AeroPurify: Autonomous air filtration UAV using real-time 3-D Monte Carlo gradient search

Outdoor air pollution is the fourth-leading cause of global mortality, claiming 4.2 million lives annually and impacting 99% of the world. Accessible and efficient solutions for pollution mitigation are lacking, despite the widespread impacts of outdoor air pollution. We present an autonomous drone air filtration system prototype, designed to detect and mitigate outdoor air pollution through utilizing a novel autonomous navigation algorithm and a custom-built data processing and transmission system (DPTS). The DPTS detects and transmits real-time particulate matter data to the navigation algorithm and has a Fibonacci Spiral-based filtration attachment that allows for air purification. Currently, one method for path finding is called gradient ascent (GA); however, this algorithm, when simulated, was time-consuming, visiting extraneous waypoints. In this paper, we propose an alternative to the GA algorithm called the gradient ascent ML particle filter (GA/MLPF) algorithm, which assists the drone in its autonomous traversal of the pollution field to find the source of the pollution. Based on the Bayesian state estimation particle filter, we hypothesized that the GA/MLPF algorithm would outperform the conventional GA algorithm by creating a time-efficient path and reducing the number of waypoints. Our results showed that the GA/MLPF algorithm did outperform the conventional GA algorithm: it took an average of 70% less time and reduced the number of waypoints by at least 28%. The GA/MLPF algorithm developed in this project is an innovative approach to tackling outdoor air pollution, and the algorithm’s mobility and effectiveness allow it to be used in many diverse environments.