This dissertation concerns the development of a feedback control framework for coordinating multiple, sensor-equipped, autonomous vehicles into mobile sensing arrays to perform adaptive sampling of observed fields. The use of feedback is central; it ...
This dissertation concerns the development of a feedback control framework for coordinating multiple, sensor-equipped, autonomous vehicles into mobile sensing arrays to perform adaptive sampling of observed fields. The use of feedback is central; it maintains the array, i.e. regulates formation position, orientation, and shape, and directs the array to perform its sampling mission in response to measurements taken by each vehicle. Specifically, we address how to perform autonomous gradient tracking and feature detection in an unknown field such as temperature or salinity in the ocean.
Artificial potentials and virtual bodies are used to coordinate the autonomous vehicles, modelled as point masses (with unit mass). The virtual bodies consist of linked, moving reference points called virtual leaders. Artificial potentials couple the dynamics of the vehicles and the virtual bodies. The dynamics of the virtual body are then prescribed allowing the virtual body, and thus the vehicle group, to perform maneuvers that include translation, rotation and contraction/expansion, while ensuring that the formation error remains bounded. This methodology is called the Virtual Body and Artificial Potential (VBAP) methodology.
We then propose how to utilize these arrays to perform autonomous gradient climbing and front tracking in the presence of both correlated and uncorrelated noise. We implement various techniques for estimation of gradients (first-order and higher), including finite differencing, least squares error minimization, averaging, and Kalman filtering. Furthermore, we illustrate how the estimation error can be used to optimally choose the formation size.
To complement our theoretical work, we present an account of sea trials performed with a fleet of autonomous underwater gliders in Monterey Bay during the Autonomous Ocean Sampling Network (AOSN) II project in August 2003. During these trials, Slocum autonomous underwater gliders were coordinated into triangle formations, and various orientation schemes and inter-vehicle spacing sequences were explored. The VBAP methodology, modified for implementation on Slocum underwater gliders, was utilized. Various operational issues such as speed constraints, external currents, communication constraints, asynchronous surfacings and intermittent feedback were addressed.
The work contained in this thesis was conducted under the advisement of Naomi Ehrich Leonard at Princeton University.