Beruflich Dokumente
Kultur Dokumente
com nets
com nets
Overview
Motivation Locally optimal handover algorithm Performance analysis of hard handoff algorithm Adaptive hard handoff algorithm Locally optimal soft handoff algorithm Conclusion
Universitt Bremen
Kommunikationsnetze
com nets
com nets
Motivation
Handover design for cellular communication networks based on signal strength measurements Find an algorithm that is optimal for handover decisions
Analytical approach, not useful in practice Optimal trade-off between system parameters (number of service failures and number of handovers)
Use a locally optimal handover algorithm (practical approach) Adapting the handover algorithm to changing environments
Universitt Bremen
Kommunikationsnetze
com nets
com nets
Mobile is moving between neighbouring cells Focus on handover algorithms based on signal-strength measurements The only reliable measurement that can be used A possible mobile trajectory [VK97] Other measurements (BER, PER) are functions of the signal strength Limitations Inter cell handovers (horizontal handovers) 2 parameters (call quality, number of handovers) 1 interface (cellular)
Universitt Bremen
Kommunikationsnetze
com nets
com nets
Only the first two criteria are considered in the study Conflict of criteria
Keeping the call quality Reducing the number of handovers
Universitt Bremen
Kommunikationsnetze
com nets
com nets
Signal strength hystersis can be used for handover decision Hyseresis effects the trade-off between call quality (signal strength) and number of handovers A hysteresis level defines when the signal quality is sufficient to force the handover
Drawback U k =0 Unnecessary handovers in areas where signals from both stations are strong Definition of the hysteresis height
Universitt Bremen
Kommunikationsnetze
U k =1
hysteresis level h
com nets
com nets
Measurement of the call quality (signal strength) Introducing a threshold that is called service failure (SF)
SNR> SF -> call quality is sufficient SNR< SF -> call quality is stale (handover should be forced)
Best trade-off between expected number of service failures and expected number of handovers Solution of the optimal algorithm leads to a Bayes formulation It will lead to a complicated and non-stationary solution It requires prior knowledge of the mobiles trajectory (location problem) A locally optimal algorithm can be derived from that
Universitt Bremen
Kommunikationsnetze
com nets
com nets
Bayes Formula
Classical probability
P( B) 0
com nets
com nets
Bayes Framework
Bayesian paradigm supplies a direct solution for non standard problems for which there is no natural classical approach A Bayesian defines a model, selects a prior, collects data, computes the posterior, and then makes predictions Various types of Bayesan analyses, standard, informative, hierarchical, continuing, and parametric Two Bayes operations
The construction of a prior distribution Updating of the prior distribution to form the posterior distribution
Universitt Bremen
Kommunikationsnetze
com nets
com nets
Two base stations B(1) and B(2) involved in the handover Distance d(i) between the mobile and the base station B(i), where i= 1,2 Measured signal strength has three components
Path loss Large-scale fluctuations (shadow fading) Small-scale fluctuations (multipath fading)
Multipath fading is not considered (average, low-pass filter) to avoid unnecessary handovers No termination (hang-up) probability is considered
Universitt Bremen
Kommunikationsnetze
com nets
com nets
Pk(i ) = i i log d k( i )
i = 1,2
k = t / ts
Signal strength of adjacent APs [PV98] Lognormal fading model Isotropic correlation of the shadow fading
E[ Z Z
Universitt Bremen
Kommunikationsnetze
(i ) k
(i ) k +m
i = i 2 ai m ] = i exp d0
2
Locally Optimal Handover Algorithms for Wireless Communications
i = m v ts
v = const.
Bremen, 23.06.05 Dipl.-Ing. Stefan Aust
com nets
com nets
Handover decision
U k = k (Ik )
{
K =1 n
Handovers
Universitt Bremen
Kommunikationsnetze
E[ N H ] = P{ k = 1} U
K =1
Locally Optimal Handover Algorithms for Wireless Communications Bremen, 23.06.05 Dipl.-Ing. Stefan Aust
com nets
com nets
Optimal policy is a set of all decision functions ,which provides the best trade-off E[NSF] and E[NH]
= (1 + 2 + ... + n 1 )
min c E[ N H ] + E[ N SF ]
Optimum trade-off curve (E[NSF] and E[NH]) can be found by solving the Bayes problem for various values of c Parameter c can be interpreted as the relative cost of handovers versus service failures Parameter c can be adapted to changing environments
Universitt Bremen
Kommunikationsnetze
com nets
com nets
PX
c ( Bn1 ) n
<| X
c ( Bn1 ) n 1
U n1=0
U n1=1
Equation can be solved numerically to find the optimal handover policy using dynamic programming (DP) Drawbacks Quite complex Non-stationary Need for prior knowledge of the trajectory of the mobile Theoretical benchmark in comparison of suboptimal algorithms
Universitt Bremen
Kommunikationsnetze
com nets
com nets
PX
c ( Bk ) k +1
U k =0
U k =1
com nets
com nets
Hysteresis algorithm Implicates unnecessary and bad handovers Locally optimal algorithm Avoids unnecessary and bad handovers (rare case) Hysteresis-threshold algorithm Avoids unnecessary handovers (but needs two parameters)
Universitt Bremen
Kommunikationsnetze
com nets
com nets
Performance Comparison
Numerical results Simulation parameters D= 2000m = 105 dBm = 30 = 0 dBm Hysteresis algorithm has the highest service failure Locally optimal algorithm Numerical results [VK97] and hysteresis-theshold have the same characteristic Hysteresis value should be small (0,1,2 dB)
Universitt Bremen
Kommunikationsnetze
com nets
com nets
Same simulation parameters Performance of optimal, locally optimal and hystersis Sampling distances 2, 5 and 10 m Sampling rate 0.5s Speed 14.4, 36 and 72 Km/h Simulation results [PV98] Optimal (dashed) and local optimal (solid) outperform hysteresis (dotted)
Universitt Bremen
Kommunikationsnetze
com nets
com nets
J ( ) = SF ( ) + H ( ) >0
Minimize the cost function
LD + = 0 H
Optimum decision
c
com nets
com nets
Estimate the active set of base stations (CDMA) Primary objective of SHO is to provide a good signal quality Design is similar to hard handover Forward/reverse channel Static policy ( ) = ( ADD _ TH , DROP _ TH )
com nets
com nets
Conclusion
Optimal handover algorithm can be determined by numerical results (complicated and non-stationary) Locally optimal handover algorithm as a practical approximation to the optimal algorithm Minimization of the Bayes formulation where parameter c is the trade-off parameter Trade-off parameter c can be used for adapting the handover algorithm to changing environments Fixing c gives operating points with nearly same gradient for different velocities Locally optimal algorithm is also suitable for soft handover decisions
Universitt Bremen
Kommunikationsnetze
com nets
com nets
Bibliography
[VK97] V. V. Veeravalli, O. E. Kelly, A Locally Optimal Handoff Algorithm for Cellular Communications, IEEE Transactions on Vehicular Technology, Vol. 46, No. 3, August, 1997 [PV98] R. Prakash, V. V. Veeravalli, Accurate Performance Analysis of Hard Handoff Algorithms, Proceedings of PIMRC 1998, Boston, September, 1998 [PV00] R. Prakash, V. V. Veeravalli, Adaptive Hard Handoff Algorithms, IEEE Journal on Selected Areas in Communications, vol. 18, no. 11, November, 2000 [PV03] R. Prakash, V. V. Veeravalli, Locally Optimal Soft Handoff Algorithms, IEEE Transactions on Vehicular Technology, vol. 52, no. 2, March 2003
Universitt Bremen
Kommunikationsnetze
com nets