Contents I. Introduction
Global navigation satellite systems are frequently used as position tracking tools. However, there are environments, for example, buildings, mountain valleys, and caves, where satellite signals cannot be readily obtained. Therefore, a personal position tracking system in such environments is highly desirable, especially for firefighting, rescue mission, and military operation. Emerging position aware calculation such as indoor mixed reality and augmented reality also requires a reliable position tracking technology. Position tracking can be achieved by using optical sensors, radio frequency (RF) beacons, acoustic transceivers, and inertial measurement units (IMUs) [1]–[3]. IMU-based position calculation is the only form of position tracking that does not require an external reference. Therefore, it is highly preferred for practical applications. Position can be calculated based on three-axis acceleration, three-axis angular rotation rate, and three-axis magnetic field strength. Such information can be obtained from a nine degree-of-freedom (DOF) IMU. Microelectromechanical systems (MEMS)-based IMUs exhibit low power, low cost, and small form factor. These features are ideal for implementing personal inertial navigation systems (PINSs). However, MEMS-based IMUs typically suffer from low accuracy and large offset and drift, which can cause a drastic position error over a short period of time. To address these challenges, human bipedal locomotion characteristics have been explored to circumvent the aforementioned drawbacks. Fig. 1(a) depicts different gait phases associated with the human walk, consisting of the striking phase, mid-stance phase, and detaching phase. For a practical application, it can be assumed that the displacement is close to zero between the striking phase and the detaching phase. Thus, the position calculation within this time period can be ceased. In addition, it has been shown that the foot heel velocity is close to zero within a short period of the time window during the mid-stance phase [4]. This stationary period can be used to reset the velocity calculation, known as the zero velocity update (ZUPT). Traditional methods employ the accelerometer or gyroscope vector output signal to measure the foot-on-ground (FoG) timing [5], [6], which is used to control the operations described earlier. However, the FoG timing determined by such methods exhibits a limited accuracy caused by IMU output ringing, foot shaking during the mid-stance phase, and other artifacts caused by foot movement during the striking and detaching phases. These effects adversely hinder the navigation performance. It has been demonstrated that the FoG timing can also be detected by using a passive ground reaction sensor array (GRSA) without suffering from the motion artifacts associated with the IMUs [7], [8]. Commercially available GRSAs typically exhibit a limited resolution and array size [7]. These characteristics limit the accuracy of the FoG timing detection, hence, adversely affecting the navigation accuracy. In order to achieve an improved navigation performance, a custom-designed high-density GRSA interfaced with a high-resolution sensor interface application-specified integrated circuit (ASIC) is proposed to obtain an accurate FoG timing. Fig. 1(b) presents the proposed PINS architecture, where a nine DOF IMU and a custom-designed high-density GRSA are embedded inside the heel region of a boot. The GRSA can provide a detailed foot-to-ground reaction pressure profile, as shown in Fig. 1(c). From the pressure profile, the foot striking moment and detaching moment can be identified as the moments when the GRSA output signals increase above and fall below a certain threshold level, respectively, as shown in Fig. 1(d). Obtaining an accurate FoG timing is critical for a substantial navigation performance improvement. The high-density GRSA can also serve as a research vehicle to explore the tradeoffs among the navigation performance, array size, and system sensing resolution. Furthermore, any residue velocity measured at the middle of the FoG time window can be used to estimate an effective acceleration offset during the previous foot swing phase. The estimated offset can be used to correct the position error accumulated in each step, thus further improving position accuracy. The real-time data from the IMU and GRSA under a walking condition are acquired and processed upon a calibration algorithm to produce an accurate navigation performance. It should be noted that the reference frame aligned with the three-axis accelerometer, three-axis gyroscope, and three-axis magnetometer enclosed in an IMU is typically referred to as the body frame. The IMU raw data output is thus expressed within the body frame. However, position tracking requires IMU outputs expressed within the navigation frame, where the three axes are aligned along the north, the east, and toward the earth center. The orientation of the IMU body frame with respect to the navigation frame is required to convert IMU output data from the body frame to the navigation frame [9], [10]. Such an orientation can be estimated by integrating the three-axis gyroscope output signals over time. However, MEMS gyroscopes’ output signals usually exhibit a large offset and drift, which can lead to an excessive orientation error over time. A three-axis magnetometer is, therefore, used to serve as an additional reference to compensate the gyroscopes inherent orientation error. Our prototype PINS employs a commercial nine DOF IMU (InertiaCube 4 from InterSense, LLC), which provides an internal data processing algorithm to estimate the orientation information in real time based on the gyroscope and magnetometer output signals, and then converts the accelerometer data from the body frame to the navigation frame. As a result, the prototype navigation system performance is based on processing the three-axis accelerometer data expressed in the navigation frame. Section II describes the GRSA design, fabrication, and characterization. Section III presents the sensor interface ASIC design and characterization results. The assembled PINS performance and system design tradeoffs are demonstrated in Section IV with the conclusion provided in Section V.
PINS assisted by GRSA. (a) Human walking gait phases. (b) PINS employing IMU and GRSA inside a boot. (c) Foot-to-ground reaction pressure profiles. (d) Detected FoG timing.
