Smart manufacturing aims to increase productivity and efficiency by integrating the physical world into the cyber world through the Industrial Internet of Things (IIoT). The IIoT now connects millions of industrial devices embedded in the physical world to the Internet (or an organisation’s Intranet) and allows to integrate the data generated by them into IT and business processes and services. This new paradigm is called Industry 4.0. The integration of physical and cyber worlds as part of Industry 4.0 is turning traditional industrial automation and control systems into cyber-physical manufacturing systems.
Currently, communication technologies used in industrial automation and control systems are by and large still wired, which has many limitations such as higher installation and maintenance costs, restricted coverage, difficulty in installing in a harsh environment [ren2021intelligent]. For this reason, wireless automation systems have attracted much attention as they overcome many of the drawbacks of wired technology. In addition, device mobility is an essential requirement for future industrial environments as it supports easy reconfiguration of the automation system in response to flexible manufacturing needs.
A big driver in the transformation from industrial automation and control into cyber-physical manufacturing systems is the introduction of private 5G wireless networks into industrial environments. Looking further ahead, it is anticipated that the shift from 5G to 6G will also stimulate a transition from Industry 4.0 towards Industry 5.0. 6G will develop better integration of automatic and high-precision manufacturing processes as well as an integration between machines and humans in the loop through low latency and high reliability [Letaief_2019]. Furthermore, 6G and Industry 5.0 may coexist with systems, including cyber-cyber, cyber-physical and physical-physical systems.
Compared to traditional wireless communication, industrial wireless communications are challenged due to metallic structures, electromagnetic interference (e.g., from electrical motor droves or welding apparatus), arbitrary movement of objects (robots and vehicles), room dimension, and thick structural pillars, etc. These challenges can significantly affect the reliability and performance of wireless communications in industrial environments. Besides, full industrial automation requires ultra-reliability and low latency communications in order to deliver sensor data and actuation commands at precise instants with designated reliability (i.e., to perform mission-critical industrial processes). In a smart manufacturing environment, it can be difficult to meet the communication reliability and latency requirements at all times. For example, the factory floor can be reconfigured from time to time according to the flexible production demands, while the network infrastructure (i.e., the wireless access points) is static in general. This creates a situation where a given area can be under network coverage in one plant configuration but might not be case in another configuration (e.g., due to the rearrangement of machineries).
In order to meet these requirements, the recently introduced concept of an intelligent reflecting surface (IRS) has drawn increased attention from both academia and industry [di2019smart]. An IRS is a key enabling technology that can significantly improve a wireless network’s performance, creating a programmable radio propagation environment. An IRS is a programmable meta surface containing a large amount of small, low-cost passive antenna arrays that can control a propagating waves’ phase, amplitude, frequency, and even polarization. An IRS can increase the efficiency of the wireless network in terms of data rate, coverage, and connectivity. For instance, if the line-of-sight (LOS) is blocked in the wireless network, an IRS can create a reflective link (i.e., a virtual LOS) to bypass the obstacles between communicating devices.
This paper aims to conceptualize the potential for IRS implementation in a smart manufacturing environment to support the emergence of Industry 5.0. In Section II, the system model is introduced. Several IRS applications are presented in Section III, specifically those relevant to smart manufacturing. We then outline the significant future research directions discussing IRS’s challenges and opportunities in modern smart manufacturing in Section IV, and provide conclusions in Section V.
Ii Intelligent Reflecting Surfaces
The concept of IRS is drawn from the concept of a meta surface, which is a 2D form of a meta material. Generally, depending on its structural parameters, this engineered man-made material exhibits unique electro-magnetic properties that cannot be obtained with conventional materials. The IRS is constructed with a large re-configurable array of passive sub-wavelength scale scattering elements (dielectric or metallic patches) that are printed on a grounded dielectric substrate. The size of these patches and their inter-element spacing is usually half of the wavelength or smaller (5 to 10 times smaller) [renzo2020smart].
