Download presentation
Presentation is loading. Please wait.
Published byWendy Franklin Modified over 9 years ago
1
On Distinguishing the Multiple Radio Paths in RSS-based Ranging Dian Zhang, Yunhuai Liu, Xiaonan Guo, Min Gao and Lionel M. Ni College of Software, Shenzhen University Department of Computer Science and Engineering, Hong Kong University of Science and Technology, Third Research Institute of Ministry of Public Security 20121029 TY
2
Outline Introduction Observation Problem definition Treatments of problem Experiment Evaluation Conclusion
3
Introduction RSS-based ranging is apt to be affected by the multipath phenomenon Previous works try to profile the environment and refer this profile in run-time – Site survey In practical dynamic environments, however, the profile frequently changes and the painful retraining is needed
4
Introduction Key observation – Given a pair of nodes, the RSS at different spectrum channels will be different By analyzing these RSS values, we are able to identify the amplitude of signals solely from the Line-of-Sight path
5
Introduction Non-Linear Curvature Fitting Problem – An ill-conditioned problem – Practical considerations to improve the conditioning shape MuD (Multipath Distinguishing) System – Real-Time indoor tracking system – Very different from us
6
Observation Multipath – Atmospheric duct – Reflection – Refraction Friss Model cannot work Profile-based approaches – Labor-Intensive – Cannot adapt to dynamic environment
7
Observation
8
Frequency diversity may help – significantly different RSS values at different spectrum channels – quite stable at the same environment By carefully analyzing these RSS, we can identify the amplitudes and phases of signals from each path, then derive the accurate distance according to the amplitude of LOS signals Observation
10
Radio propagation in free space (LOS path) – Friss Model – P t is the transmission power – G t, G r are the antenna gain of the transmitter and receiver – λ is the signal wavelength – d is the LOS path length – Path strength – Path phase Multipath & Frequency Diversity
11
Reflection and Refraction (NLOS path) – Reflection/Refraction Coefficient – path strength – Path phase is the same – d is not the LOS distance anymore Multipath & Frequency Diversity
12
Path phase – Change when the signal frequency changes phase-shift – Change when the signal path changes – Path amplitude will be affected – TelosB Frequency : 2.4G ~ 2.4835G Number of Channels : 16 Wave length : 0.125m ~ 0.1208m Multipath & Frequency Diversity
13
a = 2m λ 1 = 0.125 m b = 3m λ 2 = 0.1208 m
14
Path Phase changes when the signal frequency changes – Path a Path shift when the signal path changes – Compare Path a with Path b Multipath & Frequency Diversity
15
Assumptions – transmitters and receivers are free to dynamically adjust their frequency in run time – transmitter and receiver are well coordinated and synchronized Environment variable – Number of propagation path : n – Number of channel : m Problem definition
16
System model i ∈ [1, n], j ∈ [1, m] – Path length : d i (d 1 is the LOS path) – Reflection coefficient : Γ i (Γ 1 = 1) – Wave length : λ j – For a fixed pair of tx & rx : Problem definition
17
For a given, we have the amplitude of the vector sum is a function : Problem definition
18
In run time, we will have a RSS measurement for each channels non-linear curvature fitting problem What we want is : d 1 Problem definition
19
Use Numerical Iteration to approximate this problem – J is the Jacobian matrix – H is the Hessian matrix This is an ill-conditioned matrix – Even when n = 1 and m = 2, the condition number of the Hessian matrix is Problem definition
20
Ill-conditioned Matrix – Condition number is very large – matrix of condition number = 100 will have 100% changes on the solution when the input changes 1% Problem definition
21
Ill-conditioned Matrix – In practical calculation: Problem definition
22
Treatments of problem Derivation of fixed Unknows – Hardware Specification Manuals – Chamber Training – Online Training
23
Treatments of problem Reducing the Model Path Number (n) – NLOS paths with three or more reflections or refractions can be ignored For typical materials, Γ is around 0.5 or less Three or more reflections on a single NLOS path will consume more than 87.5% of the energy – Ignore the paths which are extremely long, say more than two times of the LOS path energy will fade inversely proportional to the square of the path length 2 times => 0.25 × 0.5 = 0.125
24
Treatments of problem Bounding the Unknowns initial value d 1 – Newton’s method is quite effective when the initial setting is close to the solution enough – Consider the limited mobility in the indoor environments – Upper bound and lower bound old real d 1 - 1 ≤ Initial d 1 ≤ old real d 1 old real d 1 ≤ Initial d i ≤ 2 * old real d 1
25
Experiment MuD System – The tracking target wears a transmitter – 3 anchor nodes as receivers connected to server – Server solve the minimize problem, apply the trilateration algorithm, and display the location – 20 x 20 m 2 – Transmission power : -10 dbm – 16 channels – Transmitter stay in one channel for 50 ms
26
Experiment
28
Evaluation The Impact of Assumed Multipath Number – n = 5 outperforms the other settings with over 65% of ranging having the error less than 20%
29
Evaluation The Impact of c – chamber approach performs the best accuracy – online training approach exhibits a similar result
30
Evaluation The Impact of Used Channel Numbers – we suggest more channels when they are available and when latency and measurement overhead are not the concern
31
Evaluation The Impact of Initial Value Setting in MuD – Only when the initial value is set in a reasonable range, a reasonable solutions is expected
32
Evaluation Accuracy – The improvement with distinguishing multipath is up to 10 times
33
Evaluation Latency of MuD Tracking System – 16 channels – Transmitter stay in one channel for 50 ms – The channel switching time is about 0.34 ms
34
Evaluation Comparison with Static Environment – They both have averaged error about 1 meter – The difference between them is less than 5%
35
Conclusion We propose to exploit the frequency diversity of the radio propagation paths to mine the phase information of the radio paths and identify the signal amplitude of the LOS path Implement a real-time tracking system Experimental results show that the ranging and localization errors are 1m in average in a 20m×20m laboratory
36
Conclusion Future work – study the scenarios when LOS path does not exist Q&A
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.