Optimization and Parallelization Methods for the Design of Next-Generation Radio Networks

Lucas Benedicic
Jozef Stefan International Postgraduate School, Ljubljana, Slovenia
Jozef Stefan International Postgraduate School, 2014


   author={Benedi{v{c}}i{v{c}}, Lucas},




The complexity of the design of radio networks has grown with the adoption of modern standards. Therefore, the role of the computer for the faster delivery of accurate results has become increasingly important. In this thesis, novel methods for the planning and automatic optimization of radio networks are developed and discussed. The state-of-the-art metaheuristic algorithms, which compare a large number of different network configurations, rely on model-based simulations for the evaluation of the solution quality and the exploration of the search space. However, current radio-network solutions, based on snapshot simulations, have major weaknesses with respect to the simulation time and flexibility provided. In particular, the size of networks that can be analyzed in a feasible time is typically very limited. The new unified framework developed in this thesis significantly outperforms the currently available solutions for snapshot-based, radio-network simulations. It brings together novel and state-of-the-art parallelization methods, in order to allow for a detailed analysis of very large networks within an acceptable amount of time for everyday planning. This is achieved by the parallel features of the framework, which are exploitable on a single multi-core CPU, as well as on a network of standard PCs with GPU devices. Clearly, the significant speedup achieved at the simulation stage allows for an increased level of detail of the simulations, which improves the accuracy of the results. Increasing the performance of the simulations involved during the objective-function evaluation is only the first step towards a practical running-time reduction for radio-network optimization. In addition to this, also the optimization algorithms have to be improved in terms of speed, but not at the expense of the quality of results. In this sense, a novel agent-based algorithm is presented and tailored to a classic optimization problem in radio networks. The algorithm, which is based on techniques of cellular automata and population-based metaheuristics, shows considerable gains with respect to the size of problem instances it may handle, as well as regarding its speed performance and solution quality. The proposed unified framework is tested on complex optimization problems, namely (i) the problem of soft-handover balancing in third-generation systems, and (ii) the parameter optimization of empirical radio-propagation models. Most mobile operators are aware of the soft-handover balancing problem, but so far, and due to its complexity, it has not yet been tackled by any modern optimization approach. This thesis identifies and formally defines the mentioned problem. Using a black-box approach, different metaheuristic algorithms are employed for solving the problem, the solutions of which show a substantial improvement of downlink and uplink balance. Another use case of the presented methods is to optimize the parameters of empirical radio-propagation models. Until now, only one parameter set was used to adapt a radio-propagation model to a complete radio network. Using the proposed design automation, the model parameters can be adjusted locally, e.g., for each cell or region in the network, and thus greatly improve the accuracy of the calculated predictions. On the one hand, the proposed approaches allow a more detailed analysis of radio networks within a reasonable time. On the other hand, the optimization of much larger radio networks is also possible.
VN:F [1.9.22_1171]
Rating: 0.0/5 (0 votes cast)

* * *

* * *

TwitterAPIExchange Object
    [oauth_access_token:TwitterAPIExchange:private] => 301967669-yDz6MrfyJFFsH1DVvrw5Xb9phx2d0DSOFuLehBGh
    [oauth_access_token_secret:TwitterAPIExchange:private] => o29ji3VLVmB6jASMqY8G7QZDCrdFmoTvCDNNUlb7s
    [consumer_key:TwitterAPIExchange:private] => TdQb63pho0ak9VevwMWpEgXAE
    [consumer_secret:TwitterAPIExchange:private] => Uq4rWz7nUnH1y6ab6uQ9xMk0KLcDrmckneEMdlq6G5E0jlQCFx
    [postfields:TwitterAPIExchange:private] => 
    [getfield:TwitterAPIExchange:private] => ?cursor=-1&screen_name=hgpu&skip_status=true&include_user_entities=false
    [oauth:protected] => Array
            [oauth_consumer_key] => TdQb63pho0ak9VevwMWpEgXAE
            [oauth_nonce] => 1477080532
            [oauth_signature_method] => HMAC-SHA1
            [oauth_token] => 301967669-yDz6MrfyJFFsH1DVvrw5Xb9phx2d0DSOFuLehBGh
            [oauth_timestamp] => 1477080532
            [oauth_version] => 1.0
            [cursor] => -1
            [screen_name] => hgpu
            [skip_status] => true
            [include_user_entities] => false
            [oauth_signature] => S86H37AxRysidw9SkDbQcfQJW8s=

    [url] => https://api.twitter.com/1.1/users/show.json
Follow us on Facebook

HGPU group © 2010-2016 hgpu.org

All rights belong to the respective authors

Contact us: