物聯網(Internet of Things)中一項重要的基本能力是物件追蹤，
意即可以提供某特定物件在任意時間點的位置，
而此定位追蹤能力更是情景感知技術的關鍵基礎。
目前雖然已有許多基於影像處理或無線訊號處理的定位演算法被提出了，
但前提是這些演算法都必須先建置基礎設施。
此外，其定位能力也僅限於這些基礎設施的感測範圍內。
然而，此假設並不符合實際的運用，
這些需要被追蹤的物件的活動範圍(如穿戴式裝置等)
在使用上不一定會被侷限于在一個固定區域內。
為了克服這項缺點，
我們提出了以相遇資訊經由粒子過濾進行的合作式追蹤演算法。
藉由電子地圖，在不仰賴基礎設施的情況下，
具備無線通訊能力與慣性感測器的電子裝置(如手機等)
即可以依此演算法協同地互相追蹤，
並運算出各個裝置的路徑軌跡。
此合作式追蹤演算法以粒子過濾追蹤法為基礎，
並將多個粒子的相遇資訊納入考量對其改進。
本篇論文的主要貢獻是我們成功地利用時空的交集資訊
，快速地將可能錯誤的粒子消除。
與原本的粒子過濾追蹤法相比，在相同的輸入資料下，
實驗結果顯示我們的演算法有顯著的進步。

One of the fundamental features of the Internet of Things (IoT) is the ability to track its users location over time as a way to provide service in the right context. While many localization and tracking techniques have been proposed using cameras and RF-based techniques,they are ultimately limited by the ability to deploy tracking infrastructure and cannot operate when outside coverage area.
To overcome this limitation, we propose encounter-based cooperative tracking for proximity-enabled wireless nodes. These nodes, carried by the subjects being tracked, roam inside a mapped area and can sense their trajectories (e.g., by inertial sensing) and each other's proximity but without access to localizing infrastructure such as beacons.
We model their locations over time on the map as particles
and apply particle filtering as the basis for tracking. Our
contribution is to augment the base technique with encounter
information to dramatically reduce the number of possible particles in the model by spatial-temporal intersection. Experimental results show our techniques to be effective even in maps that contain layouts that have many possible matches for given trajectories.

Abstract
Contents
Acknowledgments
1 Introduction
1.1 Motivation
1.2 Contributions

2 Related Works
2.1 Pedestrian Dead Reckoning
2.2 Encounter based Tracking

3 Background Theory
3.1 Review of Probability
3.1.1 ExpectedValue
3.1.2 Variance
3.1.3 Indicator Function and Dirac Function
3.1.4 Conditional Probability
3.1.5 Independence
3.1.6 Conditional Independence
3.1.7 Bayesian Rule
3.2 Bayesian Network
3.2.1 Bayesian Model
3.2.2 Inference from Bayesian Networks
3.2.3 D-Separation and D-Connection
3.3 Hidden Markov Model
3.3.1 Markov Model(FirstOrder)
3.3.2 Hidden Markov Model
3.3.3 Recursive Property of HMM
3.4 Monte Carlo Method
3.4.1 Law of Large Numbers
3.4.2 Monte Carlo Estimate
3.4.3 Monte Carlo Method
3.5 Importance Sampling
3.5.1 Basic Idea : Unnormalized/Unbiased Importance Sampling
3.5.2 Deeper Look into Importance Sampling
3.5.3 (Self-)Normalized/Weight Importance Sampling
3.5.4 Importance Sampling Diagnostics : Effective Sample Size
3.6 Sequential Importance Sampling
3.6.1 Computational Complexity of Importance Sampling
3.6.2 Choose One Recursive Proposal Distribution
3.6.3 Put It All Together
3.6.4 Algorithm Description
3.7 Sequential Importance Resampling
3.7.1 Degeneracy Problem
3.7.2 Selection of Proposal Distribution
3.7.3 Resampling
3.7.4 Generic Sequential Importance Resampling
3.8 State Estimation Problem
3.8.1 Bayesian Filter
3.8.2 Particle Filter

4 Tracking by Particle Filtering
4.1 Tracking Problem
4.1.2 Cooperative Tracking by Encounter Dependence
4.1.3 EncounterDependenceChain
4.2 Approximation in Real World
4.2.1 Description of Dynamic System
4.2.2 Tracking by Particle Filter
4.2.3 Cooperative Tracking by Encounter Dependence

5 Evaluations
5.1 ExperimentDesign
5.1.1 Metrics
5.1.2 Location Estimation Method
5.1.3 Experimental Setup
5.1.4 Experiments Design
5.2 Results
5.2.1 Results Overview
5.2.2 Experiment Results
5.2.3 Results Summary

6 Conclusions and Future Works
6.1 Conclusions
6.2 Future works

References

[1] Apple ibeacon. https://developer.apple.com/ibeacon/.

