??? Attractive forces based on maximizing communication capacity between
nodes: Fcomm = ??’aU?¤
ar , where U??— is the optimal value of the communication
utility function. For each MWSN node, the gradient is computed
along each separate data route going through the node, and the resultant
force is computed by superposition as in [39].
??? Attractive ???restoring??? forces based on penalties for exceeding the maximum
allowable communication distance between two nodes i and j:
Frestore(i, j) = uij(rj ??’ ri).
??? Information forces that depend on the gradient of the uncertainty measure,
Finf = ??’a||PT
k (r)Pk(r)|| ar , where Pk(r) is the uncertainty measure for
sampling calculated via the closed form or the Kalman Filter.
After the force calculation, the equation of motion for the i-th MWSN
node will be given by integrating a simple mass-damper model over time:
mi??ri + vi ?™ ri = Fi, where m and v are mass and damping terms respectively.
These coe?±cients have no physical meaning, but they are used to define a
dynamic equation of motion for the system in the direction of the gradient of a
global potential field function consisting of obstacles, goal locations, network
utility, and information gain. At each adaptive sampling step, the motion
of the robots stops when a minimum potential configuration is reached. A
summary of the potential field deployment algorithm is shown below:
Potential Field Sensor Node Repositioning Algorithm (POT-1)
Assumptions: The bandwidth between two nodes in the network is
governed by a known path-loss model.
Pages:
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120