A meta material unit or patch element is capable of adjusting the phase and amplitude of the reflected signals. The direction of reflected signals from each of these elements can be adjusted in the desired fashion (so as to interfere constructively or destructively at the intended location) by controlling their reflection coefficients in real-time. This phenomenon can be characterized by the concept of reflectivity which is defined as the ratio of the reflected signals to the incident signals. The reflectivity of the meta material unit can be obtained by its state, the incident angles and reflected angles. The reconfigurability of the meta material units or patch elements are achieved with the help of tunable low power electronic circuit elements such as positive-intrinsic-negative (PIN) diodes or varactor diodes or radio-frequency (RF) switches as shown in Fig. 1 [renzo2020smart]. By controlling the bias voltages of the PIN diodes, each PIN diode can be tuned between ON and OFF states which facilitates in determining the state of the meta material unit [wu2021intelligent, elmossallamy2020reconfigurable]. Similarly, by properly adjusting the bias voltage, the capacitance offered by the varactor diodes can be controlled so as to get the desired reflection coefficient properties. Another approach is to implement the phase shifter using RF switches along with lossless delay lines. An -bit RF switch is expected to provide discrete phase shifts, incurring only dB loss.
In order to program or reconfigure the smart surface or patch elements remotely, an IRS is equipped with a controller as shown in Fig. 1. The controller is connected to a base station or access point to receive relevant control and re-configuring commands. Although it is not explicitly shown in Fig. 1
, an IRS can also be equipped with sensors which helps to estimate the wireless channel conditions[renzo2020smart]. Due to the nearly passive nature of an IRS, most of the experimental works reported in the literature do not employ sensors to reduce the overall energy consumption of an IRS.
Since the number of meta material units is generally large, it is expensive and inefficient to regulate each unit individually. Therefore, adjacent meta material units are grouped with a minimum number of equal units that can be controlled independently. Each group is referred to as an element having the same state and the states of all the elements is called the configuration of the meta surface.
Iii Applications of IRS in smart manufacturing.
In the following we present a number of use cases for an IRS or multiple IRSs in a future manufacturing environment with many autonomous and mobile devices. We show where their functionality can be of particular use.
Iii-a Blockage Mitigation
Obstructions and blockage are major issue for signal coverage and can cause intermittent and poor connectivity. As a way to circumvent the obstructions, an IRS can help steer the incident signals around an obstruction and cover the area shadowed from the base station. Equipped with a large number of low-cost passive reflective elements, the IRS enables the adaptation of wireless environment to overcome the blockage and provide a strong reflective non-line-of-sight (NLOS) link.
In the following, we show the pathloss characteristics when using an IRS with a transmission frequency of 30 GHz. We assume that the communication link between the base station (BS) and the receiver is completely blocked and an IRS is set up to enable the communication between BS and the receiver, as depicted in Fig. 2. In our scenario, we assume that the IRS is composed of elements which are positioned 20m away from the BS in such a way that the IRS has a LOS link with each, the BS and the receiver. Fig. 3 shows the end-to-end pathloss as a function of the distance between the IRS and the receiver. The end-to-end pathloss is calculated using an IRS-based pathloss channel model presented in . With varying number of IRS elements, the figure shows that for any given IRS-receiver distance, over 10dB received power gain can be achieved by increasing from to .
This is a positive result to enable flexible smart manufacturing environments, where a plant re-configuration can lead to a given area being covered by wireless signals to one which is shadowed by machinery in another configuration, as illustrated in Fig. 4. In addition, workers, moving machinery, mobile robots or other vehicles are also potential blockers for a propagating signal. Hence, the signal coverage can be automatically adapted as the factory plant is reconfigured without the need of redeploying the network infrastructure. This flexibility provided by IRSs is key to enable the following features and applications that rely on high coverage availability, thus improving the communication and operational performance of a smart factory.
Iii-B Millimeter-wave and Terahertz Communications
Wireless communication for smart manufacturing is characterized by strict link and system requirements regarding number of nodes, availability, reliability and latency. For instance, closed-loop motion control automation use cases may demand cycle times lower than 1 ms and 99.9999% service availability for more than 100 nodes [etsi_ts_122104_v1650]. To support such application requirements, millimeter-wave or terahertz spectrum will provide wide bandwidths wireless channels to accommodate large number of nodes operating with high data rate and low latency. However, propagating mmWave/THz signals suffer from very high path attenuation and lower penetration through materials compared to lower frequency signals in the sub-6GHz bands, which have so far been used. It means that communication links are more vulnerable to blockage, potentially affecting their performance. Although the use of highly directional antennas can compensate some of the path loss, narrow beam widths may make the link even more vulnerable to blockage.
The mmWave/THz signals have small wavelengths that are comparable to the surface roughness of many objects, which suggests that scattering may not be neglected like it was compared to lower frequencies. Also the scattered power relative to the reflected power at mmWave/THz frequencies increases as the incident angle, and lower reflection loss (e.g., stronger reflections) are observed as frequencies increase for a given incident angle. This means that the signal energy can be more diffused in the environment and the propagation characteristics is highly dependable of surfaces and incident angles of the impinging waves. Therefore, the application of IRS for mmWave/THz could be a means to generate a more predictable and controllable channel to overcome the effects of scattering. This requires jointly optimizing the transmit beamforming and the IRS phase shift parameters, maximizing the received power [wang2020intelligent].
Iii-C Wireless Energy Transfer
As mentioned earlier, a smart manufacturing environment consists of many sensors and actuators. These actuators and sensors are usually powered by batteries, which deplete over time. Replacing batteries when they run out is a daunting task with thousands of sensors and actuators. Additionally, sensors will likely be installed in many sensitive places in a smart manufacturing environment that are not suitable for frequent battery replacements. The use of wireless energy harvesting, which allows sensors/actuators to harvest power from a signal that was intended for data transmission or power transmission, has shown to be an excellent solution to address this issue. The problem with wireless power transfer, however, is that when the distance between transmitter and receiver is large, there will be a significant loss of power. A wide range of techniques have been proposed to overcome the high power loss over long distances, such as waveform design, energy transmission and scheduling, and energy beamforming, which can be implemented at the transmitter and/or the receiver in order to improve the efficiency of wireless power transfer. Unfortunately, the above solutions are not ideal for smart manufacturing environments, since IIoT devices do not have adequate computational capabilities.
Smart manufacturing could benefit from wireless power transfer enabled and enhanced by IRS. Thanks to the deployment of IRSs close to IIoT devices, the issue of high path loss can be effectively alleviated by creating an energy-efficient charging zone for those devices, as depicted in the left side of Fig. 5. When IRSs are deployed correctly in LOS with transmitters and receivers and their beamforming capabilities are fully exploited, the received power of nearby IIoT devices can be substantially increased. An energy receiver can use the IRS’s passive beamforming system to improve the transmission efficiency of wireless energy while simultaneously enhancing the signal strength at an information receiver. Moreover, it realizes the possibility of improving both the rate and energy performance in wireless power transfer by increasing the wireless charging efficiency. This in turn helps to reduce both the transmit power and provides more flexibility in the design of transmit beamforming for the information receivers. As mentioned above, the effectiveness of passive beamformers for wireless power transfer is expected to be crucial in practice. To achieve their benefits, however, they require channel state information at the energy transmitter.
Iii-D Sensing & Localization
Sensing and localization in smart manufacturing create the opportunity for sensors to monitor individual products, providing the possibility for product customisation by tighter control, management and analysis of critical manufacturing parameters. Thus, acquiring the precise location of objects and being able to sense local information and ambient parameters in the environment in industrial settings is becoming indispensable to enable location and sensing-based services and applications. For example, the transparent production and logistics processes of smart manufacturing can be improved significantly by knowing what is happening when, where, and how. Automated guided vehicles can improve production supply, assembly lines (through transport platforms) and warehouse logistics systems. IRS offers great opportunities for precise localization (e.g., using angle of arrival method) and high-resolution sensing solutions in industrial settings since it can actively customize the propagation channels, as illustrated in right side of Fig. 5. The underlying idea of wireless sensing is based on the principle that receivers can identify the effects that sensing targets have on wireless signal propagation. The receiver then exploits the observations to understand the behaviours of targets. Unlike conventional sensing techniques, IRS-assisted sensing creates a controllable radio environment in preferred directions interacting with the sensing targets. As a result, IRS-assisted sensing does not require a LOS link between the receiver and the sensing target [Hu_2020]. On the other hand, in IRS-assisted localization, IRS is deployed between the access point (AP) and receiver in such a way that the AP can investigate a user’s reflected signal through various IRS configurations to achieve the accurate locations of users.