[2] Course note of computational statistics and data analysis.
http://www.wikicoursenote.com/
wiki/Stat341.

[3] D-separation. http://www.bayesnets.com/D-separation.html.

[4] Law of large numbers. http://en.wikipedia.org/wiki/Law_of_large_numbers.

[5] Michigan chemical engineering process dynamics and controls open electronic textbook. https://controls.engin.umich.edu/wiki/index.php/Bayesian_network_theory.

[6] Monte carlo method. http://en.wikipedia.org/wiki/Monte_Carlo_method.

[7] ARULAMPALAM, M. S., MASKELL, S., AND GORDON, N. A tutorial on particle filters for online nonlinear/non-gaussian bayesian tracking. IEEE TRANSACTIONS ON SIGNAL PRO- CESSING 50 (2002), 174–188.

[8] CONSTANDACHE, I., BAO, X., CHOUDHURY, R. R., AND AZIZYAN, M. Did you see bob?: human localization using mobile phones. In In Proceedings of ACM International Conference on Mobile Computing and Networking (MobiCom ’10) (2010), ACM, ACM, pp. 149–160.

[9] DAVIDSON, P., COLLIN, J., AND TAKALA, J. Application of particle filters for indoor po- sitioning using floor plans. In Ubiquitous Positioning Indoor Navigation and Location Based Service (UPINLBS), 2010 (Oct 2010), IEEE, IEEE, pp. 1–4.

[10] DOUCET, A., DE FREITAS, N., AND GORDON, N. An introduction to sequential monte carlo methods. In Sequential Monte Carlo Methods in Practice. Springer-Verlag, 2001, ch. 1, pp. 3 – 14.
124

[11] DOUCET, A., GODSILL, S., AND ANDRIEU, C. On sequential monte carlo sampling methods for bayesian filtering. STATISTICS AND COMPUTING 10, 3 (2000), 197–208.

[12] DOUCET, A., AND JOHANSEN, A. M. A tutorial on particle filtering and smoothing: fifteen years later, 2011.

[13] HUE, C., P. LE CADRE, J., AND PREZ, P. Tracking multiple objects with particle filtering, 2000.

[14] KONG, A., LIU, J. S., AND WONG, W. H. Sequential imputations and bayesian missing data problems. American Statistical Association 89, 425 (March 1994), 278–288.

[15] LI, F., ZHAO, C., DING, G., GONG, J., LIU, C., AND ZHAO, F. A reliable and accurate indoor localization method using phone inertial sensors. In In Proceedings of the 2012 ACM Conference on Ubiquitous Computing (UbiComp ’12) (2012), ACM, ACM, pp. 421–430.

[16] OKUMA, K., TALEGHANI, A., FREITAS, N. D., FREITAS, O. D., LITTLE, J. J., AND LOWE, D. G. A boosted particle filter: Multitarget detection and tracking. In In the European Confer- ence on Computer Vision(ECCV 2004) (2004), pp. 28–39.

[17] ORLANDE, H. R. B., COLACO, M. J., DULIKRAVICH, G. S., VIANNA, F. L. V., DA SILVA, W. B., DA FONSECA, H. M., AND FUDYM, O. Tutorial 10:kalman and particle filters. http://www.sft.asso.fr/Local/sft/dir/user-3775/documents/actes/Metti5_ School/Lectures&Tutorials-Texts/Text-T10-Orlande.pdf.

[18] OWEN, A. B. Monte Carlo theory, methods and examples. 2013.

[19] RAI, A., CHINTALAPUDI, K. K., PADMANABHAN, V. N., AND SEN, R. Zee: Zero-effort crowdsourcing for indoor localization. In In Proceedings of ACM International Conference on Mobile Computing and Networking (MobiCom ’12) (2012), ACM, ACM, pp. 293–304.

[20] SCHEINES, R. D-separation tutorial. http://www.andrew.cmu.edu/user/scheines/tutor/ d-sep.html.

[21] SYMINGTON, A., AND TRIGONI, N. Encounter based sensor tracking. In In Proceedings of the thirteenth ACM international symposium on Mobile Ad Hoc Networking and Computing (MobiHoc ’12) (2012), ACM, ACM, pp. 15–24.
125

[22] WIDYAWAN, KLEPAL, M., AND BEAUREGARD, S. A backtracking particle filter for fusing building plans with pdr displacement estimates. In Positioning, Navigation and Communication, 2008. WPNC 2008. 5th Workshop on (March 2008), IEEE, IEEE, pp. 207–212.