Finding Your Way Back: Comparing Path Odometry Algorithms for Assisted Return.

Area of Science:

• Robotics
• Human-Computer Interaction
• Assistive Technology

Background:

• Assisted return systems aid navigation by enabling path backtracking.
• These systems are particularly beneficial for blind pedestrians.
• Existing inertial odometry methods require evaluation for assisted return accuracy.

Purpose of the Study:

• To comparatively analyze inertial odometry algorithms for assisted return.
• To introduce and evaluate a novel path matching algorithm.
• To assess algorithm performance using real-world data from blind walkers.

Main Methods:

• Comparative analysis of two inertial odometry systems: deep learning (RoNIN) and turn/step counting.
• Implementation of a new path matching algorithm.
• Testing in simulated assisted return tasks using the WeAllWalk dataset.

Main Results:

• The turn/step counting odometry system achieved superior path matching performance.
• The novel path matching algorithm was tested within simulated environments.
• Performance was evaluated using inertial data from blind individuals.

Conclusions:

• Robust turn and step counting odometry is more effective for assisted return path matching than deep learning approaches.
• The developed path matching algorithm shows promise for navigation aids.
• Further research can refine these methods for practical assistive navigation solutions.