Loading...
 
[Show/Hide Left Column]

BSM 

Bayesian statistical methods for joint user activity detection, channel estimation, and data decoding in dynamic wireless networks


Axe : ComEx - Intelligent Network Structures
Sujet : Bayesian statistical methods for joint user activity detection, channel estimation, and data decoding in dynamic wireless networks
Directeurs de thèse : Frédéric Lehmann, SAMOVAR et Antoine O. Berthet, L2S
Institution : TelecomSud Paris
Laboratoire gestionnaire : SAMOVAR
Doctorant : Fakher Sagheer
Début : 20169
Productions scientifiques :
Ressources :

Contexte :
In today’s wireless communications networks, increasing the spectral efficiency and managing interference are two of the fundamental design issues. Relaxing the orthogonality constraint at the transmitter side as advised by network information theory, i.e., considering non-orthogonal multiple access (NOMA) 1, is one of the diverse solutions proposed in 5G (and beyond) to solve the first issue. This goes along with advanced signal processing techniques at the receiver side to cancel interference and hence cope with the second issue. The problem becomes even more complicated in random-access based wireless systems characterized by no or very limited coordination compared to cellular systems, and often imperfect distributed mechanisms to establish communication. In such dynamic environments, the number of active users, their location, as well as the identities and parameters that specify their modulation coding schemes and channel state, vary with time and has considerable impact on receiver’s performance. Examples of applications in communications theory have to be found in multiuser detection (MUD), spatial multiplexing schemes, or ad hoc networks. The problem of detecting the number of users in a multiuser system was addressed in the context of code-division multiple-access (CDMA) in the 2000’s 22, 23 where it was recognized that the simplifying assumption that all users were active at all time was, in general, a cause
of performance degradation and sub-optimality. Some nonlinear receivers based on successive interference cancellation need to know the strongest user to better fight back the so-called near-far effect. Moreover, identifying the active users helps the system to promptly process requests and efficiently allocate resources leading to system capacity improvements. In spatial multiplexing schemes, the total system capacity can be increased by properly selecting a subset of active users to which the power is allocated. In ad hoc networks,
optimum transmission strategies require the identification and localization of actives nodes in the neighborhood of the transmitter 2. Timely applications related to 5G and beyond include grant free access in M2M/IoT 3, cognitive radio 4, cooperative half and/or full-duplex relaying with advanced decodeand-forward protocols (e.g., dynamic and/or selective decode-and-forward without dedicated feedforward control signals and network coding in case of multisource multirelay cooperating clusters), and hybrid half-uplex/full-duplex transmissions 5. To conclude, it is worth pointing out that the problem of active user identification and data detection has deep connections with multisource-multitarget estimation in radar theory 16.
Objectif scientifique:
One way to formalize the problem of user identification, channel estimation, and data detection is to express the discrete-time baseband equivalent received signal yl 2 Cd as
yl =
X
x(k)
l 2Xl
f

x(k)
l

+ nl, l = 1, . . . ,L (1) where:
  • l denotes a channel use (time or frequency resource index) and L the total number of channel uses to transmit a codeword;
  • Xl is the finite random set encapsulating what is unknown about the active users, defined over a hybrid discrete-continuous space;
  • x(k)
l denotes a multicomponent element in Xl, e.g., x(k)
l =

k, s(k)
l ,w(k)
l

, k 2 K being an active user
identification index, s(k)
l 2 M being a coded modulated symbol, and w(k)
l 2 Cd the combination of the signal signature and the channel impulse response;
  • f(.) is a deterministic mapping whose definition takes into account the known parameters about x(k) l ;
  • nl 2 Cd is a stationary random noise sample;
Based on this definition, four problems can be formulated:
Problem 1: Reliable communications over static synchronous multiaccess channel with an unknown number of active users and perfect channel state information at the receiver (CSIR). The number of active users and their identities do not change during the reference transmission duration. The channel parameters and responses of the active users are known. The number of active users, their identity, and their data
must be jointly estimated.
Problem 2: Reliable communications over static synchronous multiaccess channel with an unknown number of active users and no or imperfect CSIR. The number of active users and their identities do not change during the reference transmission duration. The channel parameters and responses of the active users are unknown or partially known, and may vary according to a dynamical model. The number of active users, their identity, their channel, and their data must be jointly estimated.
Problem 3: Reliable communications over dynamic asynchronous multiaccess channel with an unknown number of active users and perfect CSIR. The number of active users and their identities change during the reference transmission duration according to a dynamic model characterizing the users’s activity. The channel parameters and responses of the active users are known. The number of active users, their identity, and their data must be jointly estimated.
Problem 4: Reliable communications over dynamic asynchronous multiaccess channel with an unknown number of active users and no or imperfect CSIR. The number of active users and their identities change during the reference transmission duration according to a dynamic model characterizing the users’s activity. The channel parameters and responses of the active users are unknown or partially known, and may vary according to a dynamical model. The number of active users, their identity, their channel, and their data must be jointly estimated.