Do You Smell Something?

Gas leaks in industrial settings can be highly dangerous for both workers in the area, as well as rescue personnel attempting to remedy the situation. When a leak is suspected, the job of tracking down the source falls on firefighters, who spend considerable time, at great risk, using handheld sensing instruments. The task is complicated by the fact that these instruments are not nearly as capable as animal noses, and also by environments that add complexity to the spread of the gas.

Recognizing that there must be a better way to approach these problems, a team from the Delft University of Technology developed a swarm of drones capable of identifying the source of a gas leak. Sending many drones simultaneously allows the source to be located more quickly, and without risking human lives. Moreover, the small size of the drones allows them to explore tight spaces that a person could not reach. This exploration is further enhanced by the drones flight capability, which enables them to explore in three dimensions.

These benefits do not come easy, however. These tiny devices, with minimal computing power, need to be able to navigate an unknown environment and avoid collisions with one another, in addition to detecting gases. In search of a solution to these problems, the researchers found inspiration in the humble fruit fly. With tiny brains, containing only 100,000 neurons, these insects are capable of flight, navigation, and detecting the odors given off by fruit.

This insight led the team to develop their “Sniffy Bug” algorithm for navigation. Initially, so long as none of the drones have sensed gas, they spread out as much as possible within the environment being scanned. When a drone detects gas, it communicates that fact to all of the other drones, which then begin executing a particle swarm optimization algorithm. This allows the swarm to quickly zero in on the source of the leak by leveraging its own gas concentration measurement, as well as the measurements from other drones. This method has been proven to be effective in real-world, cluttered environments.

The drone itself is a Bitcraze CrazyFlie nano quadcopter. It comes equipped with IMU sensors and four laser range sensors that are used for navigation, localization, and obstacle avoidance. A custom PCB with a Figaro TGS8100 MEMS MOX gas sensor was added to detect gas concentrations. Communication between drones was achieved with a Decawave DWM1000 ultra-wideband module.

This initial work is very promising, but the team has some challenges yet to tackle. They have not yet explored moving in three dimensions to locate gas leaks at heights, for example. With some refinement, however, these drones may find applications even beyond emergency scenarios. Perhaps one day they will seek out methane, or other gases, for scientific operations on Mars.