Dynamic Channel Allocation Using Hopfield Networks

Dynamic Channel Allocation for Mobile Communication By: Kr. Rajeev Ranjan 07438 Electronics and Comm. Engg. NIT Hamirpur (HP)

Introduction to Wireless Systems

Cellular Concept

Cellular Concept Cell Base Station (BS) or BTS Mobile Station (MS)

Cellular Concept Forward Link

Reverse Link

MS

BTS

GSM System Architecture

Channel Allocation

Channel Allocation

B A

Interfering Cell Non Interfering Cell

Interference Order

C

Channel Allocation

Channel Allocation Fixed Channel Allocation (FCA) Dynamic Channel Allocation (DCA)

Hard Condition Soft Condition

Dynamic Chanel Allocation Technique Why DCA ? • FCA systems allocate specific channels to specific cells. • Uneven Call Traffic -- available channels are not being used efficiently in FCA.

• In DCA systems, no set relationship exists between channels and cells.

Dynamic Chanel Allocation Technique Energy Function

Dynamic Chanel Allocation Technique Conventional Approach • Channel Allocation algorithm, rearranges the assignment channels in the whole network every time a new call arrives. • Complex Algorithms for deciding which available channel is most efficient. • Algorithms computationally intensive. • Requires large computing resources (Time).

These drawbacks lead us to use of Hopfield Networks

Hopfield Neural Networks Characteristics •Association or Classification •Optimization •Restoration of Patterns

Mapping Network

Objective: Assign a channel to the cell effectively along with the decrease in computation cost.

Hopfield Neural Networks Architecture

Mapping the problem into HNN Input • Call arrival or the termination in a particular cell.

Output • Assignment of channel available in the network.

Association or Mapping Network is used here in order to allocate the channel

Solution of HNN Output of Hopfield Network is given by

Energy Function is given by E(x)= - ½ Xt W X - It X + Q X

Solution of HNN Energy Function

where x

b W

input vector (channel assignments) for which an optimal solution is sought; bias vector determined by constraints; symmetric weight matrix for the neural network.

• Weight Update: • Bias Vector Update:

Simulation : Simulation carried out •over a portion of 7x7 Hexagonal Cell •2 rings of interfering cell •70 channels

To evaluate the performance: • Blocking Probability --used as a performance Index . Compared this with the existing systems of Dynamic channel allocation techniques.

Simulation Result:

Simulation Result:

Conclusion: •The HNN is an effective channel allocation technique with lower the computational costs of the algorithm (Time).

•A valuable feature of the HNN allocation technique is its insensitiveness to local faults -- implies loss of elaborative capacity of some neurons doesn’t mean that system will be blocked.

References: • Principles of Wireless Communication : By Theoder S. Rappaport. • Hopfield Neural Networks for DCA in Mobile Communication by Enrico DEL RE, Romano FANTACCI, Luca RONGA, Giovanni GIAMBENE.

IEEE Volume 3, Page(s):1664 - 1668 . • Hopfield Networks for Finding Global Optimum by Xiaofei Huang. (Proceedings of International Joint Conference on Neural Networks, Montreal, Canada, July 31 - August 4, 2005).