Divergence-based odor source declaration

Abstract

This paper explores the use of the divergence operator for odor source declaration in swarm-based algorithms. A set of simulations of a swarm of robots running the decentralized asynchronous particle swarm optimization, bacterial foraging optimization and ant colony optimization algorithms was used to generate multiple wind and odor biased vector fields to investigate the effectiveness of the divergence operator in odor source declaration. A set of real world experiments were also performed using the same swarm algorithms on a controlled environment to ascertain if the divergence operator can also be used on real data. The sparse gas sensor data acquired by the robots was interpolated using the Nadaraya-Watson estimator by means of a wind and odor biased kernel before the application of the divergence. Results show that the divergence operator excels at odor source declaration.