1. Introduction
Roughly one-third of the edible food produced for human consumption is lost or wasted globally, which is approximately 1.3 billion tons per year [
1]. The impact of food waste is not only financial, but also environmental and social. Indeed, it involves gas emission, oil degradation and energy consumption, in addition to the paradox related to food access difficulties in some less developed areas. Food losses in industrialized countries are as high as those in developing countries; however, in the latter more than 40% of the food losses occur at post-harvest and processing levels, whereas in industrialized countries, more than 40% of the food losses occur at retail and consumer levels. The amount of food waste at the consumer level is 222 million tons, nearly as high as the total net food production in sub-Saharan Africa (230 million tons) [
2]. For this reason, the promotion of consumer eco-efficient behavior and an effective, easy-to-use professional and household food management system are very important for reducing the extent of food loss.
A recent survey concerning food waste in Italy [
3] revealed that 48.2% of the wasted food had passed the expiration date, whereas 36.7% and 11.5% had been left respectively in the fridge or pantry for too long. In addition, 11.5% of waste was due to errors in meal planning and purchasing. From the same survey, a possible solution for the reduction of food waste emerged, since 46.5% of respondents said that intelligent refrigerators or cupboards would help them in planning an optimized shopping list, and determining what is in the house and what is about to expire. Suggestions to improve the use of domestic food resources were also identified: some respondents considered it helpful to receive recipes for reusing leftovers, many would have liked to get advice on how to best preserve their own food, and several said they would like to receive information on the freshness and life of the product. Most users also stated they would prefer to use modern technology to receive such information, such as e-mail and dedicated apps on mobile devices.
Motivated by the socio-economic problems discussed above and taking advantage of the survey recommendation described above, the current work depicts in detail a set of smart services deployed to enhance the functionality of a domestic chest freezer, in terms of food management, eco-efficiency and energy saving. The features implemented, which aim to reduce food and energy waste, are designed not only for the domestic environment, but also for the commercial and catering field, as the system is easily scalable. The promotion of eco-efficient behavior and smart food management is fundamentally the inner intelligence of the system.
Chest freezers are usually under a cost-driven manufacturing process, to deliver low-cost products in the mainstream markets. HighChest thus aims to drive the introduction of Internet of things (IoT) and advanced design to turn traditional chest freezers into advanced appliances able to induce eco-friendly and socially sustainable user behaviors.
2. Related Works
Chest freezers come in a wide range of sizes, ranging from approximately 1.5 cubic meters up to 8 cubic meters. They take up more space, but remain more energy efficient than upright and refrigerator freezers, thanks to their horizontal configuration. In particular, chest freezers are more energy-efficient because cool air does not rise very quickly and therefore does not escape every time the lid is opened [
4].
Smart systems for food management in household appliances are usually available for refrigerators, but few freezers perform food management or food quality assessment. To the best of the authors’ knowledge, research concerning smart freezers has mainly focused on upright freezers for domestic use. The authors found no research concerning advanced food management or promotion of eco- and energy-friendly behaviors in relation to chest freezers, either for private or for professional use. Gu et al. [
5] developed a smart fridge system to provide a well-balanced eating habits recommendation service. The system monitors food currently inside a refrigerator via radio-frequency identification (RFID) tags and a reader. The fact that the proposed system requires RFID tags is a downside, however, as such tags are rarely included in local regulations or provided with food packages available on the mass market. Luo et al. [
6] also focused on healthy nutritional habits, introducing a system for shopping lists and recipe recommendations, based on food stored in the refrigerator and users’ dietary profiles. The system uses a local database storing information on users’ nutrition habits, weight, height, age, medical record and allergies, among other things. It is equipped with a barcode reader for the tracking of products and a touch screen on the fridge’s door that acts as a user interface. Rouillard [
7] proposed a smartphone-based inventory management solution, which allows users to remotely manage food stocks. The software recognizes food through barcode scanning using cameras or by users’ vocal inputs using a speech recognition service. Additional services are provided, including reminders of expiration times of food in both the fridge and in cupboards sent using SMS and instant messaging.
The RFID Smart Fridge proposed by Noutchet [
8] is a modified fridge that includes an RFID scanner tracking all food transactions and providing reminders of product expiration, automatic shopping list generation and order replenishment. Unfortunately, these services require RFID tags on each food package. Sandholm et al. [
9] realized a system based on a fridge-embedded webcam to leverage Google Search By Image services to allow automatic recognition of foods. The system aims to improve acceptance and reduce the obtrusiveness of the food recognition process, leveraging the use of cameras. Food recognition is based on the identification of logos and text on packages, which suffers from point-of-view constraints. Sandholm et al. also proposed an automatic item position detection using an infrared (IR) proximity sensor embedded in the freezer compartment.
