A Semantic-Based Gas Source Localization with a Mobile Robot Combining Vision and Chemical Sensing

Area of Science:

• Robotics
• Sensor Fusion
• Environmental Monitoring

Background:

• Gas leak localization is critical for safety and environmental protection.
• Existing methods often struggle with complex indoor environments and sensor uncertainties.

Purpose of the Study:

• To develop an efficient mobile robot system for localizing gas emission sources in human environments.
• To leverage semantic relationships between detected gases and recognized objects for improved localization accuracy.

Main Methods:

• A sensor fusion system integrating vision and chemical sensing.
• A probabilistic Bayesian framework to manage uncertainties in object and gas recognition.
• Markov decision processes for path planning to minimize search time.

Main Results:

• The system dynamically updates potential gas source candidates based on sensor data and object recognition.
• Demonstrated efficiency in an office-like scenario through simulated and real-world experiments.
• Successfully accounted for uncertainties in object recognition and gas classification.

Conclusions:

• The proposed approach effectively localizes gas emission sources by integrating semantic information and probabilistic reasoning.
• This method offers a robust solution for gas leak detection in structured indoor environments.
• The system's efficiency was validated, showing promise for real-world applications.