Cyber Physical Systems (CPS) are becoming more ubiquitous, complex and powerful. Inherent benefit and comfort come with an environmental impact that is usually ignored when implementing these systems. This short paper intends to raise awareness of this impact due to the increasing number of connected devices and the data volume produced by their use. We rely on a specific smart home case study to illustrate the potential of considering life-cycle analysis of both the physical devices and data to set up a CPS. Our research in progress targets a design approach to converge into an equilibrium between utility, performance and minor environmental impact of smart systems.
Cyber Physical Systems (CPS) are becoming part of critical infrastructures for a large variety
of contexts including industry and our daily life (e.g., supply chain management, smart cities,
cars, etc.). According to the 2020 Cisco Report [
Life-cyle assessment (LCA) is a systematic, standardized (ISO 14040 [
Environmental concerns have been considered in several contexts. Regarding ICT in general,
in [
The aim of this paper is to question the "more is better" approach for CPS, bringing the environmental impact assessment as a first-class citizen when designing them. Disclosing the infrastructure and information life-cycle may help designers analyze the trade-ofs of CPS. We aim to go towards eco-friendly smart systems by design, to converge into an equilibrium between benefit and environmental impact.
To illustrate our purpose, we briefly discuss a smart home case: Amiqual4Home, A4H [
Smart Homes include several connected devices to provide a wide range of services from smart
management of water or electricity consumption to services for comfort like light or music
management [
In the following section, we introduce an example to raise awareness that even for apparently simple services, the required system support becomes non negligible when we consider their use at a large scale (instances and time span1) and environmental criteria such as the greenhouse gas reduction requirements and, more generally, planet boundaries. In this perspective, we put forward some design questions.
A4H is an experimental apartment2 configured to unobtrusively monitor its consumption (e.g.
electricity) and the inhabitant daily living activities such as cooking, sleeping or washing dishes.
It is a two-story, one-bedroom, one-ofice and two-bathroom apartment equipped with 219
sensors and actuators (see Table 1). The ContextAct@A4H dataset [
To clarify the environmental impact of such a CPS, it is necessary to consider the life-cycle of the devices installed in the apartment, the produced data and the scenario of use. The life-cycle of a device includes its material extraction, manufacturing, transportation, use and end-of-life. In addition to sensing and communication devices (gateways, routers), user devices, like a smartphone or a tablet, are used to control and to monitor all devices, serving as a control panel for the house. Together, they compose the smart home’s infrastructure.
The data life-cycle includes its production, transport, processing, storage and use. Data is
transferred from the sensing devices to the processing center at a cloud or, as in A4H, at the
edge. Sensors in A4H use diferent M2M communication protocols (e.g. KNX, MQQT), most
of which use a hub or gateway to create a sensor network. As illustrated in Figure 1, a server
(local in A4H) analyzes and stores sensor data and can communicate back with actuators when
needed. Data can be stored in diferent formats, for extended periods of time.
1e.g. millions of smart buildings during many years
2A4H has an area of 87m2. The reader may refer to [
It is important to recognize that system design decisions have an impact in the environmental footprint of CPS. In this paper, we argue that these decisions should be considered as part of the ifrst design decisions together with the functional analysis and expected benefits. Some choices of A4H, which is used for research purposes, are discussed here.
System configuration: deciding which and how many sensing devices should be installed in a smart home is a key decision. The number of devices can quickly grow, but the extra benefits can be minimal. In A4H redundant sensors were installed to make the system more robust and to create a versatile dataset that could be used by researchers in diferent areas, resulting in a large number of sensors installed. This configuration is related to functional requirements. In the ContextAct@A4H experiment, the considered functional service was a system to monitor elder activities at home, to support independence at home while still alerting a support network of unusual behaviours or events. The contact sensors installed at A4H, allowed monitoring the mobility, kitchen activity and bathroom use patterns. In contrast, if the main concern is the person going out and getting lost, one sensor at the entrance door and one at the door leading to the patio may be suficient for initial alerts.
In final-user smart homes, it would be important to identify precise functional requirements
and appropriate precision as well as the required fault-tolerance to limit the number of devices.
The choice of the devices comes with diferent impacts depending on the manufacturer and
on its complexity [
To illustrate our purpose, we refer to the PEP (LCA according to the ISO standard [
In the table in Fig. 2-right, we show some configurations considering various combinations of sensors, gateway and PC and their estimated energy footprint. Configurations 1 to 3 monitor all the doors and windows of the apartment whereas configuration 4 targets exclusively the front and back doors. Let’s notice that as the expected lifetime of a PC is 5 years, two of them are required for the 10 years period.
While we use the energy consumption in this paper, other criteria as resource depletion and the CO2 footprint are important. It is worth noting, that the CO2 footprint of the electricity depends on the so-called energy mix. This evolves in time and difers from one country to another and even from one region to another in some countries. According to ElectricityMap [17], accessed on March 24, 2022 at 11am GMT, at that moment, the estimated equivalent CO2 for a kwh (gCO2eqkwh) is 109 gCO2eqkwh in France, whereas it is only 27 gCO2eqkwh in Quebec. We note that in Australia there is a big diference between Flinders Island and Queensland, reporting 12 and 709 gCO2eqkwh respectively.
