Chemical source localization in real environments integrating chemical concentrations in a probabilistic plume mapping approach

• Chemical plume source localization algorithms can be classified either as reactive plume tracking or gas distribution mapping approaches. Here, we focus on gas distribution mapping methods where the robot does not need to track the plume to find the source and can be used for other tasks. Probabilistic mapping approaches have been previously applied to real-world data successfully; e.g., in the approach proposed by Pang and Farrell. Instead of the quasi-continuous gas measurement values, this algorithm considers events (detections and non-detections) based on whether the sensor response is above or below a threshold to update recursively a source probability grid map; thus, discarding important information. We developed an extension of this event-based approach, integrating chemical concentrations directly instead of binary information. In this work, both algorithms are compared using real-world data obtained from a photo-ionization detector (PID), a non-selective gas sensor, and an anemometer in real environments. We validate simulation results and demonstrate that the concentration-based approach is more accurate in terms of a higher probability at the ground truth source location, a smaller distance between the probability maximum and the source location, and a more peaked probability distribution, measured in terms of the overall entropy.

Subject headings

TEKNIK OCH TEKNOLOGIER  -- Elektroteknik och elektronik -- Robotteknik och automation (hsv//swe)

ENGINEERING AND TECHNOLOGY  -- Electrical Engineering, Electronic Engineering, Information Engineering -- Robotics (hsv//eng)

NATURVETENSKAP  -- Data- och informationsvetenskap -- Datavetenskap (hsv//swe)

NATURAL SCIENCES  -- Computer and Information Sciences -- Computer Sciences (hsv//eng)

Keyword

chemical plume source localization

Bayesian inference

chemical concentration

mobile robots

real environmen ts

Computer Science

Datavetenskap

Publication and Content Type

ref (subject category)

kon (subject category)