Implementation of Real-Time Pedestrian Navigation System using Foot-mounted IMU
Liqiang Zhang
2018-01-01
Abstract:Pedestrian navigation technologies are widely used in the situations such as security, emergency, navigation and health care, etc. In order to achieve reliable positioning results, pedestrian navigation system (PNS) is being implemented by using numerous methods, such as Inertial Measurement Units (IMUs) based [1], wireless-based [2] and vision-based [3] technologies, etc. Among these technologies, IMUs-based methods are mainly based on pedestrian dead reckoning (PDR) theory using commercial foot-mounted IMU. Besides, this method is superior to the others due to its low cost, high reliability, anti-interference, especially no external infrastructure needed. The PDR system can be divided into two categories [4] i.e. inertial navigation system (INS) and step-and-heading system (SHS). An INS is a system that tracks position by estimating the full 3D trajectory of the sensor at any given moment. An SHS is specific to pedestrians, estimating position by accruing {steplength, heading} vectors representing either steps or strides. However, the errors of the INS-based PDR system accumulate over time due to the nature of INS. Hence it is difficult for INSbased PDR system to achieve long-time localization. To solve this problem, Foxlin [5] has proposed the classical Kalman-based PDR framework named IEZ. Its key technology named zero velocity updating (ZUPT) is valid to minimize the velocity errors of the PDR system. Besides, Ilyas et.al [6] have proved that ZUPT can also minimize the roll and pitch errors simultaneously. But the heading error and sensors biases are the only key errors which are unobservable for IEZ. Hence the heading drift of the whole system is easy to happen over time. For centuries, mechanical compass has been successfully used for navigation. Thus, the electronic compass (EC) algorithm [7] is an effective way to bound the heading error. But magnetic disturbances including soft and hard disturbance [8, 9] are stochastic and changing indoors. Hence, the disturbances are difficult to be predicted. Meanwhile, they are harmful to the magnetic heading [10-12]. As a consequence, some researchers do not suggest using EC algorithm indoors. One of these researchers is Borenstein who has proposed heuristic drift reduction (HDR) algorithm [13] and its improved algorithm named heuristic drift elimination (HDE) [14]. These two algorithms only rely on the data from gyroscope and do not require the magnetometer. But compared with straight-line motion, the HDR algorithm degrades its performance while turning and curving. Meanwhile, the HDE algorithm only performs well when moving along the dominant direction of a building. These drawbacks limit their application. In SHS, pedestrian motion is divided into continuous one-step motion. The position can be calculated by accruing the vector {step-length, heading} of one-step one by one. Since the interval of one-step is short, its accumulated errors is smaller than those of INS-based PDR system. In our work, we implement the SHSbased PDR system using commercial off-the-shelf foot-mounted IMU. Our proposed stance phase detection method named dual ZUPT (D-ZUPT) [15] is utilized to detect the stance phase in order to separate each one-step from continuous walk motion. Since the pedestrian tends to walk straight to save his or her energy and many corridors are straight, HDR algorithm is utilized to correct the heading error during the straight-line motion. Meanwhile, we construct the indirect extended Kalman filter (EKF) to minimize the systematic errors using the error state vector {velocity, position, and attitude}. The scheme of our system is shown in Fig.1. Experiments including indoor normal walking as shown in Fig.2 have shown the total travelled distance (TTD) error is 7 ‰ (i.e. 1000 m travelled distance with 7 m positioning error). Fig.1 The scheme of our PDR system Fig.2 We utilize the IMU named MTW from Xsens Company to achieve the walk experiment. The TTD error of positioning is 7 ‰.