2.3 Index System Method
Index system method is a three-step process. Firstly, the index system of satellite
communication system is established by seeking for experts’ opinions. The second step
is to quantify indexes according to indexes’ characteristics. Finally, the performance is
evaluated by weighted sum of quantitative indexes. Index system method could provide
a quantitative result of the overall performance of satellite communication system.
In [10], index system construction of satellite communication system was based on
four domains: physical domain, information domain, recognition domain and social
domain. Three principles which were needed to be followed to determine indexes were
proposed in [11]. Satellite performance indexes influenced each other. Index system
was supposed to be constructed from the holistic perspective. Plus, indexes needed to
be layered and took account of actual satellite communication scenario. Liu et al. [12]
divided indexes into coverage, transmission, processing and protection according to
satellite communication process, and then quantified indexes through numerical calculation and experts’ scores. Index weights were calculated by experts’ scores, and
performance was evaluated by indexes’ weighted sum.
Index system method is greatly influenced by experts’ subjective judgment. For
instance, index weights are determined by experts’ experience. However, satellite
communication system has various application scenarios. Different scenarios have
different index weights. Scenario design would use up much labor power, material
resources and financial capacity, and index weights could not be adjusted automatically. Furthermore, once a certain real application scenario is not considered in
advance, performance evaluation targeting this scenario would be omitted.
3 Application Prospect of Neural Network in Performance
Evaluation of Satellite Communication System
Using neural network to evaluate the performance of satellite communication system
has broad prospects, such as identifying key parameters, adjusting evaluation model
adaptively and comparing different satellite constellations’ performance.
3.1 Identifying Key Parameters
In the future, a satellite communication constellation would commonly contain more
than 1000 satellites. Furthermore, a satellite communication constellation would consist
of various orbits’ satellites, inter-satellite communication links would also exist in large
numbers, the number of mobile satellite communication terminals would increase
dramatically. As a result, evaluation parameters will grow in magnitude, the interaction
between evaluated parameters will grow more and more complex, thus existing evaluation methods are difficult to identify key parameters.
Neural network method can firstly set initial weights of evaluation parameters, and
then train the neural network to correct these weights through the actual communication data. The parameters that could get larger weights can be considered as key
parameters, which are less affected by the subjective experience of experts.
242
S. Cheng et al.
Index system method is a three-step process. Firstly, the index system of satellite
communication system is established by seeking for experts’ opinions. The second step
is to quantify indexes according to indexes’ characteristics. Finally, the performance is
evaluated by weighted sum of quantitative indexes. Index system method could provide
a quantitative result of the overall performance of satellite communication system.
In [10], index system construction of satellite communication system was based on
four domains: physical domain, information domain, recognition domain and social
domain. Three principles which were needed to be followed to determine indexes were
proposed in [11]. Satellite performance indexes influenced each other. Index system
was supposed to be constructed from the holistic perspective. Plus, indexes needed to
be layered and took account of actual satellite communication scenario. Liu et al. [12]
divided indexes into coverage, transmission, processing and protection according to
satellite communication process, and then quantified indexes through numerical calculation and experts’ scores. Index weights were calculated by experts’ scores, and
performance was evaluated by indexes’ weighted sum.
Index system method is greatly influenced by experts’ subjective judgment. For
instance, index weights are determined by experts’ experience. However, satellite
communication system has various application scenarios. Different scenarios have
different index weights. Scenario design would use up much labor power, material
resources and financial capacity, and index weights could not be adjusted automatically. Furthermore, once a certain real application scenario is not considered in
advance, performance evaluation targeting this scenario would be omitted.
3 Application Prospect of Neural Network in Performance
Evaluation of Satellite Communication System
Using neural network to evaluate the performance of satellite communication system
has broad prospects, such as identifying key parameters, adjusting evaluation model
adaptively and comparing different satellite constellations’ performance.
3.1 Identifying Key Parameters
In the future, a satellite communication constellation would commonly contain more
than 1000 satellites. Furthermore, a satellite communication constellation would consist
of various orbits’ satellites, inter-satellite communication links would also exist in large
numbers, the number of mobile satellite communication terminals would increase
dramatically. As a result, evaluation parameters will grow in magnitude, the interaction
between evaluated parameters will grow more and more complex, thus existing evaluation methods are difficult to identify key parameters.
Neural network method can firstly set initial weights of evaluation parameters, and
then train the neural network to correct these weights through the actual communication data. The parameters that could get larger weights can be considered as key
parameters, which are less affected by the subjective experience of experts.
242
S. Cheng et al.