Iii-E Mobile Edge Computing
Another important use of IRS in smart manufacturing is in supporting mobile edge computing (MEC). The mobile edge computing paradigm extends the computing resources from the cloud to the network’s edge. Future smart factories will be equipped with very large number of wireless devices that generate and may need to process large volumes of data in real-time. In many scenarios, these devices do not possess the required processing power and battery capacity to process the data, and hence these processing operations can be offloaded to the network edge (preferred) or to cloud platforms. Compared with the cloud computing paradigm, MEC helps to reduce the latency or end to end delay and avoids unwanted network congestion by reducing the amount of raw data transfer to cloud platforms [mec_Zheng]. Virtual reality and processing high definition images captured by field devices are some examples. One major challenge for this application is associated with communication errors in the computation offloading over wireless links between wireless field devices and computing edge nodes/servers. Especially, devices at the cell edge or behind some blockages incur a high path loss, affecting the offloading rates, leading to higher latency, increased energy consumption, and under-utilization of edge resources.
IRS also emerges here as a possible solution to address these problems associated with the weak wireless links in a smart factory. IRS helps establish strong wireless links between the end devices and the computing edge devices in the shop floor, resulting in reduced packet losses/ re-transmissions and hence enhanced spectral and energy efficiency. As shown in Fig. 5, an IRS-assisted MEC system in a smart factory consists of one or more access points/base stations with co-located edge computing nodes/servers, a large number of passive reflecting elements controlled by its associated IRS controllers and large numbers of field devices. A virtual LOS link with enhanced channel gain can be established between the field devices and the AP by adequately tuning the IRS reflecting elements. These virtual LOS links help to offload data through the BS/AP to the MEC server more quickly. The virtual LOS link also assists in transmitting processed results or decision/control actions back from the server to end nodes, thus shortening the overall end-to-end delay. Currently, in many smart manufacturing applications, local node processing is adopted due to weak communication links in the industrial environment between wireless devices and the edge node/server, resulting in computing resources being idle at the edge server. IRS-assisted wireless channel enhancements help to exploit these powerful computational resources at the edge node better by making them easily accessible to an increased number of wireless field devices. However, the optimal resource allocation at the edge server needs to be jointly carried out with communication resources allocation and the IRS reflection coefficient adjustments to serve a maximum number of wireless field devices.
Iv Challenges & Open Issues
Iv-a Environment-aware Passive Beamforming
One of the main challenges for the succesful application of IRS in smart manufacturing is designing environment-aware and dynamic passive beamforming. In practice the design of IRS passive beamforming is determined by the discrete amplitude and phase-shift levels of each element. A beam steering process requires coordination of the phase control of individual scattering elements. In spite of limited phase shifts available at an individual scattering element of an IRS, an IRS with a large number of scattering elements can enable more flexible phase tuning. However, computational complexity is a price to pay for such flexibility. Moreover, a larger number of scattering elements means greater difficulty in channel estimation, which could hinder efficient phase control.
While exhaustive search may provide the best solution for determining the best amplitude/phase-shift levels, the approach is computationally complex, and may be infeasible for scenarios where energy savings are paramount. Efficient algorithms are therefore required to estimate the channels and control the phase shifts of all scattering elements in real-time following the dynamics of the radio environment in smart manufacturing. A practical solution as opposed to exhaustive search can be achieved by solving the problem with continuous amplitude and phase-shift values, and then calculating the closest discrete values of the obtained solutions [Wu_2020].
Passive reflective beamforming required for an IRS should also be optimized in conjunction with transmit beamforming of base stations and other active network components. In the case of a severely blocked BS-to-node link, the IRS beamforming should be guided by the BS’s transmit beamforming to maximize its signal reflection. However, in situations where the BS-to-node link’s signal attenuation is comparable to the IRS-assisted link’s, the beamforming design of the BS should aim to strike a balance between the node’s and the IRS’s beamforming directions. This mechanism will allow neighboring BSs to serve other nodes outside the IRS’s coverage area with greater flexibility.
