevance in the case of context aware environments, owing to the need of increasing overall transceiver density, which is the case of IIoT applications.
To gain insight in relation with the joint impact of node density variation and the location of elements within the indoor scenario, distributions of SNR within the complete simulation volume have been obtained for the scenario under test. To this extent, different potential transceiver nodes have been located, varying their density from 2 to 8 nodes, for  Fig. 8 (for the case of 2 transceiver nodes), Fig. 9 (for the case of 4 transceiver nodes) and Fig. 10 (for the case of 8 transceiver nodes). As a function of transceiver location, interference conditions change. In the case of the transceivers locates within the scenario, it is worth noting that their height in some cases vary as to consider potentially different locations as a function of node characteristics or application. Most of the nodes are within the 0.9 m-1.1 m range, coincident with the heights of the work benches. In the case of node #3, the height is 3.1 m and #6 the height is 2.1 m, corresponding to infrastructure nodes, exhibiting larger differences (especially in the case of node #3). An increase in the number of transceivers leads to smaller link distances, which can adapt to the existence of the interference node N owing to lower path losses. In any case, the presence of specific scatterers and large objects which lead to shadowing significantly modify SNR conditions (with average differences in excess of 5 dB for adjacent cuboids within the scenario), which can be observed with the aid of volumetric deterministic estimation.

Time domain characterization
In order to gain insight in relation with the influence of time domain characteristics in overall performance (e.g., delay metrics, coherence time or channel equalization features, among others), temporal parameters have also been estimated with the aid of 3D RL method. As an example, the estimation of delay spread, considering a fixed transmitter location (marked as T) has been obtained for the complete simulation volume. Figure 11 shows the results obtained, in the case of an operating frequency of 433 MHz for different cut-plane heights (of 0.6 m, 1.2 m, 1.8 m, 2.4 m, 3 m and 3.6 m, respectively). Variations in the specific cut plane heights lead to clearly visible variation in delay spread values, with variations in excess of 10 ns per cuboid in certain locations, owing to the contributions of the scatterer distribution to multipath propagation components.
The effect of multipath propagation can be observed in more detail by analyzing power delay profile (PDP) results obtained for locations within the indoor scenario. Several examples of PDPs have been obtained for the points indicated in Fig. 12. The case in which the frequency of operation of 433 MHz is employed is depicted in Fig. 13. All the  7 different observation points within the scenario have been considered. As it can be seen, there are relevant differences between the PDP estimations, with delay spreads varying from 35 ns to approximately 100 ns. There are also observable differences in the number of components detected, from 2 to over 30 components. This is directly related to the relative distance between the transmitter location T and the observation points. The further the observation point is (e.g., point #G), fewer components are detected, given by the maximum number of reflected rays (N = 6) allowed until ray extinction (fixed by 3D RL convergence analysis described in [65]).
Results have been obtained for all the operation frequencies and have been depicted (in this case only for points B and D) in Fig. 14. Differences can be observed mainly in detected power levels, which progressively decrease as frequency increases, as a consequence of increased path losses.

Discussion
Different received power level estimations have been obtained for the frequency bands spanning from 433 MHz to 3.5 GHz, obtaining 2D power level distributions as well as linear received power level radial. In this way, coverage/ capacity results have been obtained considering different wireless communication systems operating within these bands as a function of receiver sensitivity values (LoRa/ LoRaWAN, BLE, ZigBee, WLAN and 3G NR FR1), indicating relevant variations in achievable link distances, as a function of node location, node density and interference levels. Multipath propagation has also been characterized within the complete scenario volume, by means of delay dispersion as well as power delay profile estimation for different locations within the frequency bands under analysis.
The results indicate strong variations in dispersion delay,

Conclusions
In this work, accurate wireless channel characterization within the complete volume of indoor scenarios, compatible with industrial environments, with high density of scatterers and objects has been presented. By means of an in-house implemented hybrid 3D Ray Launching simulation code, received power level distributions, coverage/capacity estimations, interference mapping and time domain characterization results have been obtained for 433 MHz, 868 MHz, 2.4 GHz and 3.5 GHz frequency bands within a complex indoor scenario, with high density of scatterers. Measurement results have been obtained for received power levels have been obtained, showing good agreement with simulation results for all frequency bands, with average errors below 3 dB.
The proposed deterministic 3D RL wireless channel approximation can aid in wireless device/system design and planning tasks, optimizing node location and configuration, in terms of QoS/QoE and energy consumption, considering the specific characteristics of indoor industrial environments. By applying the proposed deterministic channel modelling approach, the complete scenario volume is considered, enabling the location of transceivers at any given point of the scenario under test, as well as the consideration of arbitrary distributions of interference sources. In this way, flexible network topologies can be analyzed and envisaged enabling embedding wireless communication transceivers within the most optimal locations, as a function of the time/frequency coverage/capacity estimations. Future work considers dynamic channel characterization considering elements such as human body motion or the presence of unmanned industrial vehicles or complex interference distributions, among others.
Funding Open Access funding provided by Universidad Pública de Navarra. This work was supported by the European Union's Horizon 2020 research and Innovation programme under grant agreement N°774094 (Stardust-Holistic and Integrated Urban Model for Smart Cities) and by Ministerio de Ciencia, Innovación y Universidades, Gobierno de España (Agencia Estatal de Investigación, Fondo Europeo de Desarrollo Regional -FEDER-, European Union) under the research grant RTI2018-095499-B-C31 IoTrain.

Conflicts of interest/Competing interests
The authors declare no conflict of interest or competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.