Research groups have developed several procedures to apply IoT to the issue of thermal comfort, with points of view that differ greatly from each other. Consequently, here, the authors propose a review of the most influential scientific publications that mix the fields of the IoT and thermal comfort, so the authors create added value to the scientific community that will identify the streams that have been generated naturally. In this review, the authors identified three main ways in which the IoT can be applied to thermal comfort studies: through hardware, through simulation and, still marginally, through the new Crowdsensing paradigm.
This literature review focuses in particular on the role played by the occupants, describing how researchers are moving toward a people-centric approach in which the user’s opinions and preferences matter. This point of view is slowly streaking ahead as opposed to the “one-fits-all” perspective, in which people were passive receivers of a thermal environment chosen by standards.
According to The National Human Activity Pattern Survey, conducted in 2001 but still widely considered among the scientific community, in modern society, people tend to spend 87% of their time indoors [9
]. Identifying an indoor thermal zone in which people are comfortable is an ever-current topic. The modern perception of the problem started in the 1970s, presenting two different schools of thought from the beginning. On the one hand, an initiator of the field of thermal comfort studies was Fanger, who introduced a steady-state model to relate thermal sensation to subjective parameters, such as activity and clothing levels [10
]. Since the introduction of Fanger’s comfort model, who obtained his results through experimentation in climate chambers, the accepted parameters to evaluate thermal comfort have been: indoor air temperature, mean radiant temperature, indoor relative humidity, internal air velocity, clothing insulation and metabolic rate. These parameters are incorporated in the so-called Predicted Mean Vote (PMV), an index which aims to evaluate and predict the state of thermal well-being of occupants. This method is based on the heat balance of the human body that, as explained by Höppe [11
], comprehends both the physical heat transfer process and some thermophysiological mechanisms such as sweating, shivering, etc. Fanger’s model introduced the aforementioned parameter called Predicted Mean Vote (PMV) to also relate thermal sensation to subjective parameters, and a parameter called Predicted Percentage of Dissatisfied (PPD), to estimate how many occupants would feel thermally dissatisfied. Subsequent studies showed that the PMV model fails to describe comfort temperature outside of the exact conditions of Fanger’s experiment, because it does not consider the effect of people’s adaptation. For this reason, Fanger and Toftum [12
] elaborated another model which include an expectancy factor, e
. On the other hand, Nicol and Humphreys [13
] supported that experimentations in the climate chamber do not consider other complex parameters, such as time sequence and social factors, that are fundamental in the study of thermal comfort. It was the beginning of the so-called Adaptive Thermal Comfort Model. Supporters of the latter have also been de Dear and Brager [14
], who stated that in naturally ventilated buildings, the range of accepted temperature depends in great part to other parameters, such as external temperature and people’s adaptability. Also, according to more recent studies, in the case of free-running buildings and in mixed mode ventilated buildings, i.e., in case people can adjust or change their environment with operable windows or through metabolic alterations, researchers tend to consider the use of the steady-state model inadequate [15
]. Besides, the adaptive comfort standard provides thermal design guidance and indicates the acceptable and the optimum indoor temperature ranges depending on the climate zone [17
Today, both models are regulated by ISO 7730:2005 [19
], EN 16798-1:2019 [20
] and ASHRAE 55:2017 [21
] regulations and are accepted and used worldwide, although significant differences among them keep being underlined [22
]. It is worth mentioning that current experiments are investigating a novel model, that aims to explain the effect of short-term temperature changes on human’s thermal comfort, called alliesthesia. According to this research line, thermal comfort is conceived as dynamic, and the environment as transient and non-uniform [23
]. Recent studies included other parameters in the perception of thermal comfort, i.e., thermal alliesthesia and habituation, so that new models are starting to include both a static and a transient component [26
In this context, IoT represents a great opportunity to face the issue of thermal comfort in buildings. In the last decades, IoT-based systems have already become a fundamental ally in proposing new smart solutions to ameliorate energy consumption in buildings [27
]. Taking a step forward, IoT applied to thermal comfort allows to consider the energy-matter not only passively from the point of view of the building itself, but also in its relationship with the user. Although this review does not aim to investigate this energetic topic, in many studies, thermal comfort is presented in relation to energy efficiency, hence its incorporation in this document.
