Development of an internet of things (IoT) Based four chamber smart fridge for proper storage of different pharmaceutical products based on their labeled storage conditions

Abstract

The general goal of this thesis is to present the findings of research done on an efficient way of managing temperature-sensitive medical products with the help of the internet of things (IoT) and machine learning (ML). This study was conducted considering the case of medical pharmacies in Rwanda. This is a thesis by articles that contains three articles. Each article presents a specific contribution to the whole research work. Generally, after identifying that there is poor management of temperature, a four-chamber fridge that is based on IoT was proposed. The development of this fridge passed through three main phases: The first phase was dealing with the design and the development of the whole enclosure. The second was about electronic circuit design and development. During the same phase, we have identified the best way for data transmission specifically for the case of Rwanda. During the last phase, data have been analyzed and some mathematical models were developed. In the first article, details about the development of the fridge that has four chambers have been presented. In this paper, the fridge enclosure and the electronic control circuit have been designed and implemented. In most IoT applications, data are sent to remotes database for being processed and analyzed this has been found to be associated with different challenges such as connectivity issues, data security, and data latency. The main purpose of this work was to prove that a compressed machine learning model can be developed and embedded in low resources microcontrollers. With this, we had an idea of controlling the fridge temperature by monitoring the frequency of opening and closing the door while picking some items and finally predicting what will be happing in the near future. This will make the fridge intelligent without sending data to the database. We have finally achieved our objectives and an Arduino library was developed. The result from our experiments shows that the model runs onto the controller and can predict well the internal fridge temperature at an accuracy of 96%. During the second article, we had an idea of taking our fridge as a sensor node that can communicate by sending some data to the cloud. Then our task was to find the best communication protocol that is used by considering the Rwandan situation. For any communication protocol, there is a need for a well-structured communication infrastructure for building a network of sensor nodes. This infrastructure is mainly composed of gateways, repeaters, and base transceiver stations (BTS). Considering that Rwanda is a country with a lot of hills that will not allow a line of sight communication and considering that the Global System for Mobile Communications (GSM) network covers 96% of the country, we decided to choose General Packet Radio Services (GPRS) as communication protocol. Therefore, we experimentally developed a model that can predict a life span of a GPRS-based sensor node with reference to the received signal strength indicator (RSSI) at a particular location. We acquired current consumption for the sensor node in different locations with their corresponding received signal quality and we tried to experimentally find a mathematical data-driven model for estimating the GSM/GPRS sensor node battery lifetime using the received signal strength indicator vi (RSSI). The results from the experiment showed that this model can be used to predict GPRS sensor node life span, replacement intervals, and dynamic handover which will, in turn, provide uninterrupted data service. Through the analysis, it has been even found that when there is a reduction of −30 dBm in RSSI, the current consumption of the radio unit of the node will double. In the third article, one method of analyzing data that has been sent to the cloud during the second article’s work was presented. This was a multivariate regression model which can be used to monitor the internal temperature of a pharmaceutical fridge specifically when a pharmacist is selling medical products by predicting what will be the temperature in a given period after the fridge got opened. It is clear that when a given room of a fridge is opened, the internal temperature increase. In this research, a multi-room fridge was proposed. The developed fridge had a screen that keep displaying the current internal temperature of every room and the time required for the temperature of a particular room to go beyond the acceptable range in case it is opened. We proposed that, before a pharmacist opens a fridge he/she will first check the room for a longer time. The developed model was evaluated using the coefficient of determination R2 and was found to be 77%. It has to be noted that, this study was not interested in the impact of temperature fluctuation on pharmaceutical products. The presented articles have different methodologies and this has been indicated in every article.