Machine Learning Based Adaptive Routing and Predictive Approach to Urban Garbage Management

Abstract

In this paper, a waste management system is introduced to replace the conventional waste collection process in an attempt to optimize the usage of resources. The problem of overflowing garbage cans is solved by constantly monitoring the garbage levels inside the can and deploying garbage collecting vehicles as soon as the cans are about to fill completely. The vehicles follow a pre-determined, optimized route obtained using the algorithm proposed. Another problem of garbage being spilled around the can, usually due to human carelessness, is solved by installing a device to open the lid when an object above the lid  is within a specified range, and close it automatically after a specified time. This also allows people to use the garbage can without touching it directly. Route Selection is done with predicting garbage level at the instant and a variant of Travelling Salesman. Machine learning is used to predict at what time during the day a particular garbage can is expected to be full. Expanding this prediction over all the garbage cans, the number of vehicles that need to be deployed at different times of the day can be controlled.