Murata et al. [
10] proposed an eco-feedback system to save power by detecting the opening and closing of the fridge door. The refrigerator is instrumented with internal and external IR distance sensors and a magnetic door switch sensor, to recognize opening and closing events. External IR sensors detect human approaches to the fridge, while internal IR detects food presence in specific positions.
Investigating the Espacenet Patent Data base, the US patent No US 2012/0278190 [
11] and the application patent No WO 2016/098124 [
12] claim the use of display interfaces on the appliance and tablets and smartphones for the direct or remote control of the refrigerator. Commercial appliances nowadays allow accessing the food database, control and managing expiration dates and alarms or performing on-line shopping or access media. Among the commercially-available smart refrigerators, it is worth mentioning the Samsung Family Hub™ [
13], which is provided with a wide touch screen and dedicated Apps for on-line food shopping, media streaming, advanced user touch, and vocal interaction. The LG LFX31995ST smart fridge [
14] is instrumented with a tablet for food freshness tracking, inventory, family water intake tracker and daily recipe suggestion. In some cases, cameras are used to see the fridge content, while embedded electronic interfaces allow the managing and setting of alarms in case of power supply fault or food expiration. The LG Smart ThinQ™ Refrigerator [
15], instead, embeds an 8-inch Wi-Fi LCD screen that serves as a control panel for the refrigerator, as well as an information hub and potential family organization center. The Smart ThinQ™ provides energy efficiency, a family water intake tracker and daily recipe suggestions. Nevertheless, all of the previously mentioned food management services require the manual input of information from the users. For example, describing and labeling all food, and imputing expiration dates are completely manual procedures. A dedicated Android and iOS application makes the inventory and product expiry data remotely accessible, and a ZigBee radio device enables communication with third-party smart plugs for energy monitoring.
These kinds of freezers are connected to Internet using wireless communication like Wi-Fi and GSM, in the IoT framework, to deliver alarms and information on the inner food stored, directly to the users by means of App on smartphone or tablets. Innovation often regards the process for the food recognition, the organization of a database, the technology and processes for the management of expiring foods. Nevertheless, smart systems for food management were found only for refrigerators and there are no smart chest freezers in the IoT framework. Taking in account for the current state of the art, as in
Section 2, we propose the HighChest freezer as an IoT device for advanced food management and the promotion of eco-friendly behaviors. We do believe that a smart freezer with advanced user interfaces and services for the automated food and appliance management would effectively promote a responsible use of the food and energy.
According to the current literature, many open issues remain concerning the development of highly acceptable and effective smart freezers able to improve food management, and to promote eco-efficient and socially sustainable user behaviors. A comparison between the main reviewed existing smart fridge systems and the device proposed in this work is presented in
Table 1. Comparison is only performed in relation to refrigerators, since, to our best knowledge, smart freezer technologies do not exist in literature.
6. Results
The localization system was tested to assess (i) food positioning accuracy for easy retrieval and (ii) the weight fluctuations that occur due to air humidity intake. Food localization accuracy directly affects the time spent by users seeking the desired food, thus indirectly affecting the temperature and energy loss due to the lid being open. Humid air is responsible for frost stratification and increases the chest weight, generating bias in relation to food weight measures.
Each load cell was calibrated using a certified sample weight of 1 kg. Six different samples of 0.1 kg, 0.2 kg, 0.5 kg, 1.0 kg, 1.5 kg and 4.0 kg were then used to assess the localization accuracy of the system over four different positions for each compartment (each sample tested 24 times). For each known position, the localization error was calculated, in the form of the Euclidean distance between measured and real
x and
y coordinates. This difference was more pronounced for small weights and decreased with increasing weight. The average localization errors among all compartments were 22.94 ± 2.49 cm, 9.42 ± 4.18 cm, 5.60 ± 0.93 cm, 3.60 ± 0.43 cm, 2.43 ± 1.27 cm and 1.87 ± 1.26 cm for weights of 0.1 kg, 0.2 kg, 0.5 kg, 1.0 kg, 1.5 kg and 4.0 kg, respectively. In
Figure 10 and
Table 5, the error is presented for weights more than 0.5 kg, for which the success rate in identifying the right compartment was 100%. In the case of 0.2 kg and 0.1 kg, the success rate in identifying the right compartment was respectively about 96% and 50%.
