Week4 Summary

The author starts the paper with stating the needs for tracking with Shoe-Mounted sensors. The main issue with traditional tracking framework is that it required the system to be instrumented to result in a tracking with reasonable error. To effectively track without prepared instruments, the author came up with an inertial tracking system called NavShoe. NavShoe is small, low-power, and fairly accurate up to meters in outdoor environment. NavShoe system not only tracks location of foot but also tracks orientation of foot.

Basically, tracking with only inertial sensor is nearly impossible due to the error accumulation that grows exponentially in time. NavShoe used two different stages of people walking to help reducing the error growth to linear in time with Extended Kalman Filter. The EKF gets feedbacks from Zero Velocity Update which is applied in standing stage of walking and corrects the tracking result retrospectively. However using ZUPT alone had problem with horizontal positioning. Therefore, the author made additional changes to the system such as using better gyro for yaw axis and incorporate information from other sensors such as compass and GPS. Using better gyro for yaw axis reduces the drift to reasonable level. Using compass to incorporate with gyro also reduces drift to reasonable level without the cost for better gyro. Using GPS information corrects drift in yaw through transfer alignment effect.

Next, the author explains the system used for proof-of-concept experiment. The hardware is composed of GPS, the InterSense InertiaCube3, and RF communication device. Then, the author explains the algorithm implemented in the experiment in more detail. The EKF is used in every step of algorithm to reduce error accumulation. The algorithm takes 4 different types of information at a rate of 300 times per second and incorporates the information to enhance the accuracy. If any of inertial sensing differs from the average since beginning of the current still stage above certain threshold, the algorithm uses ZUPT.

Then, the author shows test results for both indoor and outdoor environment. The results looked great and shoe-mounted tracking device could be another essential part of AR system along with head-mounted displays. Like stated in the beginning of the paper, this shoe-mounted tracking device would perform much better with incorporating visual information from HMD.

Comments are closed.