Multi-Sensor, Multi- Network Positioning

Ruizhi Chen, Heidi Kuusniemi, Yuwei Chen, Ling Pei, Wei Chen, Jingbin Liu, Helena Leppäkoski, Jarmo Takala

Currently, no single technology, system, or sensor can provide a positioning solution any time, anywhere. The key is to utilize multiple technologies. We are now exploring a multi-sensor multi-network (MSMN) approach for a seamless indoor-out-door solution. Its hardware platform is described in the previous article. h e digital signal processor (DSP) is embedded in the GPS module. All sensors are integrated to the DSP that hosts core software for real-time sensor data acquisition and real-time processing to estimate user location. A smartphone handset pro-vides wireless network measurements.

Positioning Algorithms

The multi-sensor positioning platform enables a positioning solution with a combination of GPS and reduced inertial navigation system (INS), or GPS and pedestrian dead reckoning (PDR). The reduced INS consists of a 3D accelerometer and a 2D digital compass, as a low-cost alternative to augment GNSS positioning. The reduced INS combined with GPS uses a loosely coupled Kalman filter for data integration, while the combination of PDR and GPS uses algorithms for estimating the position change with pedestrian step-length estimation.

PDR. h e PDR solution uses human physiological characteristics, implemented in a local-level frame, with equations:

where k denotes the current epoch, Y is the coordinate in East direction, X is the coordinate in North direction, S is step length, and is the heading.

h e PDR positioning algorithm includes step detection, step length estimation, determination of heading, and positioning.

To achieve an accurate heading, compass measurements are corrected with an empirical online estimated error model, which requires some training data.

WLAN and Bluetooth. FIGURE 1 describes the basic concept of the WLAN or Bluetooth locating solution using a fi ngerprint database approach. h e circles around the access point (AP) in the fi gure represent the radio coverage area and the color the signal strength. h is radio map is a simplifi ed example representing measurements from just one AP.

For the fi ngerprinting approach, the received signal strength indicators (RSSIs) are the basic observables. h e whole process consists of a training phase and a positioning phase. During the training phase, a radio map of probability distribution of the received signal strength is constructed for the targeted area. h e targeted area is divided into a matrix of grids, and the central point of each grid is referred to as a reference point. h e probability distribution of the received signal strength at each reference point is represented by a Weibull function, and the parameters of the Weibull function are estimated with the limited number of training observation samples. Based on the constructed radio map, the positioning phase determines the current location using the measured RSSI observations in real time.

Given the observation vector , the problem is to fi nd the most probable location (l ) with the maximized conditional probability, maximized by Bayesian theorem as:

We applied an assumption of Hidden Markov Models (HMM) to represent the pedestrian movement process. h e  locating problem is then translated into fi nding such a state sequence (locations) that is most likely to have generated the output sequence (the measured RSSIs) assuming the given HMM model. h e Viterbi algorithm typically solves these kinds of problems effi ciently. h is study also utilizes the Viterbi algorithm to trace the user trajectory.

Å FIGURE 2 Integration scheme for multi-sensor, multi-network positioning approach

MSMN. h e general integration scheme combining the GPS output, sensor measurements, WLAN, or Bluetooth output, and their variance estimates is depicted in FIGURE 2. A simplifi ed representation of the central fi lter combining diff erent input sources can be described with typical Kalman fi lter equations. h e measurement model is

where the state estimate vector is

with X, Y, and as previously defi ned, and S the user horizontal velocity (speed). h e measurement vector is given as

where g refers to GPS, W to WLAN/Bluetooth, acc to accelerometer, and dc to digital compass. h e matrix Hk is the design matrix of the system and the vector vk is the measurement error vector.

The recursive sequence includes prediction and update steps. h e prediction step includes the typical equations of

and

while the update step includes

Indoor Test Results

A fi eld test has been carried out on a sports fi eld, described in the accompanying article (see page 14). An indoor test was carried out in an offi ce-building corridor, but the test started and ended in an outdoor terrace area. During the test, the indoor corridor was covered with eight WLAN and three BT APs.

FIGURE 3 shows the positioning results of the GPS-only (red), Bluetooth-only (black), and WLAN-only (magenta) solutions; FIGURE 4 shows that of the integrated multi-sensor multi-network (MSMN) solution (blue) for an outdoor-in-door-outdoor test. A reference trajectory is in green in both fi gures and building outlines in grey. h e position update rate achievable by the WLAN and Bluetooth fi ngerprinting approach is only 0.1 Hz whereas the GPS-only and the inte-grated MSMN solutions are obtained every second and thus have a higher availability.

Å FIGURE 3 Pedestrian test results with GPS-only, BT-only, and WLAN-only positioning approaches with respect to a reference trajectory

Å FIGURE 4 Pedestrian test result with the multi-sensor multi-network positioning approach with respect to a reference trajectory

FIGURE 5 shows the horizontal errors obtained with the different positioning solutions over time in the indoor test. A mean horizontal error of 2.2 meters was achieved with the WLAN solution. h e Bluetooth solution is not as accurate as the WLAN solution, due to the smaller amount of BT APs; it achieved a mean horizontal error of 5.1 meters. When moving inside the corridor, the GPS solutions are used for the MSMN integration only with very low weights due to their poor quality. GPS is mainly used as a source of location outdoors where the test starts and ends. h e mean horizontal error of the GPS-only solutions during the whole test is 8.4 meters. WLAN- and Bluetooth-derived locations and the self-contained sensors are the main sources used inside the building for the MSMN positioning solution: the mean horizontal accuracy obtained with MSMN is 2.7 meters with a solution availability of 1 Hz.

The MSMN solution obviously performs much better than a GPS-only solution indoors. h e track of the pedestrian walking inside the corridor can be identifi ed clearly, which is not the case with typical approaches of GPS-only or GPS/low-cost sensors. WLAN fi ngerprinting provides good position accuracy indoors, but the MSMN solution provides the best result when taking into account positioning accuracy and the solution availabilities in both time and space domains.

Conclusions

Further development is needed for indoor areas to be able to obtain fully seamless outdoor-to-indoor location, though GPS initialization followed by sensor and WLAN/BT combination already provide very good initial results. Additional sensors and more refi ned pedestrian-specifi c algorithms will be added to further improve the positioning accuracy.

Å FIGURE 5 Horizontal errors of GPS-only, BT-only, WLAN-only and the MSMN positioning approaches with respect to time in the pedestrian indoor test