Signal Strength-Based Localization in Indoor Wireless Networks A.S. Krishnakumar Avaya Labs ask@avaya.com 5 April 2006 Copyright© 2003 2002 Avaya Inc. All rights reserved Avaya – Proprietary Use pursuant to Company instructions Outline • Introduction • Applications of Location Information • Location determination in Radio Networks • Problem definition for 802.11 Networks • Current research and examples • Limits of location determination • Conclusion ask/19-Mar-16 2 Introduction A wireless terminal is untethered and may be mobile. To deliver a variety of services, it may be desirable to know the location of a wireless terminal with some degree of precision. • Can we estimate the location – with enhancements to the end device? – Without enhancements? • What techniques are available? • How accurately can we make this determination? ask/19-Mar-16 3 Outline • Introduction • Applications of Location Information • Location determination in Radio Networks • Problem definition for 802.11 Networks • Current research and examples • Limits of location determination • Conclusion ask/19-Mar-16 4 Applications Wireless location estimation is an important enabling technology to provide value-added location-aware services: In the enterprise: – Using closest resource in the enterprise – Privileges based on security regions – Enhanced e911 services, etc. In Public spaces: – – – – Emergency services Map/route information Recreation/Entertainment information Many others ask/19-Mar-16 5 Outline • Introduction • Applications of Location Information • Location determination in Radio Networks • Problem definition for 802.11 Networks • Current research and examples • Limits of location determination • Conclusion ask/19-Mar-16 6 Location Determination in Radio Networks Location estimation in indoor environments can be based on different characteristics of radio signals • Signal strength (RSSI) (e.g., RADAR) • Angle of arrival - AOA • Time of arrival - TOA • Time difference of arrival – TDOA (e.g., Cricket) ask/19-Mar-16 7 B1 B1 B2 B2 1 2 M M 3 B3 B3 Angle of Arrival B1 B1 B2 1 B2 2 M (S1, S2, S3) M 1-2 2-3 3-1 Time of Arrival 3 B3 B3 TDOA ask/19-Mar-16 8 Received Signal Strength Is this new? It has all been done before! • Radar • LORAN • GPS • etc. So what is new ? A lot! ask/19-Mar-16 9 Outline • Introduction • Applications of Location Information • Location determination in Radio Networks • Problem definition for 802.11 Networks • Current research and examples • Limits of location determination • Conclusion ask/19-Mar-16 10 Location in 802.11 Networks Desirable Characteristics: – Use existing hardware without enhancements – Ideally no client assistance – Simple to deploy and use – Adequate accuracy Complicating Factors: – Multi-path indoor propagation environment – Heterogeneous terminals – Site Engineering ask/19-Mar-16 11 Location in 802.11 Networks Attention has been focused on RSSI-based techniques since they: • Can be implemented with currently available hardware • Are reasonably accurate ask/19-Mar-16 12 Issues in Wireless Location Estimation • Location accuracy • Deployment and cost of ownership • Management • • • • Security considerations Zero-profiling techniques Deployment for coverage vs. location estimation Techniques for model adaptation ask/19-Mar-16 13 RSSI-based Techniques Client-based approach: – Client measures the signal strength from “visible” Access Points and this information is used to locate the client Infrastructure-based approach: – Deploy wireless sniffers that monitor client activity and measure signal strength information – No client changes required – Sniffers can also be used for other monitoring and security applications ask/19-Mar-16 14 Grouping of RSSI-based Techniques Client-based Infrastructure-based † Deterministic RADAR (Bahl et al.) [4] Prasithsangaree et al. [31] Pandey et al. [30] LEASE [22] Probabilistic Youssef et al. [40] HORUS [41] Bayesian Nets [24] Abnizova et al. [1] Bayesian Nets [24] Nibble [9] † The reference numbers here correspond to the bibliography in A.S. Krishnakumar et al., CollaborateCom 2005. ask/19-Mar-16 15 Outline • Introduction • Applications of Location Information • Location determination in Radio Networks • Problem definition for 802.11 Networks • Current research and examples • Limits of location determination • Conclusion ask/19-Mar-16 16 RSSI-based Techniques • Profiling-Based (Collected data is the model) • Needs a lot of data collection to build the model – Take signal strength measures at many points in the site and do a closest match to these points in signal strength vector space. [e.g., RADAR; INFOCOM 2000] – Build a prior probability distribution at many chosen points and use posterior distributions to determine best estimate of location [e.g., Robotics; IROS 2003] • Use physical characteristics of signal strength propagation and build a model augmented with a wall attenuation factor • Needs detailed (wall) map of the building; model portability needs to be determined – [e.g., RADAR; INFOCOM 2000] based on [Rappaport 1992] • Adaptation – Environmental and other changes require model rebuilding ask/19-Mar-16 17 Steps in Profiling-based Techniques • Data Collection – Collect signal strength measurements from all the APs at many points in the area of interest • Model Generation – Generate a model; could be the parameters of a propagation model or a signal-strength vector map or something else – Given a signal strength measurement, estimate the location based on: • Euclidean distance in signal space • Maximum likelihood estimate • Some other measure ask/19-Mar-16 18 On-line • Location Determination A Deterministic Technique - RADAR Based on Profiling: • Data Collection – Collect many measurements at each location on the grid • Model building: – The same as the collected data • On-line Estimation – Select the location that is the nearest neighbor in signalstrength space to the measured signal strength vector • Reported median error ~2.9m Further details may be found in Bahl et al., Infocom 2000 ask/19-Mar-16 19 A Deterministic Technique - RADAR Based on Propagation Model: • Data Collection – Collect many measurements at different distances with and without line of sight • Model building: – Estimate propagation model parameters and wall attenuation – Use the model to generate a signal strength map • On-line Estimation – Select the location that is the nearest neighbor in signalstrength space to the measured signal strength vector • Reported median error ~4.3m Further details may be found in Bahl et al., Infocom 2000 ask/19-Mar-16 20 LEASE – Location Estimation Assisted by Stationary Emitters • Automatic adaptation to changes • “Profiling” handled automatically by using SEs • There is a mapping between client- and infrastructure-based deployments (and LEASE) – Interpretation for Client-based deployment – Sniffers co-located with APs – Points where you profile signal strength from APs = points where you place SEs • Signal Strength model for a sniffer needs to be built using measured signal strengths from SEs – Model using minimal number of SEs (“profiled” points) • Our approach to build signal strength model: – Treat the problem as a data modeling problem • Median error ~5m (Further details in Krishnan et al. Infocom 2004) ask/19-Mar-16 21 Components of the LEASE system • Uses sniffers, stationary emitters (SEs) and a location estimation engine (LEE) • SEs – Cheap, battery operated devices at known locations – Transmit a few packets periodically • Sniffers – Record signal strength from the SEs and clients – Feed this information to the LEE AP: SE: Sniffer: LEE: • LEE – (Re-)models the “radio map” for a sniffer in response to signal strength readings of SEs at sniffers – Uses models to locate clients. ask/19-Mar-16 22 A Profiling-based Probabilistic Technique • Data Collection – Collect many measurements at each location on the grid • Model building: – Histogram of signal strengths at each location (i.e. joint probability distributions) • On-line Estimation – Select the location that maximizes the probability P(location|measured signal vector) • Reported median error ~1m Further details of this technique may be found in Youssef et al., PerCom 2003 ask/19-Mar-16 23 A Probabilistic Technique without Profiling • Based on hierarchical Bayesian networks • Simultaneously estimate the location of a number of terminals • The signal strength model is a hyperparameter of the Bayesian model • Assume reasonable prior distributions and compute the posterior density given the measurements • Use the computed posterior density to estimate the quantities of interest • Currently uses Markov Chain Monte Carlo techniques • Median error ~5m ask/19-Mar-16 24 Bayesian Networks Hierarchical Non-hierarchical ask/19-Mar-16 25 Outline • Introduction • Applications of Location Information • Location determination in Radio Networks • Problem definition for 802.11 Networks • Current research and examples • Limits of location determination • Conclusion ask/19-Mar-16 26 Median Error in Estimation Method Median Error in Estimation RADAR - Profiling ~3m RADAR - Propagation ~4.3m LEASE ~5m Probabilistic - Profiling ~1m Probabilistic – No Profiling ~5m Elnahrawy et al. observed a localization error of 10 ft (median) and 30 ft 97 (percentile) over a range of algorithms, approaches and environments (SECON 2004) ask/19-Mar-16 27 Estimation Accuracy The median error values are widely variable. This raises the following questions: • Why are they different? • How do we compare these values? • Is