Summary for week 4
Pedestrian Tracking with Shoe-Mounted Inertial Sensors
InterSense developed NavShoe, which can track the user wearing a shoe on which a small inertial sensor is adapted, on the contrary to previous devices that required a specific environment. Thus, the location and the orientation of the user can be known.
The usual problem is the horizontal acceleration which is a cubic error and leads to inaccurate results. Since walking is made out of two phases : 0.5 second of stationary stance phase and 0.5 second of moving stride phase. During the first phase, the speed is zero hence there is a zero-velocity update (ZUPT) in the Extended Kalman Filter (EKF). Thus, a correction occurred after each stride leading to a linear error instead of the cubic one. The position and the orientation are also corrected. Still there is an error for the horizontal position. The solution is made out of three improvements : use a better gyro, use a magnetic compass (which will also correct another problem due to metal in the shoe) and use a GPS.
The hardware used is the InertiaCube3 with its battery on the shoe. RF is used instead of a cable, which was a big disadvantage. The aim is to have a PDA on the user instead of using a laptop. Algorithms are used for inertial navigation and zero-velocity updating, magnetometer calibration, geomagnetic modeling and heading drift correction.
Two environments have been used for testing have been used, indoor and outdoor. Indoor, the user walks 118.5 meters in 322 seconds in a house. The error is 0.3% but the accuracy is almost perfect for the height. Outdoor, the user walks 741meters, the error is also 0.3%.
For what concerns heading transfer to a head tracker or a handheld device, the issue is that it induces a magnetic deviation since the true north is not used. When the device is on the shoe, the two phases of stillness and motion allow a decorrelation of compass measurements.
Some improvements could be made, like using a GPS or optical sensors.
The main future work will be to verify that the error is the same for all trajectories and to provide a better accuracy for head-tracking sensors.
Thus, this seems to provide an efficient and light way for location and orientation. However, the final aim nowadays would certainly be to implement this technology on smartphones since almost everybody has one, thus no additional device (or almost) would be needed.
Comments are closed.