Harmful chemical compounds are released daily in warehouses, chemical plants and during environmental emergencies. Their uncontrolled dispersion contributes to the pollution of the atmosphere and threatens human and animal lives. When gas leaks occur, their timely localization is crucial to ensure a swift and effective emergency response. Moreover, the efficient monitoring of at-risk facilities can serve as an informed baseline to devise policies to reduce dangerous emissions in the long term. However, there are several risks associated with tracking and mapping toxic gasses, concerning first and foremost the safety of the people involved in the task. Autonomous robots equipped with gas sensors could make gas sensing missions safer and more efficient, carry a variety of sensing equipment and explore areas that are hard or dangerous to reach. Aerial vehicles have rarely been employed in this field because they usually move using propellers, which greatly impact the gas distribution and hinder the sensing abilities of the system. However, the employment of rotary-wing aerial vehicles offers a promising research direction because their versatility of movement can capture the tridimensional nature of the gas dispersion phenomenon.

In this dissertation, we investigate ways in which rotary-wing nano aerial vehicles can be employed in gas sensing missions, and we couple the adoption of these platforms to novel algorithmic and system design approaches to enhance the gas sensing outcome. We initially validate our bespoke gas sensing setup with a bio-inspired gas source localization algorithm, and we assess the impact of sensor placement, environmental factors and localization accuracy on sensing performance. Afterwards, we explore the application of adaptive path planning techniques to gas distribution mapping, a field where preplanned trajectories are more commonly employed for navigation. We then introduce GaSLAM, a novel algorithmic framework that aims to tackle gas source localization and gas distribution mapping simultaneously by exploiting the synergies between two algorithmic classes that were, so far, mainly treated separately. Finally, motivated by the need of speeding up urgent missions, we explore the application of a multi-robot system composed of rotary-wing aerial vehicles to the field of gas sensing. We start by qualitatively and quantitatively analyzing the impact that two vehicles with propellers have on each other's sensing performance. We then investigate the impact that different levels of information sharing among agents, movement coordination and prior environmental knowledge have on the sensing outcome. Finally, we explore the effect of different path planning approaches on multi-robot gas sensing missions that present communication constraints and where little to no information are shared among agents. Throughout the thesis we opt for an experimental approach, ensuring that all solutions that we propose are tested with repeated experiments in our experimental testbed using real robots.