Mimo Channel Modeling using a Class of Soft-Computational Techniques
No Thumbnail Available
The key challenge faced by up-coming wireless communication systems is to provide high-data-rate wireless access at better quality of service (QoS). In such a situation, Multiple-Input Multiple-Output (MIMO) wireless technology seems to be able to meet these demands by offering increased spectral efficiency. A very common form of uncer- tainty and stochastic behaviour is observed in MIMO wireless communication due to interference and correlation among channel coefficients which makes channel estimation a challenging area. There are several statistical methods of channel estimation that have provided satisfactory performance while modeling the MIMO wireless systems. Soft- computational approaches are recent additions to the list of channel estimation methods of which most of the works have primarily focused on the training-learning aspects of ANN, fuzzy systems etc. Till now, no recorded efforts have been observed regarding expansion of the abilities of such architectures beyond the training-testing realm which includes certain architectural challenges. These challenges include: (i) incorporating tem- poral behaviour in the Multi Layer Perceptron (MLP)- a feedforward ANN enabling it to track time-variations in the input signal, (ii) ensuring stability to the system by append- ing a feedback path along with the usual feedforward structure of the ANN, (iii) retaining only the contextual portion of the information with the above structure, (iv) properly capturing the fast time-varying nature of the channels, (v) combining ANN and fuzzy based systems to obtain the capability of expert-level decision making while modeling uncertainty observed in the MIMO channel and (vi) realization of a suitable system with lower implementation and time complexity. Taking these challenges into consideration, a class of soft-computational tools based on ANN in feedforward layout called MLP and feedback form called Recurrent Neural Network (RNN) and fuzzy-based composite sys- tems are explored with stress on architectural expansion so as to improve performance and precision than conventional methods while modeling the stochastic nature of the MIMO channels. Among the proposed techniques of this thesis, a fuzzy based approach, named Fuzzified Time Delay Fully RNN (FTDFRNN) establishes its superiority while modeling even a deeply faded MIMO channel in important practical applications like VOIP based transmissions. In such a framework of MIMO communication, the fuzzy based approach turns out to be the most suitable one for adaptive receiver designs optimized for high data rate wireless communication.
Supervisor: A. Mitra
ELECTRONICS AND ELECTRICAL ENGINEERING