Furthermore, several studies demonstrate that thermal discomfort has effects both on health and human performance. Lan et al. [28
] conducted some neuro-behavioural tests on users doing office work activity to monitor their productivity. It was possible to measure a quantitative relationship between productivity losses and thermal sensation and to conclude that thermal discomfort leads to reduced performance, both in case of feeling too cold and too warm. In another study, Lan et al. [29
] showed that in thermal discomfort, users reported the so-called sick building syndrome (SBS) symptoms, as well as a negative influence on the mood, the willingness to make efforts and the air quality perception. It is noteworthy that the negative effects on health and performance are not related to the distraction of subjects, but to physiological reactions to discomfort, i.e., they will occur even if subjects have become adaptively habituated to an unfavourable environment.
IoT technologies bring a technological ecosystem that has to be understood to be aware of the role of these systems in the buildings. These principles more based in computer science tend to follow a common structure. Authors have used this common structure to help to classify the different works found in the literature. IoT in buildings is composed of three layers [30
Sensing or Perception Layer, dedicated to the acquisition of information.
Network Layer, that connect data and manage the control centre.
Application Layer, that is supposed to achieve the energy management.
From the literature, it is possible to see that the studies on IoT and thermal comfort differ from each other because of one or more of these layers but maintain the general structure. For what concerns the first layer, the sensing process through the use of IoT can collect data through sensors both from the environment (in fixed locations) or from users (with personal devices carried by the participants). The literature considered also differ in the way of analysing data, although the most recurrent methods are mathematical tools and virtual modelling. About the networking, in the literature available, its complexity can vary from basic sensor networks to complete architectures that incorporate different components.
The application layer consists in the generation of outputs that, like in the sensor layer, could involve the environment, i.e., a change in setting the air temperature, or the single persons, with personal devices suitable for improving the thermal comfort of each one.
These differences also reflect the semantic duality in the conception of the IoT itself. Visions of the IoT often embraced a “Things oriented” perspective or an “Internet oriented” perspective, intending to create a bridge between the physical and digital world [31
]. Taking in mind this duality, it is possible to identify studies that involve physical smart objects among thermal comfort literature, like smart HVAC systems, personal devices and studies based on virtual modelling and simulation. In this review, the two categories will be referred to as “studies with IoT hardware” and the “Building Simulation Model”. Of course, one method does not necessarily exclude the other. So, besides these two approaches, the authors will present a few studies that use data to validate the computer model simulation, and others that propose a unified platform to combine the methods.
However, in both cases, often people are considered as passive parts of the system, one variable of the study out of many others. Differences among individuals are not considered enough and people’s response to different stimuli are often inaccurate or undermined. For these reasons, in the last decade, studies have developed a new approach to the thermal comfort matter, in which the main focus is people’s opinion and perception: The Crowdsensing paradigm.
“At present, public groups are frequently portrayed as ignorant or irrational in the face of scientific progress” [32
]. In the scientific dimension, on most occasions, “the crowd” is seen as a passive receiver of progress’ benefits or established standards. Often, users are tacitly seen by scientists as ignorant, misled or simply contrary. Even so, according to sociologists such as Williams [33
], it is important to form a citizen’s view of science, i.e., to create a sustainable development that foments a people-oriented perspective. This is particularly relevant in the investigation about the energy behaviour of buildings, in which the final aim is users’ thermal comfort. Among people’s thermal preferences, there are differences both between two people in the same environment on the same occasion (inter-individual differences) and between the same individual in the same environment on different occasions (intra-individuals differences) [34
]. That is why the extension to the thermal comfort of the new paradigm of Crowdsensing, based on every individual’s perceptions and impressions, is utterly urgent. As Erickson and Cerpa [35
] stated: “At the individual level, occupants are 100% accurate when determining if they are comfortable—only their opinion matters”.
According to Ganti et al. [36
], devices that are customer-centric, like smartphones and music players, will be the evolution of the IoT, allowing an upgrade to a societal scale.
Crowdsensing could be classified, depending on users’ involvement, into (1) participatory and (2) opportunistic sensing. In participatory sensing, people are directly involved, through reporting data, taking pictures, etc. In opportunistic sensing, the involvement is minimal, the explicit action of users is not necessary. A clear example is user-generated data in mobile social network services [37
], that helps to understand the dynamics of our society.
Before the concept of Crowdsensing appeared, Dutta et al. [38
] explained the problem and the necessity of this kind of solution well: “Mobile participatory sensing uses consumer electronics to capture, process and disseminate sensor data, complementing alternative architectures by ‘filling the gaps’ where people go but sensor infrastructure has not yet been installed”. It is clear that thermal sensors cannot be everywhere. In public places and workspaces, it surely is more common, but in most of the locations that users visit, the “gap” is unavoidable. In other words, as a way to extend researchers’ results to people, one should consider other parameters to measure the thermal comfort. Nicol and Humphreys [39
] explained this well: “And finally… do we really need to specify indoor climate? Given a full understanding of the mechanism at work, it may eventually be possible to produce thermal standards for building which do not resort to specification of the indoor climate. The characteristic of a building in relation to the local climate may be sufficient.”
One of the most fundamental constituents of the IoT is domotics, or home automation, popularly called smart home. Domotics refers to several fields, namely smart appliances, home entertainment, control and connectivity, comfort and lighting, security and, of course, energy management. Smart home, that has to be intended broadly and not just referred to in a residential context, is an incredibly fruitful market sector, considering that just in the last three years, it has doubled its sales and they are expected to grow by 60% by 2024 [40
]. Up to now, almost 95 million homes are considered smart from an energetic point of view, thanks to the diffusion of new technologies affordable for the majority of people (or at least in wealthy countries).
Smart home is a trend from 2013, but it reached its higher diffusion in the last years, due to the commercial success of devices such as Amazon Alexa
. There is something revolutionary in it, since in a very short time, IoT switched its status from news of scientific interest to fashionable for users. Alexa is an intelligent virtual assistant that works through a smart speaker, Echo
. Behind the basic functionality of the speaker itself (plays music, reads audiobooks, streams podcasts etc.), what is interesting in its configuration is the capability of controlling other smart devices. In other words, Alexa can be considered itself a home automation system, which interacts with people while wiring all the smart objects around it. It was released in November 2014, but it started to be universally popular from 2017 [41
], probably due to massive advertising campaigns that gave a sort of sentimental feeling connected to its use, such as improving family members’ relationships, summarized by the concept of “sharing is caring” [42
]. In January 2019, Amazon stated that they sold an impressive amount of 100 million Alexa devices [43
]. Simply, it has been a revolution that shifted people’s attention to domotics. Figure 1
is a graph created by the authors using Google Trends data [41
] that shows how much Alexa influenced the world of home automation: peaks in people’s interest almost perfectly overlap since the end of 2016.
In Figure 1
, the comparison also includes the trend of Google Nest. It does not affect the diffusion of smart automation as much as the commercial phenomenon of Alexa, but still, it helps to explain some peaks of penetration. In fact, Nest is another trend in the modern market, being the most popular smart thermostat worldwide. Nest is not just a programmable thermostat—it learns from every user’s preferences, being able to program itself as a consequence. It also learns about the time needed for a specific home to warm up or cool down, managing a proper schedule for every situation. Besides, Nest has a raise awareness aim, inasmuch it shows users when they are in an energy-efficient mode [44
]. It can be said that Nest learns from us and we learn from Nest. Giving a brief context, Nest was born as an invention of Tony Fadell and Mark Rogers who, in 2010, founded the Nest Labs. The brand immediately had a discrete success, and Google bought it in 2014 and it became even more popular, helped by other factors such as improvement in wireless broadband technology’s diffusion and the first standards about smart meters’ requirements in offices and homes.