some kind of normalization possible? If so, how? • Are there fundamental limits to location accuracy with this technique? • What is the dependency on factors such as distance between APs? A preliminary analytical attempt to address these questions appeared in A.S. Krishnakumar and P. Krishnan, Infocom 2005. ask/19-Mar-16 28 Theoretical Analysis of Accuracy Physical Space Probability mass S3 T -1 S0 S2 Y (x0,y0) Location Uncertainty S1 Signal Space x ask/19-Mar-16 29 Estimation Accuracy The analysis shows that the minimum value of location uncertainty depends upon: • Desired probability α • Signal variance • Propagation constant • Number of APs • Distance between APs ask/19-Mar-16 30 Outline • Introduction • Applications of Location Information • Location determination in Radio Networks • Problem definition for 802.11 Networks • Current research and examples • Limits of location determination • Conclusion ask/19-Mar-16 31 Conclusion • Indoor location determination presents challenges due multipath propagation and other factors • We can still estimate location accurately enough for many applications • Site engineering affects location accuracy • The same technique has been applied to Bluetooth networks with comparable results • About the only factor affecting location uncertainty that is in control of the algorithm designer appears to be the signal variance ask/19-Mar-16 32 Open issues and research topics • Security considerations • Zero-profiling techniques • Deployment for coverage vs. location estimation • Techniques for model adaptation ask/19-Mar-16 33 Bibliography - 1 • [Bahl Infocom 2000] P. Bahl, V.N.Padmanabhan, “RADAR: An In-Building RFbased User Location and Tracking System,” Proceedings of IEEE Infocom 2000, Tel Aviv, Israel, March 2000. • [Youssef PerCom 2003] Moustafa Youssef, Ashok Agrawala, A. Udaya Shankar, “WLAN Location Determination via Clustering and Probability Distributions,” IEEE International Conference on Pervasive Computing and Communications (PerCom) 2003, Fort Worth, Texas, March 23-26, 2003. • [Krishnan Infocom 2004] P. Krishnan, A. S. Krishnakumar, Wen-Hua Ju, Colin Mallows, Sachin Ganu, “A System for LEASE: Location Estimation Assisted by Stationary Emitters for Indoor RF Wireless Networks,” Proceedings of IEEE Infocom 2004, Hong Kong. • [Krishnakumar Infocom 2005] A.S. Krishnakumar and P. Krishnan, “On the Accuracy of Signal Strength-based Location Estimation Techniques,” to appear in the Proceeding of IEEE Infocom 2005, Miami, Florida, March 2005. • [Madigan Infocom 2005] David Madigan, Eiman Elnahrawy, Richard P. Martin, Wen-Hua Ju, P.Krishnan, and A.S. Krishnakumar, “Bayesian Indoor Positioning Systems,” to appear in the Proceedings of IEEE Infocom 2005, Miami, Florida, March 2005. ask/19-Mar-16 34 Bibliography - 2 • [Robotics] Andrew M. Ladd, Kostas E. Bekris, Algis Rudys, Lydia E. Kavraki, Dan S. Wallach, and Guillaume Marceau, “Robotics-based location sensing using wireless ethernet,” In Proceedings of the eighth Annual International Conference on Mobile Computing and Networking (MOBICOM-02), pages 227– 238, New York, September 23–28 2002. ACM Press. • [Rappaport] T. S. Rapport, “Wireless Communications – Principles and Practice,” IEEE Press, 1996. • P. Bahl, V.N. Padmanabhan, and A. Balachandran, “Enhancements to the RADAR user location and tracking system,” Technical report, Microsoft Research Technical Report, February 2000. • Prasithsangaree, P. Krishnamurthy, and P.K. Chrysanthis, “On indoor position location with wireless LANs,” In The 13th IEEE International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC 2002), 2002. • N.B. Priyantha, A. Chakraborty, and H. Balakrishnan, “The cricket location support system,” In Proceedings of the Sixth Annual ACM International Conference on Mobile Computing and Networking, 2000, pages 51–56, 2003. • M. Youssef and A.K. Agrawala, “Handling samples correlation in the HORUS system,” In IEEE Infocom, 2004. ask/19-Mar-16 35 Bibliography - 3 • [Krishnakumar CollaborateCom 2005] A.S. Krishnakumar and P. Krishnan, “The Theory and Practice of Signal Strength-Based Location Estimation,” The first international conference on collaborative computing, San Jose, California, December 2005. • T. Roos, P. Myllymaki, and H. Tirri, “A statistical modeling approach to location estimation,” IEEE Transactions on Mobile Computing, 1:59–69, 2002. • S. Saha, K. Chaudhuri, D. Sanghi, and P. Bhagwat, “Location determination of a mobile device using IEEE 802.11 access point signals,” In IEEE Wireless Communications and Networking Conference (WCNC), 2003. • A. Smailagic, D.P. Siewiorek, J. Anhalt, D. Kogan, and Y. Wang, “Location sensing and privacy in a context aware computing environment,” Pervasive Computing 2001, 2001. ask/19-Mar-16 36