Questions and answers after the presentation

Track 4

Answers by Jian Kuang and Tao Liu, GNSS Research Center, Wuhan University

Q1: Does your algorithm correctly handle "walking on escalator" ?

A1: We have shown our basic tricks in the report: the gravity constraint and the forward (body frame) constant speed constraint algorithms, i.e. assume constant velocity of the escalator motion. Before we use these algorithms, we have carried out very detailed processing in the following aspects to enhance the foot-INS performance.

First, because the IMU noise is different between dynamic and static conditions, we fine-tune the sensor parameters based on the Allan variance. The adjustment method is to meet the optimal result of zero-speed correction under long-term static data.

Second, we used an adaptive threshold method to adjust the stance phase detection results, and at the same time, modified some unreasonable detection results manually.

Third, we model the error divergence of Foot-INS. This model can help to improve the accuracy of Foot-INS significantly.

Fourth, the model between the motion speed of the escalator and the accelerometer variance was established. We carried out a sufficient detection of the time period of the escalator scenario and correctly estimated the motion speed of the escalator by this model.

We also found some re-visits of the trajectory and made use of such valuable opportunity to correct the drift of foot-INS through close loop adjustment (smoothing like SLAM).

In summary, combined with manual intervention to make the solution consist and smooth, we can handle the scene of the escalator very reliably.

Q2: Did you use map data to constrain the path ?

A2: We made it clear that no map information was used in our solution. In fact, we don't think map-matching can constrain the positioning error effectively in such shopping mall environment.


Finally, we are surprised that our solution can achieve such high positioning accuracy. It might have very good luck. If more different test data sets are used to evaluate the positioning accuracy and performance of Foot-PDR, the probability of this randomness will become relatively small.