An experimental procedure was performed, in order to measure the humid air intake, in turn measuring the freezer weight variation when keeping the lid open. Humid air enters in the fridge every time the lid is opened for food insertion or retrieval, and generates frost on the inner surfaces, increasing the chest weight and reducing the energy efficiency of the freezer, due to the low thermic conductivity of ice. The inner temperature was set to −25 °C, whereas the room temperature and relative humidity were +22 °C and 30%, respectively. The worst case of empty freezer was selected, to maximize the internal/external air exchange. A continuous weight measurement was performed, starting with the lid closed and ending one minute after the lid was open. The measure was repeated 10 times to achieve a significant data set. Neglecting the transience due to the effect of manual opening, the average bias due to opening was −0.56 kg (
Figure 11). This bias was used to filter out the effect of humid air intake on food weight estimation using the integrated scale.
7. Discussion
The smart functionalities previously described were implemented in a working prototype. The aim of our work was to embed capabilities for food management and energy saving in a domestic chest freezer, overcoming limitations found in current technology, which is so far available only in the field of augmented refrigerators. Expert evaluation and user trials were conducted by a multidisciplinary team, composed of designers, researchers and expert engineers from Whirlpool, to extrapolate key design points. These outcomes drove the prototype design proposal and the technological developments. The improvements carried out were oriented towards smart food and space management, as well as the development of an effective and intuitive interface, in order to allow conscious space management of the stored foodstuffs, promote eco-efficient use and reduce waste related to improper storage. In terms of sustainability, the developed prototype was also beneficial due to its utilization of recycled materials (such as PET) for its plastic parts and its application of PUR from renewable sources for its insulating material.
Some reviewed works proposed alternative tracking strategies in place of barcodes [
3,
6,
7,
8] because of the lack of usability and narrowness of information offered by barcodes. Nonetheless, the use of barcode technology allows the exploitation of the food packaging labels present on the majority of European Point of Sale (POS) commercial products. The employment of a pre-existing code and a low-cost barcode reader makes this codification preferable even to radio-identification (RFID) technology. Indeed, this experimentation demonstrated the opportunity to provide a believable food localization service using a low-cost scale system, instead of RFID and localization system, that requires food tag labeling. The food box tagging with RFID systems is a promising technique, for the simultaneous localization and identification of goods. Nevertheless, local regulations do not impose such technologies in the food/packaging industry and, in the majority of the world countries, only barcodes are used for the identification of food categories. The proposed localization system aims to empower smart appliances with advanced food management capabilities, independently from the presence of smart tags on food, with good positioning accuracy. If, in the near future, RFID tags are commonly used to encode and monitor food, this technology could be integrated into our system to replace or complement the barcode. On the other hand, Sandholm et al. [
7] proposed automatic item position detection using compartment-embedded IR proximity sensors. No low-cost solution based on sensors within the room is feasible at freezer operating temperatures. For this reason, we opted for the choice of using load cells under the chest’s feet. With this solution, we can simultaneously achieve the weighing of inserted products and their horizontal localization. The main purpose of our system is to identify the correct insertion quadrant for each product. Our test results, described in the previous section, show that for 0.5 kg, 1 kg, 1.5 kg and 4 kg, 100% accuracy of localization was achieved. Furthermore, Sandholm et al. and Murata et al. [
7,
8] equipped a refrigerator door with sensors to know when the door closes or opens, which they respectively used to detect wasteful usage of a fridge and to know when video frames should be captured and analyzed. We used a magnetic switch in combination with temperature sensors and a power supply detector to realize a high-level functionality able to warn the user of unintentional prolonged openings, monitor rising temperatures and, in the case of detected thawing, alert the user to not consume refrozen foods. Furthermore, the battery present on the tablet advises the user when the power source of the chest freezer is temporarily missing. Another innovative feature was implemented in our system, leveraging the combined use of door status magnetic switch and load cells to detect lid-opening events, which produce slight variation in the tare. The implemented functionality automatically resets the tare at each lid-opening event. The variation amount calculated after a drift test is used, in order to implement automatic tare reset. In addition, the weight control is also used to produce a warning if the user leans on the chest during an insertion procedure.
