Walking Trajectory Estimation Using Multi-Sensor Fusion and a Probabilistic Step Model.

Area of Science:

• Robotics
• Biomechanical Engineering
• Computer Science

Background:

• Accurate human trajectory estimation is crucial for applications like biomechanics and human-robot interaction.
• Existing methods, such as particle filters, can be computationally intensive for real-time processing.

Purpose of the Study:

• To develop a computationally inexpensive and accurate framework for estimating a walking human's trajectory.
• To improve the robustness and efficiency of human localization systems.

Main Methods:

• A non-Gaussian recursive Bayesian estimator fusing global (GPS) and inertial measurements.
• Integration of a kinematically driven step model and a maximum a posteriori-type filter.
• Utilizing zero-velocity update (ZUPT) from inertial measurement units and gradient ascent optimization.

Main Results:

• The proposed framework demonstrated comparable accuracy to state-of-the-art particle filters in simulations.
• The novel estimator proved more computationally efficient, especially at higher resolutions.
• The method offers advantages for high-dimensional state estimation problems.

Conclusions:

• The developed framework provides an efficient and accurate solution for real-time human trajectory estimation.
• This approach has potential applications in biomechanics, human safety, and human-robot teaming.
• The estimator is a viable alternative to traditional particle filters for various real-time estimation tasks.