To realize practical and efficient IRS beamforming, machine learning approaches can assist to effectively resolve the above problems by using locally observed information of the smart manufacturing environment. The high numbers of scattering elements and their sensors means that a significant amount of information can be collected during channel sensing, facilitating machine learning approaches based on large data sets. The use of data driven machine learning has the potential to minimize the overhead of information exchange between the IRS and active transceivers. In an IRS-based smart manufacturing environment, however, machine learning approaches must be designed to fit the hardware constraints. For example, passive scattering elements have limited computation/communication capability and low-power sensor/actuator nodes cannot handle high computational loads.
Iv-B Radio Resource Management
In IRS-enabled networks, one of the most important tasks is to allocate radio resources optimally. In general, IRS radio resource management is mainly concerned with power allocation, bandwidth allocation, and node-IRS connectivity. Due to the specific dynamics of interference in IRS-enabled wireless environments, power allocation is an essential component for the effective operation of an IRS. In smart manufacturing environments, numerous wireless devices embedded in machines, autonomous vehicles, and the environment coexist in close proximity, making this an even more pressing problem. In order to minimize interference while maximizing the system’s capacity, effective power allocation approaches need to be developed. On the other hand, bandwidth allocation determines the most suitable allocation of users to different subchannels to increase bandwidth efficiency. Due to the frequency-agnostic nature of IRS elements, one common IRS reflection matrix needs to be shared among subchannels, making optimization problematic. In order to address the problem, dynamic passive beamforming can be used. In this scheme, the resource blocks are dynamically assigned to different user groups with different IRS phase shifts for different time slots.
Furthermore, multi-IRS-assisted multi-user communications pose an interesting problem in how to associate users with different IRSs because the user-IRS association schemes in general determine overall network performance. When considering user-BS and user-IRS associations, as well as subchannel assignments, the optimization problem becomes much more complex. It is well known that assigning nodes to different subchannels/IRSs/BSs is an NP-hard problem [wang2020channel]. An exhaustive search of all combinations of association can yield an optimal solution, but would computationally be prohibitively complex, especially in large networks. Hence, low complexity and efficient algorithms are needed in order to achieve a desirable trade-off between performance and complexity. Machine learning can be a powerful tool to obtain a sub-optimal but high-quality solution. However, determining the optimal/sub-optimal efficiency gains introduced to a wireless networked system by IRSs in smart manufacturing, which could serve as a benchmark for low complexity algorithms remains a significant challenge.
Another major challenge in enhancing the wireless network performance using IRSs lies in associating the users/wireless devices to IRS and selecting their communication mode. Some wireless devices may have an excellent direct LOS link with the BS and hence need not be associated with any IRS. A few other devices may make use of single reflection links for better network performance and need to be associated with either the user-side or BS-side IRS (detailed in Section IV-D), while a few others may take advantage of both the single reflection and double reflection links. Most of these wireless devices in a smart factory are highly mobile, which calls for a highly challenging dynamic IRS-user association. To assign users optimally to different IRS, the channel state information (CSI) of all the communication links is essential, which is very difficult to obtain in practice. How beneficial it is to integrate sensing devices to IRS for channel sensing, making it semi-passive, is another question that needs to be investigated in detail. Another challenge lies in establishing a reliable wireless communication link between the IRS controller and the BS. Industry 5.0 forecasts the replacement of the rigid wired communication links in a smart factory. Hence, it is not advisable to employ a wired backhaul link between the IRS and the BS, especially when they are distributed on a shop floor.
To devise optimal resource allocation strategies, IRS-based communication link and network information-theoretic performance limits must be considered. The theoretical limits (e.g., capacity, throughput, latency) of the technology are currently being investigated and many aspects of IRS are still unknown. A substantial amount of effort should be devoted on investigating IRS’s theoretical limits in smart manufacturing, given that it contrasts significantly with conventional wireless environment. Additionally, the accuracy in obtaining channel state information (CSI) is an important factor in IRS-enhanced transmissions due to the nearly passive mode in which IRSs function. A precise estimation of CSI will lead to a successful joint beamforming design and resource allocation, which will further enhance the IRS system performance.
Iv-C Channel Characterization
There are two major challenges regarding the end-to-end analysis of an IRS system. Firstly, to analyze the performance limits of an IRS link, new accurate channel propagation models are needed to obtain the link budget analysis. Path loss models may depend on many parameters, including the size of the IRS or the mutual distances between the transmitter/receiver and the IRS [tang2020wireless].
Secondly, to decode the signal reflected by the IRS, the channel should be properly estimated. In addition of estimating the direct link between the transmitter and receiver, two IRS-assisted channels need to be estimated, i.e., the transmitter-IRS and IRS-receiver channels, and they cannot be separately estimated via traditional training-based approaches in general because IRSs are typically passive and cannot perform channel estimation by themselves. As a result, alternative processes are needed to perform channel estimation, while keeping complexity and overhead of IRS operations as low as possible [renzo2020smart]. The problem becomes even more challenging with large IRS arrays since the time overhead to perform the channel estimation may linearly increase with the number of IRS elements [wang2020channel].
Furthermore, such channel models and estimation processes should consider the specific industrial environment where they will be applied. In a factory floor environment, the presence of metallic surfaces on the machinery furniture and vehicles produces many dissimilarities w.r.t. the channel properties, such as path loss and multipath parameters, compared to other floor configurations, specially when using high frequency signals.
Iv-D Deployment Issues
IRS can be deployed in an industrial wireless environment using different strategies; (i) close to the distributed users or wireless devices (referred to as user-side IRS deployment), (ii) close to the base station or access point (referred to as BS-side IRS deployment), or (iii) in a hybrid style that combines both the user-side IRS deployment and BS-side IRS deployment [you2020deploy]. Each scheme has its pros and cons. User-side IRS deployment at intended locations provides enhanced network coverage mainly for the users or wireless devices within its local vicinity. In contrast, the BS-side deployment can provide extensive network coverage. One of the main motives of using IRS is to provide a virtual LOS link between the base station and the wireless devices where there are obstacles between them. The proper placement of a user-side IRS makes it relatively easy to establish a virtual LOS link between the BS and the intended local users. Placement of a BS-side IRS in such a way to establish a virtual LOS link for all its users is really difficult/ infeasible. The communication signalling overhead between the IRS controller and the BS, required for tuning the reflection coefficients of IRS elements, is relatively low for BS-side deployment since they are closer to the BS. Hybrid IRS deployment strategy aims to combine the advantages of both the user-side and BS-side deployment schemes. It also helps to exploit double reflection links (inter IRS reflection links) to provide more LOS paths between the served users/ wireless devices in a smart factory and the BS/ access point. At the same time, a hybrid deployment scheme brings additional complexity in the design, deployment and management of IRS.
However, the main challenge with these three options is exactly where and how to deploy them. The IRS location should be selected such that there sufficient NLOS paths between the IRS and the base station or access point to enable a high-rank MIMO channel [renzo2020smart]. IRS may be deployed in a centralized or distributed fashion (for a given number of reflecting elements), and it’s not yet clear which approach is the best for an industrial environment. IRS offers passive reflection without amplification; hence, these reflected signals suffer severe product-distance/ double path loss [IRS_tutorial]. Therefore IRS is usually preferred to be positioned in the vicinity of the BS/AP or the wireless end devices/users for minimal path loss. An IRS’s low cost provides the flexibility to opt for a dense deployment on a factory floor if required. However, their joint network performance optimization will be a challenging task.
In addition to the required wireless channel enhancements, IRS deployment in a smart factory needs to consider various other practical challenges. Industrial environments or spaces in which it will or can be deployed, distribution of wireless devices/ users on the shop floor, cost of installation and maintenance, hindrance created by IRSs to the flexible operation of the smart factory/ production line, IRS’s flexibility for relocation represent some of these challenges.
In this article, we have discussed the prospects of IRS-aided wireless networks in a smart manufacturing environment to support the evolution towards Industry 5.0 by unfolding their potential features and advantages through different wireless network scenarios. As IRS technology is still in its infancy, we have elaborated on the most pressing challenges as well as the potential opportunities for research into future IRS-aided wireless factory automation. Thus, it is hoped that this paper will serve as a useful and inspiring resource for future research on IRS-based smart manufacturing to unlock its full potential in a future industrial environment.