Several other brands deal with the energy topic in home and office automation. Up to now, the main competitor has been Honeywell, connected with Apple Homekit (the famous voice assistant Siri
), but recently, also Ecobee, that has a built-in Alexa Voice Service and room sensors capable of detecting occupants [45
]. Other famous enterprises like Siemens, Samsung, Xiaomi, KeenHome, etc., are trying to take a piece of this market sector but, as shown in Figure 2
, the magnitude of Nest and Honeywell is hard to reach.
It is worth noting that there exists a gap between types of applications in the market and the effective use of people, i.e., something is hindering the utilization of IoT in homes and workplaces. Although they will not be explored in this paper, researchers identified the potential reasons, namely energy consumption, device connectivity, safety and, of course, security [46
]. The most common concern is cybersecurity: people are afraid of being spied on in their own homes and workplaces and this is a big obstacle in the diffusion of the aforementioned devices. Users agree about the devices’ function of making their lives simpler, but they do not want to introduce new risks [47
1.2. Paper’s Structure
The thermal comfort studies in the last decades have been strictly related to the IoT development. Every technological advance in the IoT world corresponded to a big step forward in the understanding of the comfort conditions. Growing accuracy of the sensing tools, new technological elements in the loop and more powerful simulators marked the development of modern thermal comfort investigation. The central question of this work is understanding whether or not the novel paradigm of the IoT, the Crowdsensing, can represent a future path in the investigation about thermal comfort. Studies based on thermal surveys are normally conducted in situ, mainly because of the necessity to measure environmental parameters with proper devices. Crowdsensing broke this physical limit for other fields of study, so it would be an enormous improvement to extend its potentialities also for thermal comfort, allowing the researchers to conduct their surveys remotely.
To answer this question, the method that seemed more appropriate was analysing both how the IoT was applied to the thermal comfort study in the last ten years and how the Crowdsensing is being applied to the other fields of study.
The research methods of the articles for this review follow this duality. Applications of IoT to buildings and to energy efficiency have been the focus of the authors’ research group for years, so the starting points were the articles that the authors considered inspiring or noteworthy. It was a natural consequence noting that many of these articles mentioned the same researchers or the same investigation groups, and the main contributions were found. Then, research through academic platforms, mainly Scopus and partly Google Scholar, completed and confirmed the results.
The approach for the Crowdsensing part was quite different. As the phenomenon is quite new and not universally known, the starting point was searching for articles about Crowdsensing in general, through the mentioned academic platforms. The most cited papers were selected and the major contributions in terms of explication of how the paradigm works. From this base, articles about application to the thermal comfort topic were searched, though the keywords “Crowdsensing + Thermal Comfort”, “Crowdsensing + IoT + Thermal Comfort”, and so forth. The research was unsuccessful: the Crowdsensing application to thermal comfort was not investigated previously. After trying with “Crowdsourcing + Thermal Comfort” and “Participatory sensing + Thermal Comfort”, some results were obtained: something was moving towards that direction.
The paper is structured as follows. Section 2
investigates the thermal comfort studies with IoT hardware. The section presents a first part in which studies that propose an improvement in thermal comfort conditions are analysed, both obtained through changes in the whole environment and in the micro-environments around occupants. The second part of the section presents articles that use hardware with monitoring aims, as demonstration studies or sustained by experimentations.
explores the mimicking of IoT sensors and comfort using Building Simulation Models. The first part of the chapter simulates variations in thermal comforts due to changes both in the physical surrounding and in users’ behaviour. The second part of the section proposes studies which incorporate the simulation tools in a more complex IoT architecture, and collects studies that simulate thermal comfort in future climates. Section 4
describes the new Crowdsensing trend and explores its application to other fields of study. Tools to involve people in the loop are briefly explained and the rest of the chapter is dedicated to the first applications to the thermal comfort matter. Section 5
shows quantitative results, mainly in terms of analysis of the scientific community’s interest in these subjects. This is useful to understand if the Crowdsensing paradigm has space to grow as a scientific topic and as an application to the thermal comfort. Section 6
contains discussions about the main topics. Section 7
presents possible directions for future works.