IoT based Smart Solid Waste Bin Collection Monitoring System Using Adaptive Neuro-fuzzy Inference System (ANFIS)

Abstract

Solid waste collection is the one major parts of solid waste management. Various problems are occurring due to improper collection of solid wastes. These include unhygienic condition, air pollution, and unhealthy environment creating diseases to the general public. To solve these problems various researches were carried out on solid waste collection using different methods and algorithms. Among these some methods are shortest path method, RFID, GIS, Ant colony system, vehicle routing. These researches use time windows, scheduling, neighborhood search, detecting fullness of bin by proximity sensor. But, unhygienic condition, air pollution and unhealthy environment occurred by solid wastes is not considered by these prior researches. This research tried to solve the problem of environmental pollution by using IOT and ANFIS on different features. The main objective is to develop system which can monitor solid waste bins collection thereby improving the efficiency of waste collection management systems. In order to achieve the main objective ANFIS and IOT methods are used in this research. ANFIS is the method that contains decision making and train data which are the ability of fuzzy logic and neural network respectively. IOT is used to detect real signal around the bins using electronic sensors. Sensors used are MQ-4, MQ-35, MQ-36 and Ultrasonic to detect CH4 gas, CO2 gas, H2S gas and level of fullness of bins respectively. The detected signals are processed by Arduino uno which is used as central processor. These data is transferred to central server by using GSM module which is used as network device. The ANFIS model is created to identify the severed bins by using the data come from sensors. This model is developed using 805(eight hundred six) training data set collected by real time sensors. Back-propagation and least squares estimation algorithms are used to develop this model. The efficiency of the model in this research experiment is measured by RMSE. The research experiments are done using four function types of ANFIS method. These are Gbellmf, Guessmf, trimf and dsigmf. These functions are tested in different epochs to generate efficient model. The result of the experiment is 0.12 root mean square error with 0.118 testing error. This research demonstrated that the model developed performs the prediction task with very high accuracy (~90%). Hence, it accurately notifies the concerning body when the bins are in a condition to pollute the environment.