AN ovel concept behind cellular and personal communication systems is the frequency reuse principle. Under this principle, various channel assignment (CA) algorithms are formulated for spectral management. A good CA algorithm is required to maximize through-put performance while link quality is guaranteed to satisfy some minimum requirement. As the cell sizes are shrinking down to the order of few hundred meters, CA algorithms for future personal communication systems (PCS) may have good chance to encounter rather drastic changes in traffic patterns, both temporally and spatially. Clearly, conventional fixed channel assignment (FCA) schemes, whose performance is at best when traffic pattern is statistically static, will not function well under such circumstances. Indeed, Dynamic Channel...[ Read more ]

AN ovel concept behind cellular and personal communication systems is the frequency reuse principle. Under this principle, various channel assignment (CA) algorithms are formulated for spectral management. A good CA algorithm is required to maximize through-put performance while link quality is guaranteed to satisfy some minimum requirement. As the cell sizes are shrinking down to the order of few hundred meters, CA algorithms for future personal communication systems (PCS) may have good chance to encounter rather drastic changes in traffic patterns, both temporally and spatially. Clearly, conventional fixed channel assignment (FCA) schemes, whose performance is at best when traffic pattern is statistically static, will not function well under such circumstances. Indeed, Dynamic Channel Assignment (DCA) schemes with markov allocation, which assign channels in a call by call basis and whose assignment schemes follow changes of traffic patterns, have been demonstrated to outperform FCA schemes in throughput performance in the scenario of dynamic traffic patterns. Like FCA schemes, CA decision making in conventional DCA schemes is centralized. This can be a major factor to hinder DCA schemes be useful for PCS as the complexity of DCA schemes can become prohibitively high when sizes of PCS are large. More recently, distributed DCA (DDCA)schemes, where CA decision making is distributed at every cell site, have been proposed. Clearly, the complexity of DDCA algorithms is invariant to PCS sizes. Since every base station is capable in making CA decision, DDCA algorithms can adapt to dynamic traffic, and can resolve cellular traffic congestion in a timely manner. There have been two lines of research in the formulation of DDCA schemes. One with the use of on-line measurements and another with pre evaluation of compatibility constraints, making use of statistical assessment of propagational environment. In the case of on-line meaurements, the way to incorporate efficient channel search strategies utilised in conventional CA schemes is yet unknown. On the other hand, the schemes with pre evaluated compatibility constraints utilise values of constraints, which are worst case parameters, and hence there is a, tendency to waste the precise spectral resources.

In the project presented in this thesis, an alternative to the fixed compatibility constraint is investigated. In this approach, compatibility constraints are adaptive to the temporal and spatial variations of traffic. Such adaptivity is achieved with the use of "Fuzzy Theory". Depending on the prevailing traffic conditions, the new class of algorithms observe an aggressive policy where the worst case compatibility constraints are violated to increase the throughput. No channel rearrangements are assumed. The computational and implemen-tational complexities of this class of algorithms are more or less similar to the conventional DCA algorithms, but are able to achieve incredibly higher throughput compared to such convetional algorithms. Though there is a trade off with the quality in terms of the probability of bad links, the grade of service (a function of througput and quality) is found to be better than any conventional DCA algorithms executed without channel rearrangements. The formulations assume a distributed implementation. Each of the base stations maintain a local database of small sizes (information with regard to the channel occupancy patterns, at most for 18 neighbouring cells and the host cell). Each base stat ion can carry out channel assignment autonomously and information exchange is only with the set of neighbour base stations causing significant interference. The new class of algorithms is termed Aggressive Fuzzy Distributed Dynamic Channel Assignment (AFDDCA) algorithms. Several varients are formulated and studied via simulations.