Multiple factors and trade-ofs are involved in choosing a system configuration, including
device complexity and additional services they might provide [
Sampling rate and data footprint Sampling rate greatly impacts the amount of data
produced and transmitted. Sampling rate can be set in Hz (samples per second) or configured to
send a new measure when an event is detected. Sampling rate also has an efect in the algorithms
used to analyze the data, which need to adapt to the defined sampling approach, for example,
by using diferent windows of analysis [ 18]. In this case study, we briefly discuss data footprint
considering a 10-year lifetime. A sensor measure is a tuple < , _, >,
12 Bytes, considering integer values and ignoring any protocol overhead. Based on this, we
estimate the volume of data produced by the 21 contact sensors of A4H during 10 years with
four sampling approaches (Table 2). The Event-based sampling approach represents the scenario
where the device sends data only when a state change occurs. Relying on the three-week
experiment of A4H logged in the ContextAct@A4H dataset [
As mentioned, sensed data is transferred to and processed in a server. Architectures for data processing and learning models range from in device treatment, edge processing to hybrid or completely cloud-based systems. Each approach has its environmental footprint. In either cases, the sensor sampling approach entails a big diference in the total data produced, which is related to the resource consumption in the data life-cycle. The architecture choice depends on system and functional factors such as privacy requirements and whether or not patterns at the population level need to be analyzed (for instance, for improving public health policies).
We have described some of the CPS design decisions in A4H that afect the environmental impact of the system and we have shown how small changes can have significant impacts in its total energy consumption. This list is of course non-exhaustive but intends to raise awareness of how we can consider a life-cycle assessment of both the devices and the data in the analysis of CPS. Neither the analysis nor the decisions are easy, but the environmental factors should be one of the criteria considered along-side other requirements and restrictions like privacy, costs, and functional requirements.
It is worth to state the undeniable benefits that cyber-physical systems (CPSs) have brought in both the industry and our daily life in diferent domains such as healthcare, supply change management, security or living places. Our position is that more than ever, considering the huge potential of CPS and the current environmental challenges, the choices for introducing "smart" functions should be analysed very carefully. We claim for a "fair trade" approach with a people-planet-system perspective that integrates an impact/benefit analysis including the whole life-cycle of the smart function. Performing a Life-Cycle Assessment for a complex cyber-physical system is challenging. Nevertheless, even if a complete LCA cannot be realized, conducting an approximate study will contribute to increase the environmental responsibility and sustainability during system development. As cyber-physical systems tend to be "invisible", there is a need for awareness of the underlying infrastructure and required resources, early in the design phases. Of course, this is not to say that functionality should be compromised. A form of systemic thinking that considers both the functional definition and the environmental impact when designing solutions would help. In this paper we discussed some choices that may influence the environmental footprint if they are scaled.
In future work, we’ll focus on improving the approaches in design phases so as to introduce environment awareness together with the actual goal of the system (comfort, health, energy savings, etc.) and the induced benefits. Rebound efects should also be considered as the addition of a new service may enable other needs in terms of services, devices or data. The use of scenarios may also be useful to inquire about the various uses of the physical infrastructure. The context of the deployment is also important as the use of the system will difer from one country to another because of the origin of the electricity (nuclear, carbon, renewable, etc.). Indeed, the boundaries of the system has to be well defined as the LCA has to be feasible when looking for the required environmental information. We put forward a big challenge to motivate designers and developers to find the good enough level of services and not go to a more is better approach just because it is feasible from a technological and economical perspective. [12] Hager SE, Product environmental profile - radio window contact battery-powered, knx, 2020. URL: http://register.pep-ecopassport.org/fileadmin/tx_pepmanagement/user_upload/ HAGE-00496-V01.01-EN_pdfpep.pdf. [13] Schneider Electric Industries, Product environmental profile - window / door sensor, 2018.
URL: https://www.se.com/au/en/download/document/ENVPEP1804008EN/. [14] Hager SE, Product environmental profile - domovea server, 2021. URL: https://hager.com/nl/catalogus/download/product/asset/file/PEP_HAGER_TJA470_ HAGE-00444-V01.01-EN_17022020.PDF/. [15] Schneider Electric Industries, Product environmental profile - spacelogic knx ip router, 2020. URL: https://download.schneider-electric.com/files?p_enDocType= Product+environmental&p_File_Name=ENVPEP1909001EN.pdf&p_Doc_Ref= ENVPEP1909001EN. [16] Fujitsu, Product life cycle assessment 2021, fujitsu esprimo p9010 desktop pc, 2021. URL: https://www.fujitsu.com/global/documents/about/environment/Life%20cycle% 20analyses%20of%20Fujitsu%20Desktop%20ESPRIMO%20P9010%20June%202021.pdf. [17] electricityMap, Climate impact by area map, 2022. URL: https://app.electricitymap.org/map, accessed online on March 24, 2022. [18] N. C. Krishnan, D. J. Cook, Activity recognition on streaming sensor data, Pervasive and Mobile Computing 10 (2014) 138–154. doi:10.1016/j.pmcj.2012.07.003.