As robots move in
the 3D sampling space and acquire sensory information, the uncertainty
in their localization will a?®ect the sensor maps generated. A combined
localization-sensor field estimation problem is formulated. Robot navigation
will be driven by uncertainty measures such as the norm of Kalman
filter covariance matrices. The algorithms described here combine the uncertainty
in localization as well as in the sensor measurements to achieve
e?®ective adaptive sampling. Both of these uncertainties are especially relevant
for underwater vehicles, since position estimates are often inaccurate
due to navigational errors from dead-reckoning [5].
??? Robot navigation for maximizing the amount of raw data streamed from
the sensor nodes. Communication rates in sensor nets deteriorate as the
37
have limited bandwidth constraints between vehicles that vary with location,
2 Problem Formulation
Dan O. Popa and Frank L. Lewis
Fig. 1. Schematic diagram of fundamental problems related to adaptive sampling.
receiver moves away from the transmitter. Obviously, mobile sensor nodes
will require continuously changing path lengths, leading to variable data
rates. In some applications, such as underwater robotics, stringent communication
constraints are made much worse by interference in a common
broadcast space. Network capacity is addressed in the context of congestion
control with variable link rates.
??? Robot navigation for minimizing the amount of energy expended through
motion and communication.
Pages:
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100