In wireless multimedia sensor networks (WMSNs), N multimedia sensor nodes were deployed. Each node is randomly moved. A data transmission path was established between WMSNs and cloud platform. The routing request can be initiated by the cloud or any node, so as to establish a wireless transmission path between the nodes and the cloud, which was composed of M nodes. At this point, WMSNs have the following three characteristics:
-
(1)
The multimedia data transmission or node status has the equilibrium point. When the multimedia data is transferred, the WMSN gradually tends to balance with the nodes of the nodes, and it reaches a steady state.
-
(2)
There is the direct relationship between the speed and state of the WMSN equilibrium point and the state of the mobile node.
-
(3)
Data transfer between the nodes and the received data from the cloud can be effectively obtained by the mobile node’s collaboration.
According to the system characteristics, the WMSN can be defined as a cooperative neural network, as shown in Fig. 1. In WMSNs, the heptagonal cooperative neural networks are composed of data sending node and neighbor nodes as shown in Fig. 1, to provide reliable data transmission service. Data fusion after convergence evacuation was sent to the cloud. Users can obtain the multimedia data through the clouds. In the WMSNs, the mobile node’s 3D moving trajectory is shown in Fig. 2, which is divided into the time component, the moving vector x and the moving vector y. The whole system of WMSNs can be obtained by calculating the 3D vector product of N mobile node.
Therefore, the characteristics of the cooperative neural network are expressed as the formula (1a).
$$ \left\{\begin{array}{l}n={\displaystyle \sum_{i=1}^N{S}_{ik},\begin{array}{cc}\hfill \hfill & \hfill i\ne k\hfill \end{array}}\\ {}w\left|{}_{k=1\to N}\right.=\left\{\begin{array}{l}1,n\ge {S}_{i\to \mathrm{cloud}}\\ {}0,n<{S}_{i\to \mathrm{cloud}}\end{array}\right.\\ {}{S}_{\mathrm{CNN}}={\displaystyle \sum_{i=1}^N{w}_i{S}_i}\end{array}\right. $$
(1a)
Here, S
ik
denotes the sending signal form i node to k node. n denotes the signal received by neighbor nodes after wireless broadcasting by any node. w denotes the cooperative weight. The node would join the cooperation when the signal of some neighbor node is larger than one between i node and cloud. S
CNN is the convergence signal after cooperation to reach the cloud.
Based on signal transmission, the 3D vector analysis can be implemented in the i node of the cooperative neural network, as shown in the formula (2a).
$$ \left\{\begin{array}{l}{t}_i=w\left|{}_{i-1\to N}\right.\sqrt{{x_{i-1}}^2+{y_{i-1}}^2}+w\left|{}_{i+1\to N}\right.\sqrt{\left|{x_{i+1}}^2-{y_{i+1}}^2\right|}\\ {}{x}_i={S}_{i\to \mathrm{cloud}}w\left|{}_{i-1\to N}\right.+\sqrt{\left|{x_{i+1}}^2-{y_{i+1}}^2\right|}w\left|{}_{i+1\to N}\right.\\ {}{y}_i=w\left|{}_{i-1\to N}\right.{S}_i{\displaystyle \sum_{j=1}^i\sqrt{\left|{x_{i+1}}^2-{y_{i+1}}^2\right|}}\end{array}\right. $$
(2a)
In WMSNs with 100 nodes, a data transmission network consisting of ten nodes was forecasted from sender to receiver using the cooperation-type neural network. The results are as shown in Fig. 3. Fig. 3a shows the actual node deployment plan. Fig. 3b is the forecast results. Fig. 3c gives the analysis results of prediction errors. The results showed that cooperative neural network prediction has the higher accuracy.
Although the cooperative neural network can accurately predict the nodes in a data transmission network topology, there is a poor performance for the future state of the whole network. Therefore, combining with the opportunistic Markov chain model as shown in Fig. 4, state transition prediction model of WMSN topology dynamic was proposed according to the characteristics of dynamic Markov model, where the three transmission path multimedia is established between data sending node sender and receiving cloud.
