One of the important aspects of seamless communication for ubiquitous computing is the dynamic selection of the best access network for a multimodal device in a heterogeneous wireless environment. In this paper, we co...One of the important aspects of seamless communication for ubiquitous computing is the dynamic selection of the best access network for a multimodal device in a heterogeneous wireless environment. In this paper, we consider available bandwidth as a dynamic parameter to select the network in heterogeneous environments. A bootstrap approximation based technique is firstly utilized to estimate the available bandwidth and compare it with hidden Markov model based estimation to check its accuracy. It is then used for the selection of the best suitable network in the heterogeneous environment consisting of 2G and 3G standards based wireless networks. The proposed algorithm is implemented in temporal and spatial domains to check its robustness. The numerical results show that the proposed algorithm gives improved performance in terms of estimation error （less than 15%）, overhead （varies from 0.45% to 72.91%） and reliability （approx. 99%）as compared to the existing algorithm.展开更多
文摘One of the important aspects of seamless communication for ubiquitous computing is the dynamic selection of the best access network for a multimodal device in a heterogeneous wireless environment. In this paper, we consider available bandwidth as a dynamic parameter to select the network in heterogeneous environments. A bootstrap approximation based technique is firstly utilized to estimate the available bandwidth and compare it with hidden Markov model based estimation to check its accuracy. It is then used for the selection of the best suitable network in the heterogeneous environment consisting of 2G and 3G standards based wireless networks. The proposed algorithm is implemented in temporal and spatial domains to check its robustness. The numerical results show that the proposed algorithm gives improved performance in terms of estimation error （less than 15%）, overhead （varies from 0.45% to 72.91%） and reliability （approx. 99%）as compared to the existing algorithm.