Load cells are subject to creep over time, as reported in the producer’s datasheet. This lead to a drift in the weight measure that could be significant for long-time installation and during the appliance life. In our system this problem is overcome, by using the load cells to perform differential measurements. Every time a new food is inserted, the system is recalibrated using weight data registered before and after the lid opening. This method allows to exploit measured food weight as a differential measure on the whole content weight, for the limited time period of the lid opening, filtering out the sensor drift due to the creep.
The proposed system can locate the food placed on the freezer floor, as well as the food stacked on or placed in the baskets. The positioning algorithm assesses the barycenter of the new inserted food, and is unbiased by the presence of other items above or under the new inserted item. Indeed HighChest (as the majority of commercial chest freezer), has the compressor engine mounted on the right side, under the freezer floor. The localization errors accounted for the compartments affected by the presence of the compressor were comparable to the errors accounted for the other compartments (see
Table 5), demonstrating that the presence of items or the level/elevation of the food do not affect the localization service accuracy.
Additionally, the size of baskets is appropriate to cover the majority of food products in the market, thus enabling us to speculate that the likelihood to place very large packages, out of the boundaries of the baskets, is very low. It is clear that for large packages, for which it is necessary to remove one or more baskets, the localization system stops to correctly work, or, at least, the system can work only for the remaining baskets. Indeed the freezer design was intended to improve the food management, including the minimization of the reordering actions. The use of inner baskets, dedicated to store specific types of food, was intended to facilitate the inspection, retrieve and placement of food, without the need to remove items from the freezer. It means that the current working conditions, with defined and optimized size of baskets, should drive users to not perform displacements and changes in the position of products.
Eventually, the localization system is continuously sensible to any variation of weight in a differential mode, i.e., it is able to measure the variation of weight in relation to the activities that are performed. These activities include not only the food placement and retrieve, but also possible temporary displacements to facilitate another product placement or when the user moves the baskets along the rails for inserting food into the baskets located at the bottom. Of course, we are aware that possible “unconventional” usage could occur, such as temporary or permanent displacement of the food already stored or large size food placement. In this case, further improvement of the system will be required. In this sense, the freezer reordering procedure could be improved by the introduction of new technologies, like RFID tags. Radio tags would allow the simultaneous identification and localization of the stored food. However the localization and identification performance of this technology is nowadays problematic, as said before, because of the electromagnetic reflections of metals like the inner aluminum coating in the freezers that should be further investigated, and of the lack of standardized use of RFID on food packages.
The connectivity capability of HighChest makes it particularly suitable to be considered as an Internet of Things (IoT) device, able to improve the usability and utility of smart-freezers by sharing knowledge of foods and allowing the remote control of the appliance. The Inventory creation service was designed to be implemented on each single appliance. The chest can generate new entries in the “prefer database” or retrieve information from it either if that database runs locally on the machine, or is provided as a cloud service or hosted on a web server. In the latter case, the HighChest freezer acts as an IoT connected device able to share fundamental information for the improvement of the appliance usability and utility. Indeed, the sharing of information about a barcode reading, the associated food typology and the weight, among all the connected smart freezers, will enable the collective automatic learning of food features. This will produce a significant time saving for the users and a progressive procedure simplification for the recognition of new food typologies. The status monitoring service could be also provided in the IoT framework, connecting each user with the associated freezer. The individual access of the HighChest IoT devices to the “preferred database” and the analysis of the triggered food-expiration, temperature, food-defrost or door-opening alarms would also provide useful information for the assessment of the eco-friendly behaviors of the connected users. This could contribute to the estimation of the annual food waste and the refinement of food-waste prevention or eco-friendly awareness campaigns.
With an IoT oriented deployment, data security becomes a crucial topic. Indeed, ensuring privacy and personal data protection in the IoT framework is fundamental [
25], since smart appliances multiply the points of entry and processing for personal/sensitive data and their automated decisions, based on personal data, could directly impact on the users. Privacy risks include the opportunity that a smart freezer could provide private information about the users eating habits, revealing in some cases food allergies or intolerances. To prevent this happening, the web server or cloud platform providing the proposed HighChest IoT service manager, would have a public-key certificate. The connection with each freezer would also take place in a secure channel, implementing for example a Transport Layer Security (TLS) or a Secure Sockets Layer (SSL) protocol. Another crucial issues is the protection of the sensitive personal information (SPI), that can be used by third-part software or unauthorized entities, to identify, contact, or locate the users, as well as associate the users with the smart freezer, its content, and the usage behavior. Traditionally, authentication systems used to access cloud services or web servers made use of unique identifiers, that relates the users and the IoT associated devices [
26]. Nevertheless, this method exposes information about the user and the devices, through the connection to the IoT service manager. More complex authentication methods, including anonymous credential systems like the Identity Mixer (Idemix) technology developed by IBM Research [
27], would better protect the user privacy concerning his/her feeding habits. For example, the user and the related HighChest could for first connect to a trusted security service provider, providing anonymous credentials to connect together or with the Internet or cloud services. This would prevent the exposing of the user identity, and only the treatability of the single connections.
8. Conclusions
With the increased adoption of freezers into the home environment, the presence of solutions for food management is essential for meal planning and waste reduction, even more so than in refrigerators. In this paper, we have presented HighChest, an augmented household freezer designed to optimize the storage of frozen items, promote eco-efficient behavior and reduce wastage. To the best of the authors’ knowledge, no one before has dealt with the enhancement of domestic freezer capabilities. The proposed solution allows easier management of stored food, control of food conservation and feedback to remind users of expiry dates, preventing wastage. The monitoring of the chest’s status is another important aspect, allowing the prevention of abnormal functioning or a prompt intervention to ward off thawing.
We implemented functionalities so far applied only in refrigerators, overcoming limitations of the proposed methods to reach a working and reliable prototype. As our proposed system is easily scalable, it allows the addition of more devices in series controlled together in a single and integrated inventory, suitable for large contexts, especially those where it is essential to optimize resources and reduce waste, such as in distribution chains and catering services, as well as hospitals, nurseries and canteens.
The hardware architecture of our prototype is depicted in detail. We have introduced innovative services and features, such as power supply fault detection, food expiration prevention, a 3-D food localization system and an advanced interface for food waste prevention. The food management and freezer status functions are made accessible via a tablet-based user-friendly interface. We compared the most representative smart refrigerator concepts to our proposed freezer, highlighting the vast number of implemented functionalities in our system.
The storing management of our system is barcode technology-based, so the application of a code is necessary for all products that do not come from a POS network, such as leftovers and homemade products. Currently, the system is thought to provide printable barcodes applicable to its containers. In future work, the system will be equipped with a barcode printer to extend the range of traceable products to leftovers and homemade products. The idea is to provide the user with a set of barcodes that univocally identify the various food classes. For example, a barcode will be used to distinguish all products belonging to the type ‘pork’, another will correspond to ‘pasta’, and so on. The new code will consist of eight characters, with the first four corresponding to the type (e.g., Meat) and the second four identifying the subclass of food (e.g., Chicken). In order to automate the first insertion/registration, the list of these foods will be presented in the “Prefer” database at system installation.
The rigorous design methodology, described in
Section 3, enabled the identification of a number of users’ requirements and relative technical specifications. Some of them were implemented and described in this work. However other recommendations are still under development. In future developments, we plan to improve the system’s energy efficiency by adapting the defrosting procedure depending on the humid air intake during the loading phase, and controlling the compressor according to internal temperature, and amount and type of products placed inside the freezer. Indeed, proper control of inner temperature using the compressor affects food quality. Perishable food (e.g., fish) requires a storage temperature that is as constant as possible. Thus, rapid and energy-expensive compressor dynamics will be enabled only in the presence of delicate products, whereas an energy-saving control strategy will be used otherwise. To improve the user experience and to comply with the IoT framework, HighChest could turn the recipe suggestion service into a personalized dietary suggestion service. This functionality should consider preferences, possible allergies, intolerances and/or diseases of the user to promote a healthy diet. To allow better and more conscious space management, a dedicated app on smartphones will be created to enable mobile access to storage information, which currently is made possible only via e-mail or SMS.
Finally, behavioral and persuasive models of interaction could be designed as a future activity, to promote eco-friendly users behaviors and a healthy dietary. The ability of technological systems to act as a persuasive agent and induce behaviors on humans has been investigated from a long time, in different domains such as marketing, military training and the health industry [
28]. Sometimes, technology is used to monitor or held the people attention, to improve the teaching/learning in the educational field [
29]. In particular, many commercial devices exist for virtual coaching [
30] and researchers investigated a number of solutions to assess user’s stress or induce positive emotions through biofeedback [
31], improve the quality of life, promote healthy life, prevent chronic disease [
32] or support seniors in daily life [
28]. Most of this works rely on established psychological theories [
33,
34] on how people may be, are being, and will be influenced through the information technology (IT) designs. HighChest could contribute in the future, as a freezer embodiment, integrated in the IoT framework, for the investigation of persuasive strategies to promote eco-friendly behaviors.