The probability is determined by probabilistic chance and opportunity dimensions probability with full connection from the sender to the receiver. According to the three-dimensional prediction results of the cooperative neural network, an opportunity-type Markov chain model was defined. The 1D Markov chain model is applied to only one dimension to change, the success arrive probability of the cloud is P1Px. A two-dimensional Markov chain model is applied to only two dimensions to change, and the success arrive probability of the cloud is P1PxPy. The three-dimensional Markov chain model is suitable to three-dimensional transfer, the success arrive probability is P1PxPyPt.
In the Markov chain model, the opportunistic probability sum of the three-dimensional transmission path is 1, as shown in formula (1). The end-to-end success rate of sending packet Ps can be obtained by formula (2), as well as the outage probability P
OUT of transmission path can be calculated by formula (3).
$$ {P}_1+{P}_2+{P}_3=1 $$
(1)
$$ {P}_S=\left\{\begin{array}{l}{P}_1{\displaystyle \prod_{i=1}^a{P}_x}\\ {}{P}_2{\displaystyle \prod_{i=1}^a{\displaystyle \prod_{j=1}^b{P}_x{P}_y}}\\ {}{P}_3{\displaystyle \prod_{i=1}^a{\displaystyle \prod_{j=1}^b{\displaystyle \prod_{k=1}^c{P}_x{P}_y{P}_t}}}\end{array}\right. $$
(2)
$$ {P}_{\mathrm{OUT}}=\left\{\begin{array}{l}{P}_1{\displaystyle \prod_{i=1}^a\left(1-{P}_x\right)}\\ {}{P}_2{\displaystyle \prod_{i=1}^a{\displaystyle \prod_{j=1}^b\left(1-{P}_x\right)\left(1-{P}_y\right)}}\\ {}{P}_3{\displaystyle \prod_{i=1}^a{\displaystyle \prod_{j=1}^b{\displaystyle \prod_{k=1}^c\left(1-{P}_x\right)\left(1-{P}_y\right)\left(1-{P}_t\right)}}}\end{array}\right. $$
(3)
The cooperative neural network will lead to WMSNs gradually reach a state of equilibrium. Based on the opportunistic Markov chain model, the equilibrium state probability P
balance can be obtained as formula (4).
$$ {P}_{\mathrm{balance}}={P}_S>{P}_{\mathrm{OUT}}\left|{}_{t\to \infty}\right. $$
(4)
A single mobile node status in WMSNs included the mobile speed V
E, topology robustness R
E, and cloud communication stability S
E. Therefore, the mobile node working state can be defined as {V
E, R
E, S
E}. The V
E could be calculated through cooperative neural network of two different direction vector component predictions. R
E could be calculated through the connectivity detection using opportunistic Markov chain model as shown in the formula (5). S
E could be calculated by detecting the cloud success probability using Markov chain model as shown in Eq. (6).
$$ {R}_E={\displaystyle \prod_{t=1}^{\infty }{\displaystyle \sum_{i=1}^n\sqrt{\left|{\left({S}_{i\to i+1}w\left|{}_t\right.\right)}^2-{\displaystyle \sum_{j=1}^i\sqrt{\left|{x_{i-1}}^2-{y_{i-1}}^2\right|}{S}_i}\right|}}} $$
(5)
$$ {S}_E={P}_{\mathrm{balance}}\left|{}_{\mathrm{cloud}}\right.{\left[\begin{array}{cc}\hfill \begin{array}{cc}\hfill {V}_{E1}\hfill & \hfill \cdots \hfill \end{array}\hfill & \hfill {V}_{Ea}\hfill \\ {}\hfill \begin{array}{cc}\hfill {V}_{E1}\hfill & \hfill \cdots \hfill \end{array}\hfill & \hfill {V}_{Eb}\hfill \end{array}\right]}_{a*b}{{\left[\begin{array}{c}\hfill {R}_{E1}\hfill \\ {}\hfill {R}_{Ec}\hfill \end{array}\right]}_{c*1}}^T $$
(6)