Using Smart Virtual-Sensor Nodes to Improve the Robustness of Indoor Localization Systems

Area of Science:

• Biomedical Engineering
• Computer Science
• Signal Processing

Background:

• E-health systems require reliable monitoring for vulnerable populations (elderly, disabled).
• Wireless Sensor Networks (WSN) are crucial for indoor localization in healthcare.
• Sensor unavailability in WSN can compromise patient monitoring systems.

Purpose of the Study:

• To investigate the feasibility of using virtual sensors to replace unavailable nodes in WSN for indoor localization.
• To model and artificially reproduce the path of a monitored target when sensor data is missing.
• To assess the effectiveness of this approach, particularly for critical areas like entrances and exits.

Main Methods:

• Utilized a neural network trained on historical position data from a WSN.
• Focused on a scenario with a single unavailable sensor node.
• Trained the network with 4 minutes of non-linear movement data to simulate a 2-minute sensor outage.

Main Results:

• The neural network successfully reproduced the monitored target's path with reasonable accuracy.
• The virtual sensor approach showed particular success in critical areas like room entrances and exits.
• This method offers a viable solution for maintaining monitoring continuity despite sensor failures.

Conclusions:

• Virtual sensors generated by neural networks can effectively compensate for unavailable sensors in WSN for indoor localization.
• This technology enhances the reliability of e-health monitoring systems for patient safety.
• The approach is promising for ensuring continuous security monitoring in healthcare facilities.