Indoor navigation with smartphones

Transcription

1 Indoor navigation with smartphones REinEU2016 Conference September PAVEL DAVIDSON

2 Outline Indoor navigation system for smartphone: goals and requirements WiFi based positioning Application of BLE for positioning Map-aided navigation Using smartphone s self-contained sensors for positioning

3 Navigation with smartphones: goals and requirements Social, Retail, Safety Meet up with friends at the mall, sport events, conventions, concerts, etc., Keep track of kids & elder parents Automatically tag posts & pictures from favorite locations Accuracy requirements Room level Correct floor identification

4 Available hardware and possible approaches to navigation GNSS (A-GNSS) receiver Use of radio frequency (RF) signals, either those already present such as wireless local area networks (WLAN, cellular) or those generated by new and dedicated infrastructure (NFC, Bluetooth). Self-contained sensors: tri-axis accelerometer, tri-axis magnetometer, tri-axis gyroscope, barometer Camera. In some cases two cameras for stereo vision. Indoor map (building floor plan), radio map Network assistance, collaborative positioning Powerful processor, large memory, high quality display

5 Major components of a smartphone indoor positioning system

6 Part I WI-FI AND BLE BASED POSITIONING

7 Using WiFi for indoor positioning PROS: Existing infrastructure and end-user devices The public buildings (shopping malls, offices, agencies) as well as homes are filled with WLAN access points. Modern mobile phones and tablets carry an integrated WLAN radio. CONS: WiFi was not designed for positioning Only Received Signal Strength (RSS) can be used to compute range. RSS measurements are noisy and affected by devices s heterogeinity and orientation, signal attenuation by people. Radio map creation is a time-consuming and labor-intensive task. Using crowd-sourcing is difficult since reference positioning data usually is not available indoors.

8 Approaches for RSS based positioning

9 Comparing Methods for different positioning models based for WLAN on radio signals signal s RSS Fingerprinting Strong Pathloss models All Coverage area models A Linear State Model for PDR+WLAN Positioning

10 Radio map The area of interest is divided into cells with the help of a floor plan. RSSI values of the radio signals transmitted by APs are collected at the calibration points inside the cells for a certain period of time and stored into the radio map. Data storage requirements for radio maps for tested methods in two test buildings. There were 1292 FPs in building 1 and 3445 FPs in building 2 Reprinted from : Philipp Mu ller, Matti Raitoharju, and Robert Piche, "A Field Test of Parametric WLAN-Fingerprint-Positioning Methods, in Proc. 17th Int Conf on Information Fusion, Salamanca, Spain, 7-10 July

11 Location fingerprinting approach based on RSS measurements

12 k-nearest neighbour (knn) fingerprinting algorithm The user location estimate is chosen to be the calibration point that is closest in the RSS space x j ri j i 1 is the measured RSS from the i-th AP i x j xˆ arg min x is the i-th AP s RSS fingerprint measured at j-th calibration point Weighted KNN is a generalization with different importance weights. Larger weights are usually given to the fingerprints that are closer in RSS space. The advantages of NN algorithms are simplicity, robustness and reasonably good accuracy n i i r x j

13 Field tests: : positioning was performed in less than one month after the radio map had been created Reprinted from: Philipp Mu ller, Matti Raitoharju, and Robert Piche, "A Field Test of Parametric WLAN-Fingerprint- Positioning Methods, in Proc. 17th Int Conf on Information Fusion, Salamanca, Spain, 7-10 July

14 Field tests: positioning was performed several months after the radio map had been created Reprinted from : Philipp Mu ller, Matti Raitoharju, and Robert Piche, "A Field Test of Parametric WLAN-Fingerprint- Positioning Methods, in Proc. 17th Int Conf on Information Fusion, Salamanca, Spain, 7-10 July

15 Typical accuracy Reprinted from : Philipp Mu ller, Matti Raitoharju, and Robert Piche, "A Field Test of Parametric WLAN-Fingerprint-Positioning Methods, in Proc. 17th Int Conf on Information Fusion, Salamanca, Spain, 7-10 July 2014

16 Application of BLE for positioning BLE is superior for positioning than WiFi for the following reasons: channel hopping mechanism that makes the RSS less noisy much higher scan rate than WIFi freedom of beacon placement provides good signal geometry short handshake procedure lower transmission power BLE can provide very accurate (submeter-level) positioning in proximity mode when the transmitting power is set to low levels. In close proximity to a beacon the distance estimation based on RSS is very accurate because the signal power decreases as the lognormal attenuation model The most popular BLE beacon ecosystems are Apple s ibeacon, Google s URIBeacon and Eddystone, and Radius Networks AltBeacon

17 Part II MAP AIDED NAVIGATION

18 How map can improve positioning? In the case of pedestrian indoor navigation building floor plans represent the constraints that restrict movements. For example, people cannot walk through walls. The goal of map-aided navigation is to exploit prior information contained in maps or building plans. Question: how do we, in an automated way, do what our eyes/brains can do so well?

19 Approaches to map-matching indoors

20 Proposed solution is based Proposed solution is based on particle filtering on particle filtering 7! The sequential importance sampling (SIS) algorithm is a recursive Monte-Carlo technique to approximate the posterior pdf x by a cloud of N weighted particles. ( ) p 0: k y1 : k The probability density function is approximated by a set of samples (also referred to as particles) and its associated weights. Road map information can be incorporated into vehicle position estimation through the through constraints the on constraints the state vector on the state vector. Indoor map information can be incorporated in user position estimation /25/11

21 Application of wall constraints Algorithm: 1. Particle propagation based on PDR 2. Floor plan weighting z } { 3. WLAN weighting 4. Resampling wall Nurminen, IPIN 2013, slide 6 / 12

22 Map matching indoors: wall constraints

23 Map matching indoors: topological approach The building is represented by a node-link model It is assumed that people are located and traveling on the links It is difficult to model large open spaces Reprinted from: Gilliéron, Pierre-Yves, et al. "Indoor navigation performance analysis." ENC GNSS

24 Example of indoor navigation using WiFi and link-node model of building

25 Link-node model vs wall constraints Wall constraints Building floor plans are readily available Implementation is relatively simple The resulted position can cross walls In the case of WiFi only position measurements propagation of particles may be not accurate. Node-link model Node-link model of building has to be obtained Implementation is complex Topology is better followed Not well suited for large open spaces Numerically more stable. Small amount of particles is required Work well even when only WiFi measurements are available

26 Benefits of WiFi + map positioning Improved accuracy compare to WiFi only solution, especially the cross-track error Follows building topology and eliminates cases when estimated path crosses walls Produces smoother estimated user path by removing noise associated with WiFi positioning

27 Part III USING SMARTPHONE SENSORS FOR POSITIONING

28 Challenges in using smartphone sensors for positioning Inertial and other self-contained sensors can improve positioning on a smartphone. However, the accuracy of these sensors in a typical smartphone is low. The major difficulty for unconstrained phone is a heading offset between the direction a smartphone is facing and the direction the user is moving. There are approaches that attempt to solve this problem, however it is still an open research problem. Different transit modes such as pedestrian, elevators and escalators make the sensor based positioning even more complicated. The approaches also have to work on different types of surface and across wide group of people regardless of their height, weight, age, gender, physical conditions, etc. Smartphone sensors can be useful for positioning only if an approach can cope with changing orientation of a phone without any constraint on usage scenarios which typically include carrying phone in a pocket, bag, hand or arm-band, making a call, texting or watching display

29 Major tasks that can be performed using smartphone sensors

30 Dead reckoning Calculating current position by using a previously determined position and advancing that position based upon known or estimated displacement and heading Position and heading error accumulation The distance can be computed based on speed sensor or accelerometers The direction can be computed using gyro or magnetometer

31 Walk detection and step counting Most walk detection and step counting methods are based on assumption that pedestrian motion has a cyclic nature. The algorithms search for the repeating patterns in accelerometer or gyroscope data during walking. They can be generally categorized into the following groups: thresholding and peak detection, auto-correlation, and spectral analysis. These algorithms have to work on unconstraint smartphones by implementing the following methods: (a) allowing arbitrary pose and carrying location, (b) tracking the phone orientation, and (c) performing motion classification and adapting different algorithms for different motions

32 Step Length Estimation Simple algorithms only count steps assuming that the step length is just the average for that user. Advanced algorithms also perform accurate step segmentation and analyze the accelerometer signals to estimate the magnitude of each step individually. However, most of these systems require calibration to an individual user because everyone's gait has different acceleration profiles

33 Walking direction estimation Walking direction estimation on unconstrained smartphone is the most difficult problem that has to be solved to achieve capabilities for autonomous navigation. This is still an open problem since most of the proposed solutions can work only with IMU that have much better accuracy than modern smartphone sensors. The algorithms for walking direction estimation usually include two important tasks: (a) computation of the smartphone's pose with respect to some global frame; (b) estimation of walking direction in global frame using smartphone's sensors

34 Multiple usage scenarios and transit modes Identification of transit mode, type of activity and phone placement can significantly improve the accuracy and reliability of PDR and map-matching. For the purpose of pedestrian navigation the motion classification can be simplified by limiting the number of classes to only five: All situations when a user is static, i.e. the location does not change significantly during some time with the allowed amount of movement not exceeding the predefined threshold All cases when a user is walking except when a phone is in hand swinging. This class includes the following cases: (a) A phone is in pocket, bag, arm-band or use any other type of attachment, (b) Texting or watching screen, (c) Making or receiving a call, (d) Walking with the mobile device in a handheld bag. All cases when a phone is in hand swinging. The user is walking while holding the mobile device in his/her swinging hand. Moving on elevators or escalators. Irregular motions that the user performs while standing or sitting. During this time the user's location does not change

35 Summary of sensor based positioning Although in contemporary smartphones the sensor based positioning cannot operate as a stand-alone technology it can be very useful when combined with WLAN/BLE based positioning in a fusion algorithm Also sensors can make more accurate and stable map-aided navigation solution The major advantage of this technology is availability of accelerometers, gyroscopes and magnetometers in almost every smartphone. The major difficulty in implementation of these tasks on smartphones is freedom of movement and arbitrary placement. The major tasks that can be performed using smartphone sensors are step count, distance and walking direction computation, activity recognition, vertical movements and floor change detection. The most difficult task is a smartphone orientation tracking and walking direction estimation

36 Conclusions The most accurate navigation solution is based on a combination of WLAN, BLE, map, magnetic field and sensor based positioning. It seems likely that the future smartphone indoor positioning system will be a hybrid system with a variety of positioning modalities to augment or substitute for WLAN based positioning. Absolute position fixes will be obtained opportunistically from WLAN, Bluetooth, LTE, NFC or other future technologies. These measurements will be fed in some fusion filter which can combine them with map, magnetic field fingerprints and sensor measurements and produce accurate and reliable position estimates

37 QUESTIONS?