When gas leaks occur – for example in warehouses, environmental emergencies, or search and rescue operations – it is very important to be able to locate the source of the leak in order to take fast and effective countermeasures. This is usually done using animals, which is very costly and puts human and animal lives in danger.

Using robots to detect gas sources in 3D

Chiara Ercolani is a Ph.D. student at the Distributed Intelligent Systems and Algorithms Laboratory (DISAL) at École Polytechnique Fédérale de Lausanne (EPFL) in Switzerland. Chiara and her team believe that using robots to detect gas leaks is the best way to save lives.

One of the team’s most recent projects involved 3D gas source localization using a micro aerial vehicle (drone). 

Wind tunnels, gas plumes and drones

The experiment was conducted in a wind tunnel facility (16x4x2) m3 with laminarized wind flow of adjustable speed. An electric pump was used to disperse a mixture of ethanol and air in the wind tunnel to create the gas plume.

Gas dispersion is a 3D phenomenon, so to be effective in 3D gas source localization, the team needed to use platforms capable of 3D motion tracking.

The team experimented with sensor location, performance under various environmental conditions, and two localization strategies. The full results can be viewed in this short video summary.

For the purposes of this article, we will focus on the third aspect of the experiment, namely determining what effect each of the following localization methods would have on the outcome of the gas source localization:

  1. Ultra-Wide Band (UWB) localization system with 8 beacons
  2. Motion Capture System (MCS) with 13 cameras

Motion Capture System localization outperforms Ultra Wide Band localization

While UWB localization is easier to deploy and cheaper, this experiment makes it clear that using a motion capture system offers better performance under all tested environmental conditions. The drone’s movements are clearly much cleaner and smoother than the UWB system. As a result, the drone is faster and more efficient. 

These results show that MCS localization is more accurate than UWB localization. We also see that the drone’s movement is much more efficient when MCS is used.

Motion Analysis is the mocap partner of choice

The DISAL lab has used motion capture technology and software from Motion Analysis for many years, so when the time came to purchase a second motion capture system, the choice was clear. The laboratory now has 13 Kestrel 1300 cameras and the team uses Cortex to process performance data. 

To learn more about this research, please visit https://www.epfl.ch/labs/disal/research/3dodorsensing/

To learn more about the motion capture system and the powerful Cortex software, please visit our solution overview page.